Why is thymidine incorporation (DPM) normalized to per mg protein content while testing for cell proliferation?

Why is thymidine incorporation (DPM) normalized to per mg protein content while testing for cell proliferation?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

Sorry for this naive question. After performing the thymidine incorporation assay to test proliferation, many papers report normalization of disintegrations per minute (DPM) to per mg protein content? What does this normalization or ratio of DPM to cellular protein content exactly signify?

Thanks in advance.

I'm not sure that there is a strongly disciplined reason for the mg protein content as the normalization, but there is obviously a need for some normalization for different amounts of tissue- if you use 10X as much tissue, you should expect DPM to be 10X as high!

For example, in this somewhat old paper, They normalize not to protein content but to wet tissue mass or to intestinal crypt count (as they are looking at incorporation in intestinal crypts). So the normalization that you observe is not universal (or more likely, is a more recent innovation). They observe that the cruder measure of wet mass provides a poor normalization:

Instead there was a positive intercept of dpm/mg tissue which represented about 25-30% of the maximal tritium content of the intestine. (abstract)

That is a good reason to use a more precise normalization! They find that the more precise normalization of dpm/crypt works better, presumably because the crypt is the incorporating tissue.

I would suspect that the dpm/mg protein is simply a more precise method for normalization.

A yet more precise method would be the one used in this 1987 paper, in which thymidine incorporation is estimated as dpm / microgram DNA- here, you are actually directly comparing to the mass of the DNA, which is where thymidine is incorporated. That is a more obvious normalization than protein content.

However, for crude tissue extracts I suspect that it is more straightforward to estimate protein content than DNA content. So protein content is used as a "good enough" normalization.

Mitochondrial hypoxic stress induces widespread RNA editing by APOBEC3G in natural killer cells

Protein recoding by RNA editing is required for normal health and evolutionary adaptation. However, de novo induction of RNA editing in response to environmental factors is an uncommon phenomenon. While APOBEC3A edits many mRNAs in monocytes and macrophages in response to hypoxia and interferons, the physiological significance of such editing is unclear.


Here, we show that the related cytidine deaminase, APOBEC3G, induces site-specific C-to-U RNA editing in natural killer cells, lymphoma cell lines, and, to a lesser extent, CD8-positive T cells upon cellular crowding and hypoxia. In contrast to expectations from its anti-HIV-1 function, the highest expression of APOBEC3G is shown to be in cytotoxic lymphocytes. RNA-seq analysis of natural killer cells subjected to cellular crowding and hypoxia reveals widespread C-to-U mRNA editing that is enriched for genes involved in mRNA translation and ribosome function. APOBEC3G promotes Warburg-like metabolic remodeling in HuT78 T cells under similar conditions. Hypoxia-induced RNA editing by APOBEC3G can be mimicked by the inhibition of mitochondrial respiration and occurs independently of HIF-1α.


APOBEC3G is an endogenous RNA editing enzyme in primary natural killer cells and lymphoma cell lines. This RNA editing is induced by cellular crowding and mitochondrial respiratory inhibition to promote adaptation to hypoxic stress.


The diffusible messenger molecule nitric oxide (NO) plays multiple roles in the nervous system. Among its physiological actions, those related to neuronal survival and differentiation, as well as to synaptogenesis and synaptic plasticity, have been well characterized during the last few years (Brenman et al., 1997 Ciani et al., 2002 Contestabile, 2000 Contestabile and Ciani, 2004 Gibbs, 2003 Holscher, 1997 Keynes and Garthwaite, 2004 Thippeswamy et al., 2001). However, ample evidence also exists that NO may be an agent promoting neurodegeneration when uncontrolled production and/or inefficient scavenging results in dangerous chemical reactions with various cellular components (Contestabile et al., 2003 Dawson and Dawson, 1998 Keynes and Garthwaite, 2004 Yun et al., 1997). Most actions of NO are mediated downstream through the guanylate cyclase/cGMP system (Garthwaite and Boulton, 1995).

Nitric oxide is a negative regulator of proliferation in several cell types, including neural cells in culture (Ciani et al., 2004 Maragos et al., 1993 Murillo-Carretero et al., 2002 Nisoli et al., 1998 Peunova and Enikolopov, 1995 Yang et al., 1994). Recently, a role for NO in neurogenesis has also emerged (Cardenas et al., 2005 Chen et al., 2004 Contestabile and Ciani, 2004 Enikolopov et al., 1999 Matarredona et al., 2005 Packer et al., 2003). The actual effect of NO in the regulation of neurogenesis, however, is still poorly understood and has been the source of conflicting results and interpretations. Recent data demonstrate that NO acts, in vivo, as a negative regulator of precursor proliferation in restricted areas of the brain, the dentate gyrus of the hippocampal formation and the subventricular zone of the forebrain, where neurogenesis persists in adult mammals (Cheng et al., 2003a Matarredona et al., 2005 Moreno-Lopez et al., 2000 Packer et al., 2003 Park and Wei, 2003). In these same areas, however, NO appears to stimulate neurogenesis as a consequence of brain damage caused by ischemia or traumatic injury (Chen and Cheung, 2005 Lu et al., 2003 Zhang et al., 2001 Zhu et al., 2003). The stimulation of neurogenesis under traumatic conditions, has been related to the neuroprotective role of NO (Cardenas et al., 2005 Estrada and Murillo-Carretero, 2005 Keynes et al., 2004). To better understand how NO affects neurogenesis, it is important to study its function during development. However, relatively few studies have addressed this issue so far, and some contrasting results have emerged from those that have. Nitric oxide was found to negatively regulate neurogenesis in the toad optic tectum (Peunova et al., 2001) and to affect the balance between cell cycle progression and apoptotic elimination in the chick neural tube (Plachta et al., 2003). The only study performed so far on mammalian developmental neurogenesis, has been carried out in mice knocked down for the neuronal nitric oxide synthase (nNOS) isoform (Chen et al., 2004). In this model, it was found that impaired NO production resulted in temporarily restricted decrease of neuronal precursor proliferation in the olfactory epithelium (Chen et al., 2004). This suggests that, contrary to what has been described for adult neurogenesis (Cheng et al., 2003a Moreno-Lopez et al., 2004 Packer et al., 2003), NO may stimulate neurogenesis during early brain development. Evidence from studies in culture, however, do not support this suggestion. In a recent report based on cultures of foetal rat neocortex, exogenously provided NO depressed the proliferation of neuronal precursors (Li, 2005). In line with this result, we have recently observed that NOS inhibition increased the division rate of neuronal precursors in cultures from neonatal rat cerebellum (Ciani et al., 2004). Furthermore, under these conditions we noticed that a transcription factor essential to promote cerebellar neurogenesis, Myc (Knoepfler et al., 2002), was upregulated. The data briefly reviewed above, indicate that the conflicting roles attributed to NO in neurogenesis regulation might depend on the species, the developmental stage and the in vivo versus in vitro condition.

Cerebellar development represents a very useful model to explore the role of NO in developmental neurogenesis. Genesis of cerebellar granule neurons, the largest neuronal population of the cerebellum, is an essentially postnatal event in rodents, lasting from birth up to the third week of age and being particularly prominent during the first week of life (Altman, 1982). A major determinant of granule cell neurogenesis has been identified in the Sonic hedgehog (Shh) peptide, which is produced by Purkinje cells and promotes proliferation of granule neuron precursors through positive regulation of the transcription factor N-Myc and factors of the Gli family (Corrales et al., 2004 Kenney et al., 2003 Wechsler-Reya and Scott, 1999). Granule neurons express the neuronal isoform of NOS (Baader et al., 1997) and we have previously exploited schedules of pharmacological NOS inhibition, able to almost completely block the enzymatic production of NO in the rat cerebellum during development (Virgili et al., 1999).

In the present study, we report that neonatal NO deprivation in the rat increases the rate of cell proliferation during cerebellar neurogenesis and upregulates Myc and cyclin D1 expression only during a restricted time window that corresponds to the first three postnatal days. This effect appears to be mediated by the primary downstream effector of NO, cGMP. Its temporarily restricted spanning may be attributed to escape of cGMP from NO control with the progression of development.

Activity and expression of nNOS in cerebellar development of control and L-NAME-treated rat pups. Pups were s.c. injected with L-NAME (60 mg/kg/day, divided in two daily administrations) for various neonatal intervals (P1-P3, P3-P5, P5-P7) while control pups received vehicle. Animals were killed on the last day of treatment. (A) Maturation of calcium-dependent NOS activity in developing cerebella (white bars) and almost complete inhibition obtained by L-NAME treatment (black bars). Values are the mean ± s.e.m. of at least three experiments for each time point. * P<0.05, compared with the control at P3, #P<0.001 compared with the corresponding control (Bonferroni's test after ANOVA). (B) Total RNA was extracted from cerebella of control and L-NAME-treated rat pups and equal amounts of RNA were used for real-time PCR performed with specific primers for nNOS RNA as described in the Materials and Methods section. These results, expressed as a percentage of control at P3 are the mean ± s.e.m. of four different experiments performed in triplicate. * P<0.05, ** P<0.01 compared to P3 (Bonferroni's test after ANOVA). (C) In western blot experiments, cerebellar extracts from control and L-NAME-treated rat pups were probed with an antibody specific for nNOS. Bands were quantified by optical densitometry and normalized for the amount of β-actin. These results are the mean ± s.e.m. of four different experiments. * P<0.05, ** P<0.01 compared with P3 (Bonferroni's test after ANOVA).

Activity and expression of nNOS in cerebellar development of control and L-NAME-treated rat pups. Pups were s.c. injected with L-NAME (60 mg/kg/day, divided in two daily administrations) for various neonatal intervals (P1-P3, P3-P5, P5-P7) while control pups received vehicle. Animals were killed on the last day of treatment. (A) Maturation of calcium-dependent NOS activity in developing cerebella (white bars) and almost complete inhibition obtained by L-NAME treatment (black bars). Values are the mean ± s.e.m. of at least three experiments for each time point. * P<0.05, compared with the control at P3, #P<0.001 compared with the corresponding control (Bonferroni's test after ANOVA). (B) Total RNA was extracted from cerebella of control and L-NAME-treated rat pups and equal amounts of RNA were used for real-time PCR performed with specific primers for nNOS RNA as described in the Materials and Methods section. These results, expressed as a percentage of control at P3 are the mean ± s.e.m. of four different experiments performed in triplicate. * P<0.05, ** P<0.01 compared to P3 (Bonferroni's test after ANOVA). (C) In western blot experiments, cerebellar extracts from control and L-NAME-treated rat pups were probed with an antibody specific for nNOS. Bands were quantified by optical densitometry and normalized for the amount of β-actin. These results are the mean ± s.e.m. of four different experiments. * P<0.05, ** P<0.01 compared with P3 (Bonferroni's test after ANOVA).


Abstract Heterogeneity of smooth muscle cell (SMC) phenotype and function is rapidly emerging as an important concept. We have recently described that phenotypically distinct SMC subpopulations in bovine pulmonary arteries exhibit unique proliferative and matrix-producing responses to hypoxic pulmonary hypertension. To provide better understanding of the molecular mechanisms contributing to this phenomenon, experimental studies will require a reliable in vitro model. The purpose of the present study was first to determine if distinct SMC subpopulations, similar to those observed in vivo, could be selectively isolated from the mature arterial media, and then to evaluate whether select SMC subpopulations would exhibit heightened responses to growth-promoting stimuli and hypoxia. We were able to reproducibly isolate at least four phenotypically unique cell subpopulations from the inner, middle, and outer compartments of the arterial media. Differences in cell phenotype were demonstrated by morphological appearance and differential expression of muscle-specific proteins. The isolated cell subpopulations exhibited markedly different growth capabilities. Two SMC subpopulations grew slowly in 10% serum and were quiescent in plasma-based medium. The other two cell subpopulations, exhibiting nonmuscle characteristics, grew rapidly in 10% serum and proliferated in plasma-based medium and in response to hypoxia. Certain colonies of the nonmuscle-like cell subpopulations were found to grow autonomously under serum-deprived conditions and to secrete mitogenic factors. Our data, demonstrating that phenotypically distinct cells with enhanced growth potential exist within the normal arterial media, support the idea that these unique cells could contribute selectively to the pathogenesis of vascular disease.

For many years, the arterial media was believed to be composed of a phenotypically homogeneous population of SMCs. However, the possibility that the arterial media might be composed of phenotypically and functionally diverse subpopulations of SMCs is raised by the fact that vascular SMCs, especially in response to pathological stimuli, are required to perform numerous functions, including contraction, extracellular matrix protein synthesis, and replication. In disease, essential functions of the vessel, such as contraction, must be maintained while, at the same time, the reparative process is carried out. It seems likely that these diverse functions could best be served by multiple cell subpopulations with different functional capabilities rather than by modulation of a single SMC phenotype into functionally different phenotypes.

Recent in vivo studies in the systemic circulation have demonstrated that morphologically and immunohistochemically distinct SMC phenotypes exist within the arterial media of various mammalian species. 1–6 Within the pulmonary circulation, we have recently reported that the normal media of developing and mature bovine arteries is also composed of multiple unique SMC subpopulations based on differential expression of muscle-specific markers 7 and diverse profiles of tropoelastin mRNA expression. 8 In a series of developmental studies, analysis of the pattern of muscle-specific cytoskeletal protein expression demonstrated that the identified SMC subpopulations progressed along unique developmental differentiation pathways, suggesting the existence of distinct developmental lineages for medial SMC subpopulations. 7 We have also shown that the phenotypically distinct SMC subpopulations exhibit markedly different proliferative and matrix-producing capabilities in response to the stimuli associated with hypoxia-induced pulmonary hypertension. 9,10 Although it is clear now that the arterial media is a heterogeneous organ composed of SMCs with distinct phenotypes and functional properties, the molecular mechanisms contributing to the existence of distinct SMC subpopulations remain uncertain. To provide better understanding of this phenomenon, experimental studies will require a reliable in vitro model in which the isolated cell subpopulations can maintain unique characteristics of interest over time in culture.

The purpose of the present study was to selectively isolate the phenotypically distinct cell subpopulations (both smooth muscle and “nonmuscle-like”) identified in vivo 7 and then to evaluate whether the nonmuscle-like cell subpopulations would exhibit heightened responses to growth-promoting stimuli as well as to hypoxia. Using selective isolation techniques and selective media, we were able to reproducibly isolate from mature bovine arterial media (both pulmonary and systemic) four phenotypically distinct medial cell subpopulations. Two cell subpopulations could be generally classified as smooth muscle, and two could be classified as nonmuscle-like, based on their morphological and biochemical differences. The isolated cell subpopulations were found to exhibit markedly different growth capabilities under various conditions and to maintain these properties over multiple passages in culture. These data raise the possibility that nonmuscle-like cells with enhanced growth capabilities exist in the normal arterial media, exhibit unique responses to pathophysiological stimuli, and thus contribute selectively to the pathogenesis of vascular disease.

Materials and Methods

Isolation of Cell Subpopulations

Main pulmonary arteries and aortic arches were obtained from adult (2-year-old) cows (n=8). Segments of the main pulmonary artery (proximal to its bifurcation) and the aortic arch (immediately distal to the subclavian artery) were used for cell isolation. These vascular segments were cut open and mechanically stripped of adventitia. To ensure complete removal of the adventitia, a thin portion of the outermost media was also discarded. Endothelium was removed by gentle scraping of the luminal surface of the vessel with a scalpel blade.

We have previously demonstrated that the intact mature bovine arterial media is composed of at least four phenotypically unique cell subpopulations that reside in distinct medial layers. 7 The subendothelial media is predominately populated by nonmuscle-like cells the middle media, by SMCs and the outer media, by two phenotypically distinct cell subpopulations, one smooth muscle and the other nonmuscle-like. In order to isolate the four cell subtypes identified in vivo, we first separated the arterial media into the three previously described layers: (1) a very thin subendothelial layer (termed here L1), (2) an intermediate-sized middle layer (termed L2), and (3) a thick outer layer (termed L3) (Fig 1A , and Reference 7 7 ). We found that these three medial layers could be mechanically separated from one another because of distinct mechanical properties of each layer, apparently due to specific patterns of cell arrangement and elastic lamellar distribution (Fig 1B ). After separation of the media into three layers, cells were grown from each layer by explant techniques as previously described. 11 Tissue explants were maintained in complete DMEM (Sigma Chemical Co) supplemented with 200 U/mL penicillin, 0.2 mg/mL streptomycin, and either 10% CS (HyClone Laboratories) or 10% plasma-derived serum (Cocalico Biologicals, Inc). Plasma-derived serum was used to selectively obtain cells with unique growth capabilities as previously described 12 and, therefore, was carefully screened before use to make certain it lacked PDGF-derived mitogenic activity.

Since our goal was to obtain pure subpopulations of smooth muscle and nonmuscle-like cells previously identified in the intact mature vascular media, 7 we first selectively isolated individual cell colonies with a distinct, although uniform, morphological appearance from primary culture using “ring-based” techniques (see below). We then examined expression of smooth muscle–specific markers in each isolated cell subpopulation. Only cell subpopulations with uniform morphological appearance and uniform patterns of expression (or lack thereof) of smooth muscle markers were selected for subsequent experiments.

To isolate individual cell colonies growing from tissue explants in primary culture, plastic rings (5 to 10 mm in diameter, greased on the bottom) were placed over each cell colony of interest. Cells within the ring were trypsinized and transferred to a 24-multiwell plate for expansion. Simultaneously, a small portion of cells was plated for immunostaining analysis as described below.

All studies were carried out using cells at passages 1 to 8. Cell cultures were tested for mycoplasma contamination using a Gen-Probe Mycoplasma T. C. Rapid Detection System (Gen-Probe Inc) and were negative.

Bovine endothelial cells were obtained from mature main pulmonary artery as previously described. 13

Immunofluorescence Analysis

Cells isolated from distinct medial layers were assessed for expression of muscle-specific contractile and cytoskeletal proteins, α-SM-actin, smooth muscle myosin heavy chains (termed here SM-myosin), and metavinculin. Indirect single- or double-label immunofluorescence staining techniques were used. Monoclonal anti–α-SM-actin antibodies (clone 1A4) were purchased from Sigma. Rabbit antibodies against bovine aortic SM-myosin were kindly provided by Dr R. S. Adelstein (National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, Md). 14 We have previously shown that these antibodies (termed anti–SM-myosin antibodies in this article) react strongly with smooth muscle SM-1 and SM-2 isoforms of myosin heavy chains in bovine arterial tissue and do not recognize nonmuscle isoforms. 7 Affinity-purified rabbit anti-metavinculin antibodies were previously described and shown to react specifically with metavinculin but not vinculin. 7 As a control, monoclonal anti-vinculin antibodies 15 that react with both vinculin and metavinculin were used. Monoclonal anti-BrdU antibody (Becton Dickinson) was used at a dilution of 1:200. Biotinylated horse anti-mouse IgG and avidin/biotin–horseradish peroxidase complex (Pierce) were applied at dilutions recommended by the supplier. For assessment of endothelial marker expression, affinity-purified rabbit antibodies against von Willebrand factor (DAKO Corp) were used.

Cells grown on Tissue-Tek chamber slides (Nunc) in 10% CS to confluence were then growth-arrested (0.1% CS) for 48 hours, fixed in absolute methanol at −20°C for 10 minutes, and processed for indirect immunostaining as follows. For double-label immunofluorescence staining of α-SM-actin and SM-myosin, fixed cells were incubated with a cocktail of monoclonal α-SM-actin and polyclonal anti–SM-myosin antibodies (diluted 1:100 and 1:1000, respectively) for 1 hour at room temperature. After 3 washes in PBS, cells were incubated with a cocktail of biotinylated anti-mouse IgG and FITC-conjugated anti-rabbit IgG (both diluted at 1:100, and both purchased from Sigma) for 1 hour at room temperature. The staining was accomplished by incubation with streptavidin–Texas Red (1:50, Amersham Corp). For double-label immunofluorescence analysis of Ki-67 and SM-myosin expression, staining with Ki-67 antibody was accomplished first by using a biotin/streptavidin–Texas Red detection system and then by staining with polyclonal anti-metavinculin antibodies and FITC-conjugated anti-rabbit IgG as secondary antibodies. For single-label immunostaining, affinity purified rabbit anti-metavinculin antibodies or monoclonal anti-vinculin antibodies were used at dilutions 1:10 and 1:50, respectively. FITC-conjugated anti-rabbit or anti-mouse IgGs (1:40, Sigma) were used as secondary antibodies. Controls were performed in which primary antibodies were replaced by nonimmune rabbit serum or nonimmune mouse ascites (Sigma) at the same dilutions as primary antibodies.

The stained cells were examined with a Nikon Optiphot epifluorescence photomicroscope. Colored photomicrographs were taken with Ektachrome 160T film. Black and white micrographs were taken on Kodak T-MAX 400 film.

SDS-PAGE and Western Blotting Assays

Protein extraction and SDS-PAGE (using 12% polyacrylamide gels) were performed according to the method first described by Laemmli. 16 Twenty micrograms of protein was loaded onto each lane. Protein transfer to a nitrocellulose membrane was carried out according to Towbin et al. 17 To detect α-SM-actin in protein extracts, monoclonal anti–α-SM-actin antibodies (clone 1A4, Sigma) at a dilution of 1:2000 were used as primary antibodies, followed by anti-mouse IgG conjugated with horseradish peroxidase (diluted 1:4000, Sigma). Detection of the bound antibodies was performed using an ECL kit (Amersham Corp). The NIH Image program (National Institutes of Health) was used for quantitative densitometric scanning of the blots. Results of three separate Western blots were analyzed. Data were expressed in relative scan units the content of α-SM-actin in L1-cells equals 100 relative scan units.

Cell Growth Assays

Cell Growth in the Presence and Absence of Serum

Cells (at passages 3 to 7) were plated onto 24-multiwell plates at a density of 20×10 3 cells per well in DMEM supplemented with 10% CS. On day 1, four wells were trypsinized and counted in a standard SPotlite hemacytometer (Baxter). The remaining wells were rinsed with PBS, and 0.5 mL of DMEM supplemented with either 0.1% CS or 10% CS was added. Cells (in four wells) were trypsinized and counted at different time points. Data were expressed as follows: cell number×10 4 /well. Population doubling time was calculated in exponentially growing cells as described elsewhere. 18

Cell Growth in Plasma-Derived Serum

Cells from distinct subpopulations were assessed for their capability to proliferate in medium supplemented with 10% plasma using the same method as described above for growth in 10% serum.

Response to Purified Mitogens (DNA Synthesis Assay)

DNA synthesis of distinct cell subpopulations in response to various purified mitogens was determined by measuring [ 3 H]thymidine incorporation under serum-free conditions. Cells were plated at 20×10 3 cells/well onto 24-multiwell plates in complete DMEM supplemented with 10% CS. The next day, cells were rinsed with PBS and growth-arrested in 0.1% CS for 72 hours. For additional 24 hours, serum-free DMEM supplemented with 0.5 μCi/mL [ 3 H]thymidine (ICN Biochemicals, Inc) and the following purified mitogens were added: PDGF-BB (10 ng/mL, Bachem Fine Chemicals), IGF-I and IGF-II (both at 100 ng/mL), and bFGF (30 ng/mL, Bachem). Mitogen doses used in the present study were found to elicit maximal responses in bovine SMCs (data not shown). As positive and negative controls, [ 3 H]thymidine incorporation in 10% serum and in 0.1% serum was assessed. Four wells with cells of the same subpopulation were assessed for each mitogen. Measurement of [ 3 H]thymidine incorporation was performed as described elsewhere. 19 Cell counts were concurrently obtained from four additional wells. Incorporation of [ 3 H]thymidine into DNA was expressed as disintegrations per minute (dpm) per cell and/or per 1000 cells.

Effects of Heparin on Cell Growth

The effect of heparin on cell growth in different cell subpopulations was assessed as previously described. 20 Briefly, cells were plated onto 24-multiwell plate at 1×10 4 cells/well in 10% serum–supplemented DMEM. Heparin (1 to 1000 μg/mL) was added on day 1 and again on day 3. Cells were counted in four wells on days 1 and 5.

Effects of Hypoxia on DNA Synthesis and Cell Growth

Cells were plated at 20×10 3 cells/well onto 24-multiwell plates in complete DMEM supplemented with 10% CS. After reaching confluence, cells were rinsed with PBS, and 0.5 mL of DMEM containing 0.1% CS was added to each well. Cells were then placed in sealed humidified gas chambers (Bellco Glass). Chambers were purged with 21% or 3% O2 with 5% CO2 balanced with N2. After gassing for 20 minutes, chambers were placed in the 37°C incubator for 48 hours. After this incubation period, chambers were open, [ 3 H]thymidine (0.5 μCi/mL) was added to wells, and the chambers were sealed and gassed again. After an additional 24 hours of incubation, measurement of [ 3 H]thymidine incorporation was performed as described elsewhere 19 and expressed as dpm per cell. To determine not only DNA synthesis but also growth, cells were plated at 20×10 3 cells/well in 24-well plates, four wells were counted on day 1, and the rest were incubated under normoxic (21% O2) or hypoxic (3% O2) conditions for 3 more days and then counted.

Coculture Experiments

To determine if any of the isolated cell subpopulations secreted growth-promoting and/or growth-inhibiting factors, coculture experiments using “source” and “target” cells were performed as previously described by R. Majack (Cook et al 21 ). Briefly, target cells were plated within a 10-mm plastic ring placed in the center of 35-mm tissue culture dish. On the next day the source cells of interest were plated around the periphery of the ring. After reaching subconfluence, both source and target cells were incubated in serum-deprived medium for 72 hours. Medium from target cells was then withdrawn, the ring was removed, and conditioned serum-free medium from the source cells was allowed to spread over the target cell population. The remaining ring of grease prevented the cells from migrating. After coculture for 48 hours, BrdU was added for additional 24 hours. The percentage of BrdU-positive nuclei in the target cell population was assessed by immunocytochemistry.


Cells Isolated From Distinct Medial Layers Exhibit Unique Morphological Appearances and Patterns of Cytoskeletal Protein Expression

In order to determine whether multiple SMC subpopulations with unique phenotypic characteristics similar to those observed in vivo could be isolated from a single specific vascular site, the arterial media was first separated into distinct layers (Fig 1B and Reference 7 7 ), and then tissue explants derived from each medial layer were incubated in culture medium with either 10% complete calf serum or with 10% plasma-derived serum. The cells growing from tissue explants of subendothelial, middle, and outer media were then assessed for morphological appearance and expression of contractile and cytoskeletal proteins as described below.

Since previous in vivo studies had demonstrated the presence of nonmuscle-like cells in the subendothelial compartment of the mature bovine arterial media, 7 our objective was to selectively isolate these cells in culture. When subendothelial tissue explants were incubated in 10% complete calf serum, two morphologically distinct cell subpopulations were observed. Cells of one subpopulation (hereafter termed L1-cells), constituting <10% of all cell colonies, appeared small and “rhomboidal” in shape and, at confluence, formed a dense multilayer network (Fig 2A ). These L1-cell colonies were selectively isolated from primary culture using plastic rings (see “Materials and Methods”) and assessed for contractile protein expression. Immunofluorescence analysis demonstrated that L1-cells expressed little to no α-SM-actin and virtually no SM-myosin (Fig 3A ). Cells in the other subpopulation identified in primary cultures from the subendothelial media appeared bipolar or spindle-shaped. At confluence, these cells formed the “hill-and-valley” pattern traditionally described for SMCs (Fig 2B ). Immunofluorescence analysis of these cells demonstrated intense staining with both anti–α-SM-actin and anti–SM-myosin antibodies similar to that observed for SMCs derived from the middle (L2) media (Fig 3B ).

Because small rhomboidal cells lacking expression of smooth muscle markers represented only a minority of cell colonies in primary cultures from subendothelial media grown in 10% serum, attempts were made to selectively increase the number of this cell type in primary culture. Since use of plasma instead of serum had previously been shown to allow selective growth of unique SMC subpopulations, 12 we incubated subendothelial tissue explants in medium supplemented with 10% plasma-containing medium. Under these conditions, rhomboidal cell colonies were observed to predominate (>90%). Immunofluorescence analysis demonstrated the same pattern of α-SM-actin and SM-myosin expression as seen in L1-cell colonies selectively isolated from primary cultures grown in serum. When rhomboidal L1-cells initially grown in 10% plasma were subsequently subcultured in 10% serum–containing medium, their characteristic rhomboidal appearance and pattern of immunostaining did not change. Thus, using selective isolation techniques and/or culture medium supplemented with 10% plasma, we were able to isolate and expand in culture a subpopulation of nonmuscle-like cells similar to those identified in the subendothelial media of mature bovine arteries in vivo. 7

Virtually all cell colonies isolated from the middle (L2) media explants grown in 10% serum exhibited uniform morphological characteristics. These cells (hereafter termed L2-SMCs) exhibited a spindle-shaped bipolar appearance and, at confluence, showed the hill-and-valley pattern traditionally described for vascular SMCs (Fig 2B ). No cell growth was observed when middle media explants were incubated in 10% plasma. Immunofluorescence analysis demonstrated intense staining of these cells with both anti–α-SM-actin and anti–SM-myosin antibodies (Fig 3B ). The phenotype of this cell subpopulation was consistent with that described in the middle media of the mature bovine arteries in vivo. 7

Primary cultures of cells isolated from the outer (L3) media in serum-based medium were observed to be composed of two morphologically distinct cell subpopulations (Fig 2C ). One cell subpopulation (constituting ≈70% of all cell colonies) consisted of spindle-shaped cells (hereafter termed L3-“S” cells), which formed at confluence a hill-and-valley pattern (Fig 2e ). Another cell subpopulation (constituting ≈30% of all cell colonies observed) consisted of cells that exhibited a “rounded,” cobblestone, or epithelioid morphology (termed L3-“R” cells) and which at confluence formed a monolayer (Fig 2d ). Colonies of spindle-shaped L3-“S” cells and rounded L3-“R” cells were selectively isolated from primary culture using plastic rings (see “Materials and Methods”) and processed for immunofluorescence analysis of contractile protein expression. The two cell subpopulations exhibited markedly different patterns of contractile protein expression: spindle-shaped L3-“S” cells stained intensely with both anti–α-SM-actin and –SM-myosin antibodies (Fig 3C ), whereas rounded epithelioid L3-“R” cells demonstrated only a moderate intensity of staining of stress fibers with α-SM-actin antibodies but virtually no SM-myosin staining (Fig 3D ).

When explants from the outer (L3) media were incubated in 10% plasma, >90% of cell colonies exhibited a rounded epithelioid morphological appearance. When these cells were subsequently subcultured in 10% serum–based medium, they maintained their rounded morphological appearance and biochemical characteristics and never acquired bipolar appearance or expressed SM-myosin.

Because the two muscle cell subpopulations, L2- and L3-“S” SMCs, isolated from the middle and outer media, respectively, exhibited similar morphology and both expressed α-SM-actin and SM-myosin, we questioned whether they could be differentiated on the basis of expression of other muscle-specific proteins. Since metavinculin expression differentiated two SMC subpopulations in vivo, 7 we analyzed the expression of this protein by the two morphologically spindle-shaped cell subtypes (L2 and L3-“S” SMCs) in vitro. We found that only the subpopulation of spindle-shaped cells from the outer media (L3-“S” SMCs) expressed metavinculin in culture (Fig 4 ). These findings are again consistent with the previous in vivo observations. 7

To make certain that the isolated nonmuscle-like L1- and L3-“R” cell subpopulations were not contaminated with endothelial cells derived from either the intimal surface of the vessel (for subendothelial L1-cells) or from the vasa vasorum (for outer media L3-cells), immunofluorescence analysis with antibodies against the endothelial cell marker, von Willebrand factor, was performed. No von Willebrand factor staining was noted in nonmuscle-like L1- and L3-“R” cultures (data not shown).

The unique morphological characteristics of all four cell subpopulations were maintained over multiple passages in culture (experiments presented in the present study were carried out on cells at passages 1 to 8). Never were any of the four cell subpopulations observed to revert to another morphological phenotype under similar culture conditions. Cells at early passages could be frozen and thawed for further use without causing changes in morphological appearance.

Quantitative Analysis of α-SM-Actin Content in Distinct Cell Subpopulations: Western Blotting Assay

Because immunofluorescence analysis demonstrated that all four cell subpopulations stained positively with antibodies against α-SM-actin, although with markedly different levels of intensity, we sought to quantitatively assess the relative content of α-SM-actin in different cell subpopulations using Western blotting analysis. Relative content of α-SM-actin was the lowest in L1-cells, 1.82-fold higher in L3-“R” cells, 4.4-fold higher in L2-SMCs, and 6.6-fold higher in L3-“S” SMCs compared with L1-cells (data not shown).

Phenotypically Distinct Cell Subpopulations Exhibit Markedly Different Growth Capabilities Under Identical Conditions

To determine if morphologically and biochemically distinct cell subpopulations exhibited different growth capabilities in response to growth-promoting and growth-inhibitory stimuli, the rate of cell proliferation was measured in response to serum and/or plasma stimulation after serum withdrawal as well as in response to various concentrations of heparin. The effect of heparin and hypoxia on cell growth was also assessed. Additionally, DNA synthesis was determined by measuring [ 3 H]thymidine incorporation under conditions of serum stimulation and serum withdrawal and in response to various purified mitogens as well as in response to hypoxia. Finally, the effect of coculture of different cell phenotypes on DNA synthesis of a specific cell subset was determined.

Cell Growth in Response to Stimulation by Serum and/or Plasma

As shown in Fig 5A , cells isolated from distinct layers of arterial media and exhibiting distinct morphological and biochemical characteristics demonstrated markedly different growth responses to 10% serum stimulation. Subpopulations of L1- and L3-“R” cells grew more rapidly than did the L2- and L3-“S” SMC subpopulations. Population doubling times during log-phase growth for the colonies of L1- and L3-“R” cells were 40.7±2.4 hours and 42±2.8 hours, respectively, whereas subpopulations of L2- and L3-“S” SMCs exhibited doubling times of 87±4.7 hours and 109±8.2 hours, respectively. Interestingly, we found that two cell colonies isolated from the subendothelial (L1) media of two different adult animals (termed here colonies L1–1 and L1–5) grew in 10% serum–containing medium faster than other L1-cell colonies and showed a population doubling time of only 29.5 hours and 32 hours, respectively (data not shown). At day 10, the saturation density of L1-cells was 3.6-fold higher than that of L2-SMCs and 4-fold higher than that of L3-“S” SMCs. The saturation density of L3-“R” cells was 2.1-fold and 2.6-fold higher than that of L2- and L3-“S” SMCs, respectively.

In medium supplemented with 10% plasma instead of serum, L1- and L3-“R” cell subpopulations demonstrated an ability to proliferate, although at a slower rate than in 10% serum. In contrast, L2- and L3-“S” SMC subpopulations were quiescent under these conditions (Fig 5B ).

In serum-deprived (0.1% CS) medium, only two cell colonies derived from subendothelial media were observed to proliferate autonomously over prolonged periods in culture and hereafter will be termed L1-AUT cells. Cell counts in 0.1% serum over a 10-day period demonstrated that the number of L1-AUT cells increased ≈8-fold, whereas cell number from other colonies of L1-cells as well as L3-“R”, L2-, and L3-“S” cells remained virtually unchanged (Fig 5C ). After 72 hours of serum deprivation, DNA synthesis (assessed by both BrdU and [ 3 H]thymidine incorporation) was low in all cell subpopulations, except for L1-AUT cells, in which DNA synthesis was found to be markedly increased (Fig 6B ).

DNA Synthesis in Response to Purified Mitogens

Cell subpopulations were found to differ in their response to peptide mitogens (Fig 6A ). PDGF-BB markedly increased DNA synthesis in L3-“R” and L1-cell subpopulations and, to lesser extent, in L2-SMC and especially in L3-“S” SMC subpopulations. IGFs (both -I and -II) stimulated DNA synthesis in L1- and L3-“R” cells but not in L2- or L3-“S” SMCs. No significant differences between IGF-I and -II were noted. When bFGF was tested, it was found to increase DNA synthesis in all cell subpopulations but to lesser extent than did PDGF. Interestingly, L1-AUT cells, in which DNA synthesis under serum-deprived conditions was markedly elevated, were found to be quite unresponsive to the mitogens used in the present study: PDGF-BB had little, although distinct, stimulatory effect, whereas bFGF and IGFs had no stimulatory effect on DNA synthesis in these cells (Fig 6B ).

Relationship Between Proliferation and Expression of Muscle-Specific Markers in L2- and L3-“S” SMC Subpopulations

Although L2- and L3-“S” SMC subpopulations ex-hibited somewhat similar growth responses to serum (Fig 5 ), their markedly different expression of a muscle-specific differentiation-related cytoskeletal protein metavinculin (Fig 4 ) led us to investigate the correlation between cytodifferentiation and proliferation in response to 10% serum stimulation in these two cell subtypes. We found marked differences in the relationship between cytodifferentiation and proliferation (as defined by the expression of smooth muscle myosin heavy chains [SM-myosin] and the proliferation-associated nuclear marker, Ki-67) in these two SMC subtypes. We observed that under growth-arrested conditions (0.1% CS, 72 hours), the majority of cells in both L2- and L3-“S” SMC cultures expressed SM-myosin. However, after serum stimulation, L2-SMC cultures retained a high percentage of SM-myosin–positive cells, whereas in cultures of L3-“S” SMCs the number of SM-myosin–positive cells decreased by approximately half. Furthermore, we found that serum-stimulated L2-SMCs simultaneously expressed both Ki-67 and SM-myosin antigens at a high frequency (78.9% of all replicating cells expressed SM-myosin), whereas in L3-“S” SMCs the majority (90.8%) of replicating cells were SM-myosin–negative (data not shown). It is important to note that after serum stimulation, the majority of both L2- and L3-“S” SMCs continued to express α-SM-actin.

Cell Growth in Response to Heparin

Because we observed marked differences in responsiveness to promitogenic stimuli among distinct medial cell subpopulations, we also tested whether differences in response to growth-inhibitory factors would be detectable. We examined the effect of heparin on cell growth of two cell subpopulations (L1-cells and L2-SMCs) that exhibited significant differences in serum-stimulated growth. We found that heparin did not affect the growth rate of slow-growing L2-SMCs under serum-stimulated conditions even when applied in high concentrations (1000 μg/mL). In contrast, heparin exerted dramatic growth-inhibitory effects on fast-growing L1-cells even at concentrations as low as 1 μg/mL (Fig 7 ).

Inhibition of L1-Cell Growth by Coculture With L2-SMCs

Because L1- and L3-“R” cell colonies were observed quite infrequently in actively proliferating primary cultures yet exhibited high growth rates when grown after selective isolation (and therefore somewhat “purified”), we attempted to study whether their growth in primary cultures was inhibited by factors secreted by other cell subpopulations. For this purpose, we chose to assess growth potential of the fastest growing L1-cell colonies, L1-AUT (population doubling time, 29.5 hours and 32 hours, respectively) in coculture with L2-SMCs under various serum concentrations (0.1%, 1%, 3%, and 5% serum). Under all the conditions tested, DNA synthesis in L1-AUT cells was significantly inhibited by coculture with L2-SMCs (Table 1 ). Even when stimulated by 5% serum, DNA synthesis in L1-cells cocultured with L2-SMCs was inhibited to the level of that in serum-deprived (0.1% CS) medium.

DNA Synthesis in Response to Hypoxia

Because in previous in vivo studies 9 we demonstrated differential proliferative responses of phenotypically distinct medial SMC subpopulations to hypoxia-induced pulmonary hypertension, we attempted to evaluate the proliferative response of the isolated distinct medial cell subpopulations to hypoxia. We found that under serum-stimulated conditions, DNA synthesis was increased in L1- and L3-“R” SMCs in response to hypoxia, whereas it was decreased in L2- and L3-“S” SMCs (Fig 8A ). Growth assays performed on autonomously growing L1-AUT cells demonstrated a significant increase in cell number under hypoxia (3% O2) in serum-deprived medium compared with normoxia (21% O2) (Fig 8B ).

Selected Colonies of L1- and L3-“R” Cells Secrete a Mitogenic Factor

The possibility that certain vascular SMC subpopulations secrete mitogenic factors and that others do not has been raised by previous studies. 22 Thus, we were interested in determining whether any of the isolated cell subpopulations secreted mitogenic factors. We hypothesized that those cell subpopulations that were able to proliferate in plasma-based medium were most likely to secrete mitogenic factors. Using coculture techniques, we assessed the effect of serum-free medium conditioned for 48 to 72 hours by different colonies of L1-cells, L2-SMCs, L3-“R” cells, and L3-“S” SMCs on growth-arrested L2-SMCs. As shown in Fig 9 , L2-SMCs had low rates of DNA synthesis (as defined by BrdU nuclear incorporation) when cocultured with other L2-SMCs or with L3-“S” SMCs. In contrast, when cocultured with certain colonies of L1- and L3-“R” cells (two of seven L1-cell colonies tested, and three of nine L3-“R” cell colonies tested), DNA synthesis in L2-SMCs increased at least 5-fold, suggesting the secretion of a mitogenic factor(s) by these selected colonies of L1- and L3-“R” cells. We also found that serum-deprived medium, conditioned by these cell colonies for 48 to 72 hours and then applied to the culture of growth-arrested L2-SMCs, had the same mitogenic effect as that observed in live cocultures. Freezing and thawing of this conditioned medium did not affect its growth-promoting effect.

Previous studies have demonstrated that certain phenotypes of vascular SMCs secrete a PDGF-like mitogenic activity. 22–24 However, our preliminary results show that the mitogens present in the conditioned medium from selected L1- and L3-“R” colonies are not any of the PDGF isoforms but rather are two unique members of heparin-binding family of growth factors (A.A. Aldashev, unpublished data, 1996).


The present study demonstrated that at least four distinct cell subpopulations, each bearing a strong resemblance to a specific arterial cell subpopulation previously observed in vivo, 7 could be reproducibly isolated from the media of mature bovine main pulmonary arteries and aortas. The cell subpopulations with unique characteristics (see Table 2 ) were obtained from specific compartments of the arterial media (subendothelial, middle, and outer layers), which could be separated mechanically. Our view that these cell subpopulations represented phenotypically distinct cell lines and were not the result of “phenotypic modulation” in culture was based on the following: (1) characteristic morphological, immunobiochemical, and proliferative properties observed at early passages in a specific cell subpopulation were maintained over time in culture (ie, conversion from one phenotype to another was not observed) (2) each cell subpopulation was isolated from a specific medial layer of the same vascular segment and exhibited biochemical and functional characteristics similar to those of cells observed in the corresponding medial layer in vivo (3) the unique morphological, biochemical, and growth characteristics of each cell subpopulation were maintained under different culture conditions (in complete serum or in plasma) and (4) the findings were consistent among 8 different animals.

The findings regarding arterial SMC diversity reported in the present study in a large mammalian species support and extend previous experimental in vitro data that came primarily from avian and rodent species. 12,25–30 The present study, however, provides the first link between in vitro and in vivo observations by demonstrating that phenotypically and functionally unique vascular SMC subpopulations observed in vivo could be isolated in culture and retain, at least to a certain degree, similar biochemical and functional characteristics. For instance, in vivo one specific SMC subpopulation, residing in the outer media, was found to express the cytoskeletal protein metavinculin and to be resistant to the growth-promoting effects of hypoxic pulmonary hypertension. In vitro, only one cell subpopulation (L3-“S” SMCs), isolated from the outer media, expressed metavinculin (Fig 5 ). This SMC subpopulation also exhibited minimal responses to exogenous mitogens, and its growth capability was suppressed by hypoxia. In addition, previous in vivo studies demonstrated the existence of cell subpopulations in the normal mature vascular media that failed to express smooth muscle–specific markers. 7 In vitro, two nonmuscle-like cell subpopulations were derived from the subendothelial and outer media (L1- and L3-“R” cells, respectively). Additionally, we found that L1- and L3-“R” cells exhibited a highly proliferative phenotype and, unlike “traditional” SMCs, proliferated under hypoxic conditions, a response similar to that observed during hypoxic pulmonary hypertension in vivo. Thus, cell subpopulations described in the present in vitro study appear representative of resident cell subpopulations in the vascular media, raising the possibility that molecular mechanisms conferring unique cellular responses to pathological stimuli in vivo can be addressed in vitro.

The idea raised by others that the normal arterial media is composed not only of differentiated SMCs but also of nonmuscle-like cells is supported by our results. 12,25–33 In general, the distinct cell subpopulations identified could be subdivided into two major categories: muscle and nonmuscle-like (Table 2 ). However, our data also suggest that unique cell subtypes exist within each of these two general categories. For instance, within the general category of nonmuscle-like cells, differences in morphology, in growth pattern, and in mitogen secretion were observed. Within the “muscle” group, differences in the expression of cytoskeletal proteins (ie, metavinculin) and in the regulation of contractile protein (SM-myosin) expression during proliferation were noted. Similarly, in vivo these two SMC subpopulations exhibited different biochemical characteristics, as well as distinct proliferative and matrix-producing responses to hypertensive stimuli. 7–9 It appears possible, or even likely, that distinct cell subtypes, existing within the general categories of muscle and nonmuscle-like cells, will be found to contribute in unique ways to vascular wall homeostasis under normal and pathological conditions.

Recent in vitro experiments in the systemic circulation have demonstrated the existence of nonmuscle-like cells with enhanced proliferative capabilities in the subendothelial space of normal aortas and suggest a distinct role for these cells in the pathogenesis of intimal thickening. 27,31,33,34 Villaschi et al 33 isolated relatively undifferentiated cells from the intimal aspect of normal adult rat aortas, very similar to the L1-cells described here. Holifield et al 34 demonstrated that a subpopulation of nonmuscle-like cells contributes selectively to the intimal thickening following balloon injury in canine arteries. The nonmuscle-like cells described in the present study were isolated from both the subendothelial and the outer compartments of the media. They exhibited heightened growth potential, unique proliferative responses to hypoxia, and the capacity (in at least certain colonies) to secrete mitogens and thus influence the proliferative phenotype of other SMC subpopulations. The existence of functionally unique cell subpopulations within distinct compartments of the vessel wall suggests that the vascular response to a given pathological stimulus could be cell specific, could be localized, and could differ significantly depending on the inciting injury.

The observation of unique cell subsets with markedly enhanced growth capabilities in animal species leads to speculation about the potential role similar cells (if they exist) could play in human vascular diseases, such as arteriosclerosis, postangioplasty restenosis, and hypertension. At least certain stages in the pathogenesis of these vascular diseases are characterized by accelerated proliferation of SMCs. 35,36 However, whether all medial cells can contribute equally to pathophysiological processes or, alternatively, whether there exist specific cell subpopulations that contribute selectively is still not clear. Studies in atherosclerosis, initially by Benditt and Benditt 37 and later by other investigators, 38–40 as well as experimental studies in primary pulmonary hypertension by Tuder et al 41 suggest selective participation of unique cell subtypes in the formation of pathological lesions in these diseases. Another possibility, raised by the present in vitro data, is that injury-induced stimulation of a vascular cell subpopulation with enhanced proliferative potential could contribute to vascular wall thickening not only through selective expansion of a specific cell subpopulation but also through secretion of mitogens by these cells and recruitment of other medial SMCs into the proliferative phenotype.

The isolation of cell subpopulations with markedly enhanced growth capabilities in vitro from the normal, mature, quiescent vascular media (for instance, L1-AUT cells) raises questions as to what controls the growth state of these cells in the intact vessel wall. In primary cultures, these cells were observed rather infrequently and were never found to “overgrow” other cell subtypes. We found that DNA synthesis of L1-AUT cells was dramatically inhibited by coculture with L2-SMCs even under serum-stimulated conditions (Table 1 ), suggesting that in vivo L1-AUT cells could be maintained quiescent by other SMC subtypes. This finding may account for the fact that bovine arterial SMCs have been cultured by investigators for over 20 years, yet to our knowledge, no other reports have described in culture phenotypically and functionally unique SMC subpopulations. We also found that even very low concentrations of heparin (1 μg/mL) dramatically inhibited serum-stimulated growth of L1-cells. The secretion of heparin-like molecules by endothelial or SMCs has been previously reported, 42 suggesting that this can be an additional candidate factor in maintaining the quiescent state of L1-cells in the normal intact vessel wall. Decreased production of inhibitory factors, such as might occur in vascular injury, could allow selective expansion of a specific cell subpopulation with enhanced proliferative potential.

If, as we suggest, the phenotypically and functionally distinct cell subpopulations we have isolated indeed represent cells of unique genetic lineages, the question arises as to the embryological origin of these cells. In avian embryonic arteries, unique cell types, which maintain distinct functional differences in culture, have been shown to be derived from either splanchnopleural mesoderm or cardiac neural crest. 28,43–45 In these studies, however, unlike ours, phenotypically and functionally unique arterial cell subpopulations were isolated from different segments of the systemic circulation (aortic arch versus abdominal aorta) rather than from the same arterial segment. A study by Bergwerff et al 29 demonstrated, based on the pattern of α-SM-actin expression, a clear phenotypic segregation of cell types, derived from neural crest versus splanchnic mesoderm, within the avian vascular media. We observed a similar developmental process of phenotypic segregation of distinct arterial cell types in the bovine species. 7 To our knowledge, however, no lineage analysis of SMCs in the vascular media of large mammals has been performed. An intriguing possibility that the cells observed in the subendothelial space of the avian arteries may arise from endothelial cells has recently been raised by DeRuiter et al 46 and discussed by Majesky and Schwartz. 47 Our findings of distinct cell subpopulations within arterial media at the same vascular site in the large mammalian species emphasize the need for further studies to determine the precise embryological origins of these cells.

The observation of multiple phenotypically and functionally distinct vascular SMC subpopulations raises important questions as to why cellular diversity exists and how it is maintained during normal vascular development. Numerous functions, including contraction, proliferation, and synthesis of extracellular matrix proteins, are required of vascular SMCs under both normal and pathological conditions. Diverse vascular functions may require different SMC phenotypes. Accordingly, it is highly likely that in response to pathological stimuli, some medial SMC subpopulations maintain, at least initially, their normal contractile function, whereas others exhibit increased proliferation and/or increased synthesis of extracellular matrix proteins. The current in vitro and previous in vivo studies strongly support the concept that numerous phenotypically unique subpopulations of cells exist in the vascular media, may perform different functions, and thus may contribute in unique ways to vascular homeostasis both under normal conditions and in vascular disease.


Cardiomyocytes bind, internalize, and activate prorenin, the inactive precursor of renin, via a mannose 6-phosphate receptor (M6PR)–dependent mechanism. M6PRs couple directly to G-proteins. To investigate whether prorenin binding to cardiomyocytes elicits a response, and if so, whether this response depends on angiotensin (Ang) II, we incubated neonatal rat cardiomyocytes with 2 nmol/L prorenin and/or 150 nmol/L angiotensinogen, with or without 10 mmol/L M6P, 1 μmol/L eprosartan, or 1 μmol/L PD123319 to block M6P and AT1 and AT2 receptors, respectively. Protein and DNA synthesis were studied by quantifying [ 3 H]-leucine and [ 3 H]-thymidine incorporation. For comparison, studies with 100 nmol/L Ang II were also performed. Neither prorenin alone, nor angiotensinogen alone, affected protein or DNA synthesis. Prorenin plus angiotensinogen increased [ 3 H]-leucine incorporation (+21±5%, mean±SEM, P<0.01), [ 3 H]-thymidine incorporation (+29±6%, P<0.01), and total cellular protein (+14±3%, P<0.01), whereas Ang II increased DNA synthesis only (+34±7%, P<0.01). Eprosartan, but not PD123319 or M6P, blocked the effects of prorenin plus angiotensinogen as well as the effects of Ang II. Medium Ang II levels during prorenin and angiotensinogen incubation were <1 nmol/L. In conclusion, prorenin binding to M6PRs on cardiomyocytes per se does not result in enhanced protein or DNA synthesis. However, through Ang II generation, prorenin is capable of inducing myocyte hypertrophy and proliferation. Because this generation occurs independently of M6PRs, it most likely depends on the catalytic activity of intact prorenin in the medium (because of temporal prosegment unfolding) rather than its intracellular activation. Taken together, our results do not support the concept of Ang II generation in cardiomyocytes following intracellular prorenin activation.

Cardiac angiotensin synthesis depends on renin of renal origin, both under normal and pathological conditions. 1–4 Circulating kidney-derived renin diffuses into the cardiac interstitial space, 5,6 and/or may bind to renin receptors. 7–9 In addition, the heart may sequester prorenin, the precursor of renin, from the circulation. Prorenin could contribute to angiotensin generation at cardiac tissue sites, either because of its inherent catalytic activity, which is the consequence of temporal prosegment unfolding, 10,11 or following its local conversion to renin. In support of this concept, we have recently demonstrated that cardiac myocytes and fibroblasts are capable of binding and internalizing recombinant human renin and prorenin via mannose 6-phosphate/insulin-like growth factor II (M6P/IGFII) receptors, and that prorenin following its internalization is rapidly activated to renin by proteolytic cleavage of the prosegment. 12,13

Recombinant as well as native human renin and prorenin contain the M6P signal that is required to bind to M6P/IGFII receptors. 14–16 These receptors also contain binding domains for IGFII and retinoic acid, 17,18 and binding of the latter agonists to M6P/IGFII receptors results in second messenger activation and growth inhibition, respectively. 19,20 Moreover, proliferin, which, similar to (pro)renin, binds to M6P/IGFII receptors via its M6P group, 21 has been reported to induce endothelial cell chemotaxis via these receptors in a G-protein- and mitogen-activated protein kinase-dependent manner. 22 In this respect, it is of interest to note that renin binding to mesangial cells resulted in enhanced 3 H-thymidine incorporation and plasminogen-activator inhibitor-1 (PAI-1) release, without intermediate angiotensin generation. 9

M6P/IGFII receptor-mediated accumulation of renin and activated prorenin in cardiac cells may result in intracellular angiotensin generation, and such intracellular angiotensin synthesis could underlie the stretch-mediated release of angiotensin II (Ang II) that has been demonstrated in myocytes. 23,24 However, to allow intracellular Ang II synthesis, the intracellular presence of angiotensinogen and ACE is also required, and in previous studies we were unable to detect angiotensinogen synthesis by cardiomyocytes. 25

In the present study, we set out to investigate whether prorenin binding, internalization, and activation by neonatal rat cardiomyocytes results in a cellular response, either directly (without intermediate Ang II generation), via binding to M6P/IGFII receptors, or indirectly, via the generation of Ang II and subsequent AT receptor activation. We measured protein and DNA synthesis following incubation of cells with recombinant human prorenin with or without angiotensinogen. Experiments were repeated in the presence of M6P, eprosartan, and PD123319, to antagonize M6P/IGFII -, AT1 -, and AT2 receptors, respectively. For comparison, we also studied the effect of Ang II in the presence and absence of these antagonists. Finally, we investigated Ang I and II generation during incubation of myocytes with prorenin and angiotensinogen, taking into consideration that prorenin itself displays catalytic activity (ie, without prosegment cleavage).


Cell Culture

All experiments were performed according to the regulations of the Animal Care Committee of the Erasmus University Medical Center Rotterdam, in accordance with the “Guiding Principles” of the American Physiological Society.

Primary cultures of neonatal Wistar rat (Harlan) cardiomyocytes were prepared as previously described. 12,13 Cells were seeded in noncoated 24-well plates (Corning Costar), giving a confluent monolayer of spontaneously beating myocytes at 1.5×10 5 cells/cm 2 after a 24-hour incubation in 1.5 mL medium (consisting of DMEM and Medium 199 [4:1], supplemented with 5% fetal calf serum [Life Technologies], 5% horse serum [Sigma], 100 U/mL penicillin, and 100 mg/mL streptomycin [Roche]). Thereafter, cells were incubated for 48 hours in medium supplemented with 5% horse serum and for 24 hours in serum-free medium. Before the start of each experiment, cells were rinsed 3 times with 1 mL warm (37°C) phosphate-buffered saline. Next, myocytes were incubated for 24 hours at 37°C with 250 μL serum-free medium, supplemented with 1% bovine serum albumin (BSA), and containing 100 U/L (≈2 nmol/L) recombinant human prorenin (a kind gift of Dr S. Mathews, Hoffmann-LaRoche, Basel, Switzerland) and/or 150 nmol/L human angiotensinogen (Sigma) in the presence or absence of 10 mmol/L M6P, 1 μmol/L eprosartan, or 1 μmol/L PD123319. For comparison, experiments with 100 nmol/L Ang II (Bachem) were also performed. Cells incubated without prorenin, angiotensinogen, or Ang II served as control.

Protein and DNA Synthesis

Protein and DNA synthesis rates were determined in triplicate by quantifying [ 3 H]-leucine and [ 3 H]-thymidine incorporation during the last 6 hours of the above 24-hour incubation period in the presence of Ang II or prorenin and/or angiotensinogen. 26 Total cellular protein and DNA were quantified after solubilization as described previously using BSA and salmon sperm as standard, respectively. 26

Angiotensin Generation

To measure angiotensin generation, myocytes were cultured in 6-well plates, and incubated at 37°C for 4 hours with 1 mL medium containing 10 or 100 U/L recombinant human prorenin and/or 150 nmol/L angiotensinogen. Cells incubated without prorenin or angiotensinogen served as control. After 1 hour and 4 hours of incubation, 75 μL medium was rapidly mixed with 6 μL inhibitor solution 27 and frozen at −70°C. Cells were collected after 4 hours (when the cellular levels of activated prorenin are maximal 12,13 ) as described before. 27 Ang I and Ang II levels in medium were measured by radioimmunoassay (detection limit 40 and 20 fmol/mL, respectively). 25,27 Ang I and II levels in cell homogenates were measured by radioimmunoassay after SepPak extraction and reversed-phase high-performance liquid chromatography separation (detection limit 0.4 and 0.2 fmol/10 6 cells). 25,27 For comparison, cellular angiotensin levels following a 4-hour incubation with 100 nmol/L Ang II were also measured.

Statistical Analysis

Data are expressed as mean±SEM. Statistical analysis was by ANOVA, followed by post-hoc evaluation according to Dunnett where appropriate. Statistical significance was accepted at P<0.05.


Protein and DNA Synthesis

The total cellular protein and DNA contents of myocytes, following a 24-hour incubation with vehicle, were 147±14 and 8.3±1.0 μg/well (n=15–16), respectively. Incorporation of 3 H-leucine and 3 H-thymidine during the last 6 hours of the 24-hour incubation period amounted to 27901±3707 and 57403±6262 dpm/well (n=16), respectively. None of the receptor blockers affected protein or DNA content or incorporation of 3 H-leucine and 3 H-thymidine (n=8 for each blocker, data not shown). Prorenin alone and angiotensinogen alone were without effect (Figure 1). Prorenin (2 nmol/L) combined with angiotensinogen increased total cellular protein (P<0.01), 3 H-leucine incorporation (P<0.01), and 3 H-thymidine incorporation (P<0.01), whereas Ang II increased the latter (P<0.01) only. The effects of prorenin combined with angiotensinogen were not observed at a prorenin concentration of 0.2 nmol/L (n=10, data not shown). All effects of prorenin combined with angiotensinogen as well as the effect of Ang II on 3 H-thymidine incorporation were blocked by eprosartan but not PD123319 or M6P (Figures 2 and 3 ). PD123319 tended to enhance the effects of Ang II on total cellular protein, but the difference was not significant (Figure 3).

Figure 1. Effect of 2 nmol/L prorenin, 150 nmol/L angiotensinogen, 2 nmol/L prorenin plus 150 nmol/L angiotensinogen, and 100 nmol/L angiotensin (Ang) II on total cellular protein content, total cellular DNA content, [ 3 H]-leucine incorporation, and [ 3 H]-thymidine incorporation in cardiomyocytes. Data are expressed as percentage change from control (mean±SEM of 8–16 experiments). PR indicates prorenin Aog, angiotensinogen. *P<0.01 versus control.

Figure 2. Effect of 2 nmol/L prorenin plus 150 nmol/L angiotensinogen on total cellular protein content, total cellular DNA content, [ 3 H]-leucine incorporation, and [ 3 H]-thymidine incorporation in cardiomyocytes in the presence of vehicle (none), the AT1 receptor antagonist eprosartan (Epro 1 μmol/L), the AT2 receptor antagonist PD123319 (1 μmol/L), or the M6P/IGFII receptor antagonist M6P (10 mmol/L). Data are expressed as percentage change from control (mean±SEM of 8 experiments). #P<0.05 versus none.

Figure 3. Effect of 100 nmol/L Ang II on total cellular protein content, total cellular DNA content, [ 3 H]-leucine incorporation, and [ 3 H]-thymidine incorporation in cardiomyocytes in the presence of vehicle (none), the AT1 receptor antagonist eprosartan (Epro 1 μmol/L), the AT2 receptor antagonist PD123319 (1 μmol/L), or the M6P/IGFII receptor antagonist M6P (10 mmol/L). Data are expressed as percentage change from control (mean±SEM of 8 experiments). #P<0.05, *P<0.01 versus none.

Angiotensin Generation

Ang I and Ang II were undetectable in medium or cells under control conditions and following incubations with either prorenin alone or angiotensinogen alone. At 1 hour after the addition of 2 nmol/L prorenin combined with 150 nmol/L angiotensinogen to the cells, Ang I and Ang II levels in the medium were 4278±207 and 372±23 pmol/L (n=3), and after 4 hours these levels amounted to 4704±462 and 795±102 pmol/L. When using 0.2 nmol/L prorenin in combination with 150 nmol/L angiotensinogen, the Ang I and Ang II levels in the medium were 656±161 and 85±14 pmol/L after 1 hour (n=3), and 876±43 and 129±18 pmol/L after 4 hours. Cellular Ang I levels measured after 4 hours of incubation with prorenin and angiotensinogen or 100 nmol/L Ang II were below the detection limit (n=3 for each condition). Cellular Ang II levels were also undetectable following a 4-hour incubation with 0.2 nmol/L prorenin combined with 150 nmol/L angiotensinogen. However, at a 10-fold higher prorenin concentration, as well as following a 4-hour incubation with 100 nmol/L Ang II, cellular Ang II levels amounted to 1.3±0.5 and 1.9±0.3 fmol/10 6 cells, respectively (n=3 for each condition). These levels represent <0.5% of the levels in the medium.


Prorenin binding to M6P/IGFII receptors located on the cell surface of neonatal rat cardiomyocytes does not result in enhanced protein or DNA synthesis. Enhanced protein and DNA synthesis were observed however when exposing the cells to prorenin combined with angiotensinogen, suggesting that these effects depend on angiotensin generation. Because prorenin binding to M6P/IGFII receptors is followed by internalization and rapid intracellular activation to renin, without subsequent release of renin to the medium, 12,13 we reasoned that intracellular angiotensin synthesis might underlie these findings. Such intracellular angiotensin generation will not occur in the absence of prorenin, because neonatal rat cardiomyocytes do not synthesize renin in detectable amounts. 25 However, the addition of 10 mmol/L M6P to the medium, which fully blocks prorenin internalization and activation, 12,13 did not block the prorenin plus angiotensinogen-induced effects on DNA and protein synthesis. Remarkably, blockade was observed with the AT1 receptor antagonist eprosartan, while the AT2 receptor antagonist PD123319, like M6P, was without effect. In view of the virtual lack of internalization of receptor antagonists, 28 the most likely explanation for these findings is that the inherent catalytic activity of prorenin, caused by temporal unfolding of the prosegment, results in extracellular Ang II generation and subsequent AT1 receptor stimulation. In support of this concept, Ang I and Ang II could be detected in nanomolar concentrations in the medium during incubation of the cells with 2 nmol/L prorenin and 150 nmol/L angiotensinogen.

The lack of effect of prorenin binding per se was unexpected, because several groups have reported that binding of M6P-containing glycoproteins to M6P/IGFII receptors results in a cellular response in a G-protein-dependent manner, eg, chemotaxis or increased c-fos expression. 19,22 Moreover, binding of renin to human mesangial cells was found to enhance 3 H-thymidine incorporation and PAI-1 release, independent of angiotensin generation. 9 The receptor mediating the latter effect has not yet been identified and is not necessarily the M6P/IGFII receptor. 9 Furthermore, the IGFII analog [Leu 27 ]IGFII, which binds to the M6P/IGFII receptor with equal affinity as IGFII, induced chemotaxis but not DNA synthesis. 29,30 Therefore, taken together, it is still very well possible that prorenin binding to M6P/IGFII receptors elicits other cellular responses than protein and DNA synthesis.

The addition of human angiotensinogen to the medium did not result in protein or DNA synthesis. This suggests that neonatal rat cardiomyocytes, like mouse cardiomyocytes, 31 do not posses enzymes (eg, cathepsins) capable of cleaving human angiotensinogen into Ang I and des-angiotensinogen. Ang I generation only occurred when combining human prorenin with human angiotensinogen, and this generation was limited to the extracellular compartment. The absence of intracellular Ang I, despite the proteolytic cleavage of prorenin to renin in myocytes, is in agreement with previous studies showing neither binding of angiotensinogen to cardiac and vascular membrane fractions, 1,8 nor angiotensinogen internalization. 27 Apparently, the intracellular presence of Ang II in myocytes must be explained on the basis of AT1 receptor-mediated endocytosis, 32–34 rather than intracellular Ang II generation. In support of this contention, Ang II was also detected in cell lysates following a 4-hour incubation with 100 nmol/L Ang II. Furthermore, the low cellular Ang II levels during incubation with prorenin plus angiotensinogen (<0.5% of the levels in the medium) also argue against synthesis and/or storage of Ang II in myocytes. 23

In the present study, extracellular Ang I generation occurred in a prorenin concentration-dependent manner. At the highest prorenin concentration tested (2 nmol/L), the Ang I levels in the medium reached a steady state within 4 hours. The levels were in the order of 5 nmol/L, which is within the range expected based on the concept that <2% of prorenin is catalytically active. 27 We previously reported that, in the absence of serum, Ang I–II conversion by ACE on myocytes is responsible for approximately 50% of Ang I metabolism by neonatal rat cardiomyocytes, and that the Ang I half life under these conditions is 1 hour. 25 The Ang II half life is much longer, 26 and this may explain why the medium Ang II levels in the present study continued to rise between 1 and 4 hours. Importantly however, the Ang II levels in the medium after 4 hours of prorenin plus angiotensinogen incubation were less than 1 nmol/L, and it is unlikely, in view of the steady-state Ang I levels, that these levels would have become much higher on longer incubation. Yet, despite these relatively low Ang II levels, the effects of prorenin combined with angiotensinogen on protein and DNA synthesis were equal to or stronger than those of 100 nmol/L Ang II. There are several explanations for this apparent discrepancy. First, Ang I–II conversion by ACE may occur in close proximity of AT1 receptors and may thus result in higher Ang II levels in the microenvironment of these receptors than in the medium. 35 Second, long-term exposure to low levels of Ang II (as a consequence of continuous Ang II generation) might be more efficient to induce cellular responses than short-term exposure to high levels of Ang II, for instance because the latter results in rapid downregulation of AT1 receptors. 36 Finally, because neonatal rat cardiomyocytes possess both AT1 and AT2 receptors, Ang II may also stimulate AT2 receptors, and this could counteract the AT1 receptor-induced effects. 26,37 It is possible that ACE-dependent local Ang II generation predominantly leads to AT1 receptor activation, because ACE is located in close proximity of AT1 receptors, 38 whereas exogenous Ang II results in equal AT1 and AT2 receptor activation. In agreement with this concept, as in our previous study, 26 the Ang II-mediated effect on total cellular protein increased in the presence of PD123319 (Figure 3).

The effects of locally generated and exogenous Ang II on protein and DNA synthesis rate were of modest proportion and exceeded those on total cellular protein and DNA content, suggesting that they may have been counterbalanced, at least in part, by protein and DNA degradation.

In conclusion, the partial catalytic activity of prorenin is responsible for the enhanced protein and DNA synthesis observed in cardiomyocytes during their incubation with prorenin and angiotensinogen. We found no evidence for intracellular angiotensin generation in these cells, nor did prorenin binding to M6P/IGFII receptors per se result in cell proliferation.

This study was supported by the Dutch Kidney Foundation, grant no. NSN C99.1874.


Rheumatoid arthritis (RA) is a chronic inflammatory disease characterized by T-cell mediated inflammation which contributes to the destruction of cartilage and bone in the joints [14–16, 25]. In animal models, CIA induced by injection of type II collagen in complete Freund’s adjuvant (CFA), shares a number of relevant features with human RA. By evaluating the molecular and cellular effects of green tea polyphenols or EGCG in different autoimmune models, these studies have brought to the forefront the beneficial effects of EGCG in modulating inflammation in animal models of experimental autoimmune encephalomyelitis (EAE) [18], lupus-like and other immune-mediated glomerulonephritis [17, 35], spontaneous non-obese diabetic mice [37] and Sjogren’s syndrome [38]. Our studies indicated that green tea extract or EGCG administration improved symptoms of arthritis, pathological features, and decreased serum CII-specific IgG2a antibody levels. In addition, EGCG treatment markedly reduced inflammation-related cytokine production including IFN-γ, IL-6 and TNF-α whereas it increased the production of IL-10. The present report adds mechanistic insights to previous observations that EGCG can be beneficial in arthritis models [20].

EGCG has been reported to modulate lymphocyte, neutrophil, macrophage and dendritic cell functions [16, 21, 39–43]. In in vitro cultures, EGCG altered B and T cell proliferative responses [16, 41, 43, 44]. Using a mixed lymphocyte culture assay it was evident that green tea extract significantly inhibited the proliferation of murine lymphocytes after stimulation with a potent T cell mitogen [45]. Our studies demonstrate that the total number of cells isolated from dLN were significantly lower in EGCG-treated than vehicle-fed CIA mice (P < 0.01). EGCG also suppressed the proliferation of autoreactive T cells and to a lesser extent, mitogen-stimulated T cells as measured by 3 H-thymidine incorporation. These studies parallel the changes observed in the CII antigen-specific T cell division assays using the tracking dye CFSE, which indicated that EGCG suppressed T cell proliferation. In a previous study, it was found that EGCG might inhibit T cell proliferation through modulating the IL-2/IL-2R system [43]. IL-2 can induce IFN-γ production and IL-2 blockade can lead to the inhibition of IFN-γ production [28, 43, 45], and this might underlie some of the observations in the current study. Recently, it was found that EGCG also inhibits B lymphocyte proliferation and induces B lymphocyte apoptosis [44]. Consistent with this, we found that EGCG administration reduced total dLN B cell (CD5 − B220 + ) frequencies as well as percentages of major splenic B cell subsets including follicular (Fo, CD21 int CD23 hi ) and marginal zone (MZ, CD21 hi CD23 lo ) B cells. Taken together, we conclude that EGCG suppresses both T and B cell expansion induced by CII in arthritic mice as EGCG effected both the frequency and absolute numbers of cells.

Previous reports have described the correlation between the decreased function and/or percentage of CD4 + CD25 + Treg cells in patients with RA and clinical disease activity [46]. Moreover, CD4 + CD25 + T cells isolated from arthritic animals were capable of exerting suppressor function in in vitro assays [47], while it has been shown that the depletion of CD4 + CD25 + cells could lead to the spontaneous development of autoimmune diseases and increased severity of symptoms in CIA mice [48]. In an EAE model, EGCG also reduced the production of IFN-γ, IL-17, IL-6, IL-1β and increased Treg numbers in lymph nodes and spleen [18]. In agreement with previous findings, the current study showed that EGCG–fed mice exhibited increased percentages of CD4 + CD25 + Foxp3 + Treg cells when compared with vehicle-fed CIA mice. These results are also consistent with the findings of other investigators who have shown that in vitro treatment with EGCG modestly enhances Foxp3 and IL-10 mRNA expression in a CD4 + T cell line [49].

This study is in agreement with previous reports that IDO expressing innate immune cells can help generate Tregs. For example, human IDO expressing plasmacytoid DC triggered by TLR ligation induced the generation of CD4 + CD25 + Foxp3 + Tregs from CD4 + CD25 − T cells [50]. Other studies also support this result as CD11c + CD11b + DCs isolated from peyer's patches of orally tolerized mice with type II collagen appeared to be necessary for the expansion and differentiation of CD4 + CD25 + T cells, which suppressed CII-specific T-cell proliferation [31]. These studies support our observation that CD11b + DCs isolated from EGCG-fed mice express high levels of IDO and can facilitate the generation of Ag-specific CD4 + CD25 + Foxp3 + Tregs. Blocking IDO activity with the specific inhibitor, 1-MT, significantly abrogated the proportion of CD4 + CD25 + Foxp3 + T cells induced by CD11b + DCs from EGCG-fed CIA mice. Thus, these data support the hypothesis that EGCG-induced IDO-expressing CD11b + DCs can generate Tregs from CD4 + CD25 − cells in CIA mice.

Previously, antioxidants and antioxidative enzymes have been shown to reduce cartilage damage in animal models of RA, with Nrf2 being a major player [51, 52]. Moreover, Nrf2 deficiency leads to an acceleration of the effector phase of arthritis [52]. It has also been reported that Nrf2 activity inversely correlates with disease in RA patients [53]. Our data demonstrate that EGCG treatment significantly increased pNrf2 activity, and increased expression of HO-1, a Nrf2 target gene. Devesa and colleagues have reported that induction of HO-1, which is protective against joint destruction, can exert partial anti-arthritic effects in CIA [54]. Recently, it has been found that pharmacological up-regulation of HO-1 causes a robust anti-inflammatory response in a model of non-autoimmune arthritis in mice [55], and might prove to be a novel therapeutic target in treatment of chronic inflammatory diseases. Thus, the finding that EGCG enhanced Nrf2 activity resulting in increased levels of HO-1 is of considerable significance.

Of note, activation of Nrf2 in T cells by tert-butylhydroquinone (tBHQ), inhibits production of the Th1 cytokine, IFN-γ [56]. Maicas et al. reported that deficiency of Nrf2 resulted in increased migration of pro-inflammatory cells into the joints during the development of arthritis, with significant elevations in TNF-α and IL-6 levels compared with wild type controls [51]. This is also consistent with reports indicating that CD4 + T cells from Nrf2 null mice secreted increased amounts of IFN-γ whereas levels of IL-4, IL-5 and IL-13 are decreased [51]. In addition, several reports have demonstrated that a deficiency in Nrf-2 activity results in greater sensitivity to oxidative and inflammatory disorders such as asthma, colitis and sepsis [57–59]. Recently, a chronic granulomatous disease (CGD) patient was found with undetectable IDO metabolic activity, increased Th17 cells as well as impaired transcription factor Nrf2 activity [60]. CGD is an inherited immunodeficiency characterized by a hyper-inflammatory response and an inability to produce reactive oxygen intermediates (ROI), which might lead to impaired counter-regulation by the IDO pathway and insufficient Nrf2 activation [32, 61]. Interestingly, our data revealed that EGCG-fed CIA mice had significantly decreased levels of IFN-γ IL-1β, IL-6 and TNF-α and increased IL-10 levels in joint homogenates and serum as compared to vehicle-fed CIA mice supporting the observation that increased Nrf2 activity correlates with suppression of the inflammatory response.

Collectively, our studies and previous work support a model whereby EGCG-induced Nrf2 activation may skew T cells from a Th1/Th17 phenotype to a Th2 and Treg phenotype and this finding warrants further investigation. The current study is the first to report a relationship between the effects of EGCG treatment and the induction of IDO expression, an activity which can then upregulate antioxidant pNrf-2 activity in mice with arthritis. Further studies are required to determine the clinical relevance of these findings and a systematic testing of potential therapeutic targets in this regulatory cascade.

Results and Discussion

Labeling of Cellular DNA with EdU.

Cells incorporate 5-bromo-2′-deoxyuridine (BrdU) and 5-iodo-2′-deoxyuridine (IdU) into their DNA in the place of 5-methyl-2′-deoxyuridine (thymidine). We reasoned that 5-ethynyl-2′-deoxyuridine (EdU) should also be efficiently incorporated into replicating DNA and its terminal alkyne group should be available to react with organic azides in the major groove of the double helix, without steric hindrance. We synthesized EdU and fluorescent azide derivatives required for its detection in cells by using click chemistry (Fig. 1). Cells labeled in culture with 10 μM EdU overnight showed very intense nuclear staining after reaction with Alexa568-azide under Cu(I)-catalyzed click reaction conditions (Fig. 2 B, E, and H). In contrast, cells not labeled with EdU displayed no detectable staining if subjected to the same fixation and staining conditions as EdU-labeled cells (Fig. 2 A, D, and G). EdU staining was greatly reduced in cells in which DNA replication was blocked by treatment with hydroxyurea (Fig. 2 C, F, and I see also J–M), demonstrating that the nuclear staining in EdU-labeled cells is the result of EdU incorporation during DNA synthesis. Also, in synchronized HeLa cells, the extent and intensity of DNA labeling after a pulse of EdU during S phase correlates with the EdU concentration and the length of the pulse (data not shown). EdU has been previously evaluated as an antiviral compound (7). These early studies concluded that EdU was not incorporated into bacterial and phage DNA (8) its incorporation into eukaryotic DNA, however, was not investigated. We find that EdU is strongly incorporated into the DNA of proliferating mammalian cells. EdU is also incorporated into DNA replicated in vitro in Xenopus egg extracts (data not shown, see Materials and Methods) by simple addition to the cycling extract preparation.

Use of 5-ethynyl-2′-deoxyuridine (EdU) to label DNA in cells. (A) Structure of 5-ethynyl-2′-deoxyuridine, a thymidine analogue that carries a terminal alkyne group instead of a methyl in the 5 position of the pyrimidine ring. (B) Schematic of the click reaction for detecting EdU incorporated into cellular DNA. The terminal alkyne group, exposed in the major groove of the DNA helix readily reacts with an organic azide (R can be any fluorophore, hapten, electron-dense particle, quantum dot, etc.) in the presence of catalytic amounts of Cu(I). (C) Structures of the fluorescent azides used in the present study. Unlike azide 2, azide 3 is cell-permeable, allowing cells to be stained while alive, without fixation and/or permeabilization. Fluorescein azide 4 can be easily prepared in large amounts by using inexpensive starting materials. See text and Materials and Methods for details.

Detection of EdU incorporated into the DNA of cultured NIH 3T3 cells (A–M) and HeLa cells (N–P) by fluorescence microscopy. NIH 3T3 cells were incubated in media without EdU (A, D, G, J, and L), media supplemented with10 μM EdU (B, E, and H) or media with 10 μM EdU and 10 mM hydroxyurea to block DNA synthesis (C, F, I, K, and M). In A–M, the cells were fixed and then reacted with 10 μM Alexa568-azide (Fig. 1 C, compound 2) for 10 min. The cells were then counterstained with Hoechst to reveal cellular DNA, washed, and imaged by fluorescence microscopy and differential interference contrast (DIC). Note the strong specific and low nonspecific azide stain in the presence (H) and absence (G) of EdU, respectively. Not all nuclei in H are labeled after overnight incubation with EdU, suggesting that only cells that went through S phase became labeled. Blocking DNA replication with hydroxyurea abolishes EdU incorporation almost completely (I). If cells labeled with EdU in the presence of hydroxyurea are photographed with longer exposure times (3-s exposure time in L and M, compared with 40-ms exposure time for G–I under otherwise identical illumination and camera settings), low levels of EdU incorporation can be seen in a small fraction of the hydroxyurea-treated cells. (N–P) Still images from a time-lapse recording of EdU staining of live cells by using the cell-permeable TMR-azide (Fig. 1 C, compound 3). Live HeLa cells labeled with 10 μM EdU were stained with 500 nM TMR-azide in the presence of Cu(I) in PBS. Time is shown in minutes in the upper right corners of M–P. Note that Cu(I) is cytotoxic and the cells do not survive the staining reaction. See text and Materials and Methods for details.

The [3 + 2] cycloaddition reaction used to detect EdU in cells is catalyzed by Cu(I) ions. Because of the low solubility of Cu(I) salts, Cu(I) is often introduced in [3 + 2] cycloaddition reactions in a liganded form (9). We found that the best EdU staining was obtained by generating Cu(I) from Cu(II) in situ, using ascorbic acid as reducing agent. The intensity of the staining obtained was proportional to the concentration of the fluorescent azide and that of ascorbic acid in the staining reaction (data not shown). In the case of ascorbic acid, for a fixed concentration of Cu(II) of 1 mM (in the form of CuSO4), the intensity of the EdU stain increased as the ascorbic acid concentration was increased to 50–100 mM.

The staining reaction between cellular EdU-labeled DNA and fluorescent azides proceeds very quickly, being half-maximal in <5 min under the staining conditions we used (see below). This fast kinetics might be because EdU-labeled DNA is a multivalent, polymeric click chemistry substrate. The triazoles formed from the reaction between the ethynyl group and the fluorescent azide are Cu(I) chelators, increasing the local concentration of the catalyst and accelerating the reaction of neighboring ethynyl groups. This effect has been described for other multivalent [3 + 2] cycloaddition substrates (10).

EdU detection proceeds smoothly with a variety of fluorescent azides we synthesized, including aliphatic azides (compounds 2 and 3 in Fig. 1) and aromatic ones (compound 4 in Fig. 1, which can be easily synthesized in a single step in large amounts as required for some whole-mount staining procedures). The EdU detection method is also compatible with immunostaining. We recommend performing a regular immunostaining protocol after first performing the EdU stain (see below).

We find that under the conditions used (1 mM CuSO4, 100 mM ascorbic acid, and 10 μM fluorescent azide), the staining reaction does not proceed to completion and only a fraction of the ethynyl groups present on DNA is reacted before the staining mixture loses activity. As shown in Fig. 3 A, if EdU-labeled cells first stained with Alexa488-azide (green) are subsequently reacted with fresh staining mixture containing Alexa594-azide (red), the DNA is strongly stained both green and red. The two staining patterns display perfect colocalization (see Fig. 3 A Right, showing an overlay of the two successive EdU stains), demonstrating the high reproducibility of the EdU detection method. The intensity of the EdU stain can thus be increased through repeated incubation with fresh detection mixture, without a change in the staining pattern.

Reproducibility of EdU labeling (A) and comparison with BrdU (B). (A) NIH 3T3 cells labeled by incubation overnight with 2 μM EdU were fixed and reacted successively with 10 μM Alexa488-azide and 10 μM Alexa594-azide, respectively, as described in Materials and Methods. The cells were imaged by fluorescence microscopy. (Left) Alexa488-azide stain. (Center) Alexa594-azide stain. (Right) Overlay of the Alexa488 and Alexa594 images. The complete colocalization of the two colors indicates that the two successive azide stains detect ethynyl groups that have the same distribution within the cell nucleus. (B) NIH 3T3 cells labeled by incubation overnight with 2 μM EdU and 2 μM BrdU were fixed and reacted with 10 μM Alexa594-azide, followed by standard BrdU immunodetection by using an anti-BrdU monoclonal antibody and an Alexa488-conjugated secondary antibody. (Left) BrdU stain. (Center) EdU stain. (Right) Overlay of the BrdU and EdU images. Note that each cell that incorporated BrdU also incorporated EdU. The BrdU micrograph was taken by using a five-times longer exposure than the exposure used for EdU (500 ms compared with 100 ms) under identical digital camera settings.

We next asked how EdU compares with the conventional BrdU-labeling method. Cultured NIH 3T3 cells were labeled with both EdU and BrdU by overnight incubation in media containing 2 μM each of the two deoxynucleoside analogues. The cells were then fixed, stained with Alexa594-azid,e and then processed for BrdU immunofluorescent detection. As shown in Fig. 3 B, all cells staining with BrdU also show EdU staining. In addition, the EdU staining is significantly more intense than the BrdU staining.

Detection of EdU Without Fixation.

The Alexa568-azide used earlier is not cell-permeable and required cell fixation and permeabilization to stain EdU-labeled cells. To detect EdU-labeled DNA under native conditions (i.e., without fixation and permeabilization) we developed a cell-permeable tetramethylrhodamine (TMR) azide (Fig. 1). After confirming that TMR-azide can enter live cells, we used it to stain live cells that had been labeled with EdU. The staining mixture containing TMR-azide was added to cells on a fluorescent microscope stage at time t = 0, to initiate the staining reaction, which was then followed by fluorescence confocal microscopy. As shown in Fig. 2, TMR-azide reacts rapidly with EdU-labeled DNA in cells, in the presence of Cu(I) generated in situ from Cu(II) and ascorbic acid, suggesting that the click reaction takes place efficiently on native cellular DNA. The staining reaction is half-maximal in <5 min. This experiment underestimates the staining rate, because we used only 500 nM TMR-azide in the detection reaction (as opposed to the usual 10–50 μM azide concentration we normally use for staining). The lower TMR-azide concentration used in this experiment was required to allow imaging of cellular DNA by fluorescence microscopy, against the background of unreacted TMR-azide in the staining solution.

It should be noted that Cu(I) is toxic to cells and that Cu(I)-catalyzed click reaction conditions result in cell death. We believe, however, that even if the staining conditions were nontoxic, live cell microscopy of EdU-labeled and azide-stained cells would be of limited utility. The reaction between ethynyl groups on DNA and fluorescent azide molecules results in covalent attachment of many large chemical groups to cellular DNA. If cells were to survive staining, such a chemical insult would likely trigger a massive DNA damage response and subsequent cytotoxicity. Despite Cu(I) toxicity, we envision that the ability to stain EdU-labeled cells rapidly and without fixation will be useful in a number of situations: (i) when permeabilization and fixation degrade the structure of the specimen (ii) for assaying DNA synthesis at the end of live cell experiments, without removing the cells from the microscope stage (iii) in high-throughput screening assays, when reducing the number of steps in the staining protocols is important. Additionally, staining cells without fixation and permeabilization can prevent cell detachment and loss from coverslips.

Use of EdU to Assay DNA Synthesis in Animals.

We next wanted to determine whether EdU can be used to detect DNA synthesis in animals. One hundred micrograms of EdU in PBS were injected i.p. into adult mice and tissues were harvested and fixed 96 h after injection. Paraffin sections were stained with Alexa568 azide for 15 min, washed, and then imaged by fluorescence microscopy. Because of the high intensity of fluorescence, large sections can be quickly scanned by using low-magnification objectives, even a simple dissecting microscope equipped with fluorescence. As shown in Fig. 4, EdU strongly labels proliferating cells in vivo (Fig. 4 A and B, showing sections of small intestine). It is well suited for detecting very low levels of cell proliferation in tissues that undergo very little to no cellular turnover (Fig. 4 C and D, showing sections of mouse brain).

Labeling DNA in vivo by using EdU. An adult mouse was injected i.p. with 100 μg of EdU in PBS. Tissues were harvested, fixed, and sectioned 96 h later. Tissue sections on slides were then reacted for 10 min with 10 μM Alexa568-azide. The images shown are overlays of a DIC image of the sectioned tissue, a fluorescent image of cellular DNA (Hoechst stain blue), and a fluorescent image of EdU-labeled DNA revealed by reaction with Alexa568 azide (red). (A and B) Mouse small intestine. Villi are seen in transverse section in A and in longitudinal section in B. The cells with red nuclei are descended from cells that had been in S phase during the EdU pulse and thus incorporated EdU into their DNA. (C and D) Mouse brain. The vast majority of nuclei on brain sections were not labeled with EdU, confirming that the label does not detectably incorporate into the DNA of nondividing cells. Each of the two images shows a field of cells containing an EdU-labeled nucleus, belonging to a cell of undetermined type.

Finally, we wanted to determine whether EdU can be used to detect cell proliferation in large, fresh tissue and organ explants. Mice were injected i.p. with 100 μgrams EdU and their small intestine was harvested 24 or 96 h later. The fresh intestine was then directly stained with the cell-permeable TMR-azide, followed by fixation and washing to remove unincorporated TMR-azide. As shown in Fig. 5, after 24 h, EdU staining can be detected strongly in cells present at the base of the villi, which is the location described for the actively dividing cells of the small intestine mucosa. Ninety-six hours after the EdU pulse, the labeled cells have multiplied and migrated distally, away from the base of the villi and toward their tips. This whole-mount imaging of cellular turnover in small intestine villi is consistent with classical measurements and demonstrates that EdU can be used to assay cells proliferation in fresh tissues rapidly and with high sensitivity. In this context, EdU allows the rapid and sensitive staining of large organ fragments in whole mounts this will make the study of organ and tissue dynamics much easier than by earlier methods.

Exploring cell proliferation and tissue dynamics in animals using EdU. Whole-mount fluorescent images of mouse small intestine, stained to detect EdU incorporation. Two adult mice were each injected i.p. with 100 μg of EdU in PBS and their small intestines were removed after 24 and 96 h, respectively. A freshly harvested 2-cm-long segment of the small intestine (A) was opened with a longitudinal cut, rinsed, and immediately stained for 10 min with 100 μM TMR-azide in the presence of Cu(I), without fixation or permeabilization. The intestine piece was then fixed, washed to remove unreacted TMR-azide, and imaged at low magnification on a dissecting microscope equipped with fluorescence. (B and C) After 24 h, the EdU shows strong incorporation in cells located at the base of each villus. (D and E) Ninety-six hours after the EdU pulse, the labeled cells have moved away from the base, near the tips of the villi. See text and Materials and Methods for details.

In some experiments it would be desirable to perform two different DNA-labeling pulses and be able to distinguish between the two. We reasoned that the alkyne and azide functionalities can be swapped between the deoxynucleoside analogue and detection reagent to allow two-color DNA staining based on click chemistry. We synthesized 5-azido-3′-deoxyuridine (AdU) as described for 5-azido-dUTP (11). Like EdU, this analogue was incorporated into cellular DNA. We detected it by reacting with a terminal alkyne conjugated to the Alexa568 fluorophore, under staining conditions identical to those used for EdU (see Materials and Methods). Although we were able to detect AdU-labeled DNA in cells by fluorescence microscopy, the background was, for unknown reasons, much higher than in the case of EdU (data not shown), as reported for other instances in which click reactions were performed on cells metabolically labeled with azides (12). The routine use of AdU as a second color for DNA labeling will have to await the development of a detection protocol that affords lower background staining. The combination of BrdU and EdU (Fig. 3 B), meanwhile, offers a simple and robust way to perform double-DNA labeling.

We envision that the superior structural preservation allowed by EdU relative to BrdU staining should facilitate high-resolution electron microscopy studies of cellular chromatin. For example, EdU-labeled DNA can be imaged by electron microscopy by using an azide conjugated to eosin (or another fluorophore that generates singlet oxygen efficiently), followed by photoconversion of diaminobenzidine into an osmiophilic precipitate (13). Immunoelectron microscopy of EdU-labeled chromatin under nondenaturing conditions can also be achieved by using an azide conjugated to a good epitope (such as digoxigenin) followed by immunostaining.

A recent method called photoactivation localization microscopy (PALM) (14) generates nanometer-resolution images of biological molecules in cells by repeated cycles of photoactivation of “caged” fluorophores followed by imaging until fluorophore bleaching. The amount of fluorescent signal generated during EdU detection can be easily controlled (by varying ascorbic acid and fluorescent azide concentrations, and by the ability to perform successive stains), in a manner similar to photoactivation of fluorescence used for the PALM technique. The EdU-labeling method is thus well suited for optical imaging of cellular DNA at nanometer resolution.

Supplementary Material

We are indebted to Marina Bacac, Carlo Fusco, and Roland Meier for invaluable discussions and critical reading of the manuscript. The help and suggestions of Werner Held and Pedro Romero regarding NK experiments are gratefully acknowledged. We also thank Cynthia Dayer for help with fluorescence stereomicro­scopy. L.A. Mauti was supported by a MD-PhD grant from the Swiss National Foundation. This work was supported by Swiss National Science Foundation grant 31003A-105833 and the NCCR Molecular Oncology (to I. Stamenkovic).


ZFP36 dynamics during T-cell activation

ZFP36 expression is induced upon T-cell activation (Raghavan et al., 2001). We examined its precise kinetics following activation of primary mouse CD4 +T cells by Western analysis with custom ZFP36 antisera generated against a C-terminal peptide of mouse ZFP36. Protein levels peaked

4 hr post-activation and tapered gradually through 72 hr, and were re-induced by re-stimulation 3 days post-activation (Figure 1A). ZFP36 expression depended on both TCR stimulation, provided by anti-CD3, and co-stimulation, provided by co-cultured dendritic cells (DCs) (Figure 1B). A similar pattern of transient ZFP36 induction occurred in activated CD8 +T cells (Figure 1—figure supplement 1A).

HITS-CLIP as a transcriptome-wide screen for ZFP36 function in T cells.

(A) Immunoblots with pan-ZFP36 antisera after activation of naïve CD4 +T cells in DC co-cultures, and with re-stimulation at day 3. Antibody and MW markers are shown on the left. NS* indicates a non-specific band. (B) Immunoblotting with pan-ZFP36 antisera 4 hr after activation of naïve CD4 +T cells, testing dependence on TCR stimulation (α-CD3), and co-stimulation (DCs or α-CD28). (C) ZFP36 HITS-CLIP design. (D) Representative autoradiogram of ZFP36 CLIP from activated CD4 +T cells using pan-ZFP36 antisera, with pre-immune and no-UV controls. Signal in Zfp36 KO cells is due to capture of ZFP36L1 RNP complexes. (E) The most enriched binding motifs and (F) annotation of binding sites from WT and Zfp36 KO cells. (G) Overlap of binding sites in WT and Zfp36 KO cells, stratified by peak height (PH). CLIP data are compilation of 4 experiments, with 3–5 total biological replicates were condition. (H) RNAseq in WT and Zfp36 KO CD4 +T cells activated under Th1 conditions for 4 hr. Log2-transformed fold-changes (KO/WT) are plotted as a cumulative distribution function (CDF), for mRNAs with 3’UTR, CDS, or no significant ZFP36 HITS-CLIP sites. Numbers of mRNAs in each category (n) and p-values from two-tailed Kolmogorov-Smirnov (KS) tests are shown. RNAseq data is a compilation of 2 experiments, with 3–4 biological replicates per condition.

Western blot analysis showed multiple bands at

40–50 kD, indicating several isoforms. Notably, isoforms running above the predicted molecular weight (MW) of ZFP36 (36 kD) pre-dominated early after activation, and are consistent with previously reported hyperphosphorylation (Qiu et al., 2012). In addition, partial conservation of the immunizing peptide in ZFP36L1 and ZFP36L2 raised the possibility of paralog cross-reactivity. Western analyses with recombinant constructs confirmed ZFP36, ZFP36L1, and ZFP36L2 are readily detected with our custom antisera (henceforth, pan anti-ZFP36 Figure 1—figure supplement 1B). Commercial paralog-specific antibodies were identified, and Western analysis showed that both ZFP36 and ZFP36L1 were induced by T-cell activation (Figure 1—figure supplement 1B–C). ZFP36L2, expected to run at

62 kD, was not detected under any conditions examined. Analysis of Zfp36 KO T cell lysates with pan ZFP36 antisera showed

50% reduced signal compared to WT. We conclude that the residual signal is likely due to persistent expression of ZFP36L1, which is highly homologous and of similar size to ZFP36. Collectively, these results demonstrate activation-dependent expression of ZFP36 and ZFP36L1 in T cells, and suggest activated Zfp36 KO T cells have partial loss (

The characterization of Zfp36 as an immediate early response gene in various cell types established transcription as a mechanism of its activation-induced expression (Lai et al., 1995). Interestingly, Zfp36 (RPKM = 22.5) and Zfp36l1 (RPKM = 8.2) mRNAs are robustly present in RNAseq data from naïve CD4 +T cells, despite an absence of detectable protein by western blotting. ZFP36L2 was not detected in any conditions examined, but its mRNA was also detected in naïve cells (RPKM = 30.2). These observations indicate that post-transcriptional mechanisms regulate expression of ZFP36 paralogs in T cells.

Transcriptome-wide identification of ZFP36 target RNAs in CD4 +T cells

The striking pattern of ZFP36/L1 expression in T cells led us to develop ZFP36 HITS-CLIP as a screen for its biological function (Figure 1C–F Figure 1—figure supplement 2). Notably, ZFP36/L1 RNPs isolated by CLIP from activated CD4 +T cells sera exhibited high molecular weight (MW) complexes resistant to detergent, heat, and RNAse, consistent with a pattern previously observed in ZFP36 CLIP in macrophages (Figure 1D, Figure 1—figure supplement 2A [Sedlyarov et al., 2016]). This RNP signal pattern was dependent on UV irradiation, and was observed with two different anti-pan-ZFP36 but neither pre-immune sera.

Given prior evidence that ZFP36 regulates T-helper type-1 (Th1) cytokines (e.g. TNF-α and IFN-γ), we next generated a comprehensive map of ZFP36/L1 binding sites by HITS-CLIP using anti-pan-ZFP36 in WT CD4 +T cells, activated for 4 hr under Th1-polarizing conditions (Ogilvie et al., 2009 Carballo et al., 1998). 5132 robust binding sites were defined, requiring a peak height (PH) >5, and support from at least 3 (of five total) biological replicates and two different pan-ZFP36 antisera (Supplementary file 1A [Shah et al., 2017 Moore et al., 2014]). Consistent with identification of bonafide ZFP36/L1 binding events, HITS-CLIP recovered the known AU-rich ZFP36 consensus motif at high significance, along with reported binding sites in Tnf, Ifng and other targets (Figure 1E Supplementary file 1A [Brewer et al., 2004]). Globally, ZFP36/L1 sites confirmed a preponderance of 3’-UTR binding (>75%), and showed substantial binding in coding sequence (CDS

6.5%) and introns (5.4%) (Figure 1F). Separate analysis of low and high MW RNP complexes showed similar transcript localization and motif enrichment, all consistent with ZFP36 binding (Figure 1—figure supplement 2B). This analysis indicates the presence of large, stable ZFP36 complexes in vivo, consistent with stable multimers (Cao et al., 2003 2004). Subsequently, CLIP reads from different MW regions were pooled to maximize dataset depth.

To examine possible paralog specificity, we also mapped ZFP36L1 sites by HITS-CLIP in Zfp36 KO CD4+T cells under identical conditions. As in western analysis (Figure 1—figure supplement 1C–D), Zfp36 KO samples showed reduced but significant CLIP signal compared to WT (Figure 1D, Figure 1—figure supplement 2A), representing ZFP36L1-RNA complexes. Sites in WT and Zfp36 KO cells showed very similar enriched motifs and transcript localizations, indicating that ZFP36 and ZFP36L1 have similar binding profiles in vivo (Figure 1E–F, Figure 1—figure supplement 2C, Supplementary file 1B). Majorities of robust sites (53%) and target mRNAs (66%) identified in WT cells were found independently in Zfp36 KO cells, and site overlap was far greater (>90%) for peaks of increasing magnitude (Figure 1G). A subset of potential ZFP36L1-specific sites was identified only in Zfp36 KO cells (n = 675 Supplementary file 1C), although these showed similar features to ZFP36 sites overall (Figure 1—figure supplement 2C third panel). Thus, these analyses do not exclude paralog specificity at some sites, but indicate broadly overlapping in vivo RNA binding for ZFP36 and ZFP36L1 reflecting their high homology.

Secondary enriched motifs revealed additional properties of ZFP36/L1 target sites. The second top motif resembled the known recognition sequence for polyadenylation factor CFI(m)25 (Venkataraman et al., 2005). Accordingly, ZFP36 binding in 3’UTRs was most concentrated in the vicinity of expected polyA sites,

50 nucleotides before transcript ends (Figure 1—figure supplement 2D). Analysis of CDS-specific binding revealed the AU-rich ZFP36 motif, along with strong enrichment of the 5’ splice site (5’-ss) consensus (Figure 1—figure supplement 2E). Cross-link-induced truncations (CITS) clarified that CDS peaks are centered within coding exons, but supporting CLIP reads often spanned the exon-intron boundary. Thus, at least a subset of CDS binding by ZFP36 occurs prior to pre-mRNA splicing in the nucleus.

ZFP36 represses target mRNA abundance and translation during T-cell activation

We next employed RNA profiling strategies to determine the functional effects of ZFP36/L1 binding. RNAseq analysis in WT and Zfp36 KO CD4 +T cells activated under conditions identical to our HITS-CLIP analyses uncovered two main effects. First, a small number of mRNAs that were silent in WT cells, including numerous immunoglobulin loci, were detected at low to moderate levels in KO cells (Figure 1—figure supplement 3A). These mRNAs lacked evidence of ZFP36/L1 binding, both in HITS-CLIP data and direct motifs searches, suggesting they are not direct targets. Given established chromatin regulation of many of these loci (i.e. Ig genes) and the on-off nature of the changes, dysregulated transcriptional silencing is a potential explanation, but is likely to be a secondary effect of ZFP36 loss of function. The second main effect emerged from a global analysis, which showed ZFP36 binding in 3’UTR (p=4.44×10 −16 Kolmogorov-Smirnoff [KS]) and CDS (p=5.33×10 −11 ) correlated to subtle but highly significant shifts toward greater mRNA abundance in Zfp36 KO cells, relative to mRNAs with no binding sites (Figure 1H). This correlation was not observed for mRNAs with binding exclusively in introns, and we did not find evidence in these data that ZFP36/L1 binding correlated with altered usage of proximal splice or polyA sites (not shown). The same pattern was also observed when considering a more stringently defined of sites set overlapping statistically robust CITS (Figure 1—figure supplement 3B).

The overall trend in transcriptome profiling is consistent with evidence that ZFP36 represses RNA abundance (Lykke-Andersen and Wagner, 2005). However, stratification of sites by the magnitude of ZFP36 CLIP binding allowed resolution of potentially complex effects. For 3’UTR binding, ZFP36 targets overall showed a significant shift in abundance, but mRNAs containing the top 20% most robust sites (ranked by peak height [PH], see Materials and methods) showed no significant effect. Thus, a higher degree of binding correlated with less effect on RNA abundance in the absence of ZFP36 (Figure 1—figure supplement 3C). This trend was not observed for CDS sites, where the top 20% showed a similar shift to sites overall (Figure 1—figure supplement 3D). Thus, our analyses show a trend of negative regulation of RNA abundance in this context, but with blunted effects for highly robust binding sites in 3’UTR (see Discussion).

We next examined in more detail the effects of ZFP36 regulation for HITS-CLIP targets with highly robust 3’UTR binding in T cells. Activation marker CD69, apoptosis regulator BCL2, and effector cytokines TNF and IFNG showed significantly increased protein levels in Zfp36 KO versus WT T cells 4 hr after activation (Figure 2A, Figure 2—figure supplement 1). Of these, only Bcl2 showed increased mRNA abundance. Tnf, Ifng, and Cd69 were all among the top 20% of targets as defined by CLIP binding magnitude (PH), thus supporting the trend in our global analyses that some highly robust binding targets show little regulation at the level of mRNA abundance in this context. The effects on protein level in the absence of changes in mRNA abundance suggested regulation of translation. We tested this principle by constructing GFP fluorescent reporters with an intact 3’UTRs (WT-UTR), or variants with the CLIP-defined ZFP36 binding site deleted (Δ-UTR Figure 2B), for Cd69, Tnf, and Ifng. In 293 cells, ZFP36 strongly repressed protein expression for all three WT-UTR reporters, while showing weaker (Tnf and Ifng) or no (Cd69) repression of mRNA levels. Of note, the Δ-UTR constructs showed increased protein levels both in the presence and absence of ZFP36. This indicates that additional, endogenous factors are regulating these AU-rich sites in 293 cells, though Western analyses confirmed that ZFP36 paralogs are undetectable (Figure 1—figure supplement 1A). In addition, ZFP36 over-expression exerted

2 fold repression of Δ-UTR constructs, which may indicate incomplete ablation of binding or secondary effects of ZFP36 over-expression. However, repression of WT-UTR constructs was consistently greater than for Δ-UTR variants, demonstrating specific ZFP36 repression of the defined binding sites. In summary, these heterologous assays independently confirmed ZFP36 regulation of CLIP-defined targets, and support specific effects of translation, in addition to RNA stability.

ZFP36 regulates target protein levels in T cells.

(A) Levels of mRNA and protein in Zfp36 KO and WT T cells and ZFP36 CLIP tracks measured 4 hr post-activation for targets with robust 3’UTR ZFP36 binding. RNA values are mean RPKM ± S.E.M. of 4 biological replicates. Protein values are mean fluorescence intensities (MFI) ± S.E.M. for 3–4 mice per condition. (B) Design of GFP reporters with WT 3’UTR (WT-UTR) or one lacking the ZFP36 binding site (Δ-UTR). (C) WT-UTR or Δ-UTR reporters were co-transfected into 293 cells with Zfp36 (+) or vector alone (-). 24 hr post-transfection, reporter mRNA and protein levels were measured by RT-qPCR and flow cytometry, respectively. Values are mean ± S.D. of 4 biological replicates in each condition. Data for Ifng reporters show one representative experiment of three performed. Tnf and Cd69 reporters were analyzed in one experiment. Results of two-tailed t-tests: *=p < 0.05 **=p < 0.01 ***=p < 0.001.

To directly test a role for ZFP36 in translational regulation in T cells, we next performed ribosome profiling of WT and Zfp36 KO CD4 +T cells activated for 4 hr under Th1 conditions (Figure 3A and Figure 3—figure supplement 1A [Ingolia et al., 2009]). We observed robust ribosome association of Zfp36 mRNA in WT cells that was lost completely downstream of the engineered gene disruption in Zfp36 KO cells (Figure 3—figure supplement 1B), consistent with accurate identification of translating mRNAs. Globally, there was a subtle but significant shift toward greater ribosome association in Zfp36 KO cells for mRNAs bound by ZFP36 in 3’UTR (p=1.04×10 −11 KS) or CDS (p=1.58×10 −11 ), relative to mRNAs with no ZFP36 binding (Figure 3B). These shifts mirror those for global RNA abundance, with two notable exceptions. First, mRNAs with ZFP36 binding in CDS showed a significantly larger shift in ribosome association than those with 3’UTR binding (Figure 3B). Second, in contrast to effects on RNA abundance, the top 20% most robust ZFP36 binding sites in 3’UTR showed similar effects on ribosome association to sites overall (Figure 3—figure supplement 1C). Thus, ZFP36 target mRNAs show increased ribosome association in Zfp36 KO cells.

ZFP36 regulates target ribosome association.

(A) Ribosome profiling of Zfp36 KO and WT CD4 +T cells. (B) Changes in ribosome association between Zfp36 KO and WT cells plotted as a CDF. (C) Change in translation efficiency (ΔTE) between Zfp36 KO and WT was calculated as a delta between log2(KO/WT) from ribosome profiling and RNAseq datasets. The distribution of ZFP36 targets in mRNAs ranks by ΔTE is shown (left), along with normalized enrichment scores and FDRs from GSEA (right). Intron-bound mRNAs are shown as a representative gene set that show no enhanced TE in Zfp36 KO cells. (D) Normalized coverage of ribosome profiling reads for Tnf and Ifng mRNAs in Zfp36 KO and WT cells, with p-values from binomial tests. (E) Normalized coverage of ribosome profiling reads across all mRNAs for Zfp36 KO and WT cells. Ribosome profiling data are a compilation of two experiments, with four total biological replicates per conditions.

Levels of ribosome-associated mRNA are related to total abundance. To evaluate changes in translational efficiency (ΔTE) in Zfp36 KO versus WT T cells, ribosome profiling fold-changes were normalized to those from RNAseq. We then used Gene Set Enrichment Analysis (GSEA) to examine the distribution of ZFP36 targets among all detected mRNAs ranked by ΔTE (Subramanian et al., 2005). Importantly, this analysis compares the observed ΔTE of CLIP-defined ZFP36 target mRNAs to mRNAs with no detected ZFP36 binding sites. ZFP36 3’UTR and, more significantly, CDS binding targets were strongly enriched for increased TE in Zfp36 KO cells (Figure 3C). In addition, ZFP36 targets with highly robust 3’UTR binding showed more significant effects on TE than ZFP36 3’UTR targets overall. This enrichment was not observed for mRNAs with intronic binding sites, indicating specificity for 3’UTR and CDS binding. As a striking confirmation of these results, normalized ribosome coverage on robust 3’UTR targets Tnf and Ifng was significantly higher in Zfp36 KO cells than WT (Figure 3D), despite no detectable difference in overall mRNA abundance (Figure 2A). Crucially, ribosome coverage averaged across all mRNAs was not appreciably different between KO and WT cells, indicating specific effects on ZFP36 targets (Figure 3E). Notably, the pattern of ribosome association along these and other transcripts is remarkably consistent between WT and Zfp36 KO cells, but with altered magnitude. Mechanistically, this observation indicates that ZFP36 prevents association of mRNAs with ribosomes, but does not impact elongation. These results indicate repression of mRNA target translation by ZFP36 during T- cell activation, likely at the level of initiation.

ZFP36 negatively regulates T-cell activation kinetics

ZFP36 target mRNAs pointed to multilayered control of T cell function, including its reported regulation of effector cytokines (e.g. Il2, Ifng, Tnf, Il4, Il10 Figure 4—figure supplement 1). Novel targets spanned direct components of the TCR complex (e.g. CD3d, CD3e), co-stimulatory and co-inhibitory molecules (e.g. Cd28, Icos, Ctla4), TCR-proximal signaling (Fyn, Sos1, Akt1), and transcriptional response (e.g. Fos, Nfatc1, Nfkb1). As an unbiased assessment, we examined the distribution of ZFP36 targets in high-resolution gene expression time courses of CD4 + T-cell activation (Yosef et al., 2013). ZFP36 targets were enriched for mRNAs, like its own, that were rapidly induced after T-cell activation, and targets were depleted among mRNAs with stable expression or delayed induction (Figure 4A). Gene Ontology (GO) enrichments spanned many basic metabolic and gene regulatory functions, in addition to signal transduction, cellular proliferation, and apoptosis (Figure 4B Supplementary file 2).

ZFP36 regulates T-cell activation kinetics.

(A) Gene expression patterns from a T-cell activation time course (Yosef et al., 2013) were partitioned by k-means, and enrichment of ZFP36 3’UTR and CDS targets was determined across clusters (Fisher’s Exact Test). Mean expression of genes in the three clusters most enriched (left) or depleted (right) for ZFP36 targets is plotted. (B) Enriched GO terms among ZFP36 HITS-CLIP targets (full results in Supplementary File 2). (C) Proliferation of naïve CD4 +Zfp36 KO and WT T cells in the indicated time windows after activation, measured by 3 H-thymidine incorporation (D) Fractions of apoptotic annexin-V+ and (E) proliferating Ki67 +CD4+T cells 24 hr post-activation. Mean ± S.E.M. is shown circles are individual mice (n = 3–4 per genotype). (F) Proliferation of BG2 TCR-transgenic CD4 +T cells cultured with DCs pulsed with cognate (β-gal) or irrelevant (OVA) peptide. Mean ± S.E.M. is shown (n = 5 mice per genotype). (G) Proliferation of CD4 +T cells co-cultured with syngeneic (C57BL6/J) or allogeneic (Balb-c) DCs. Mean ± S.E.M. of three replicate cultures is shown. (H) Levels of CD69 and CD25 after activation of Zfp36 KO and WT naïve CD4 +T cells. Mean ± S.E.M. is shown (n = 3–4 mice per genotype). (I) Naïve and effector subsets 40 hr post-activation in Zfp36 KO and WT CD4 +T cells. Representative plots are shown (top), along with mean ± S.E.M and circles for individual mice (n = 4 per genotype). For (C–I), results of two-tailed t-tests: *=p < 0.05 **=p < 0.01 ***=p < 0.001. Data are representative of three (H) or two (C–G, I) independent experiments.

Functional clustering of ZFP36 targets in proliferation and apoptosis prompted us to investigate potential regulation of T cell proliferation. In thymidine incorporation assays, naïve Zfp36 KO CD4 +T cells showed greater proliferation than WT from 16 to 24 hr post-activation (Figure 4C). Similar results were obtained with CD8 +T cells (Figure 4—figure supplement 2A). This increase reflected decreased apoptosis (Figure 4D) and increased numbers of proliferating cells (Figure 4E) in KO versus WT cultures. We examined whether an action on IL-2 might account for enhanced proliferation, as increased IL-2 production in Zfp36 KO T cells has been reported (Ogilvie et al., 2005), and our HITS-CLIP data confirmed direct interaction. Zfp36 KO T cells proliferated more than WT both in the presence of excess recombinant IL-2 or neutralizing anti-IL-2 antibody, as well as in different Th polarizing conditions, indicating the effect is not solely IL-2-dependent (Figure 4—figure supplement 2B–C). In summary, T cells from Zfp36 KO mice show enhanced proliferation shortly after activation under all conditions examined.

Anti-CD3 is not a physiologic stimulation, so we next examined proliferative responses to MHC-peptide-mediated TCR binding. First, we bred WT and Zfp36 KO mice with a transgenic, class-II restricted TCR specific for a β-galactosidase-derived antigen (BG2). BG2 transgenic Zfp36 KO cells showed greater proliferation than WT across a broad titration of cognate peptide, but not irrelevant peptide (Figure 4F). Second, Zfp36 KO T cells also showed greater proliferation than WT in response to allogeneic DCs (Figure 4G). Therefore, Zfp36 KO cells show an exaggerated proliferative response upon MHC-peptide stimulation over a range of signal strengths.

Analysis of canonical T-cell activation markers revealed enhanced induction of CD69 and CD25 in Zfp36 KO versus WT cells over the first 24 hr post-activation (Figure 4H). At 40 hr post-activation, a greater proportion of Zfp36 KO versus WT had transitioned from a naïve to effector surface phenotype (Figure 4I). Notably, thymidine incorporation data showed enhanced proliferation of Zfp36 KO cells early after activation, but similar rates in Zfp36 KO and WT cells after 24 hr (Figure 2C). Collectively, these results show accelerated activation kinetics in the absence of ZFP36.

ZFP36 regulation of T-cell activation is cell-intrinsic

The accelerated activation of Zfp36 KO T cells could in principle reflect the activity of other cell subsets or inflammatory signals in Zfp36 KO mice. To test for a T cell-intrinsic function, we generated mixed bone marrow (BM) chimeras, allowing isolation of WT and KO T cells that develop in the same in vivo milieu (Figure 5A). Naïve Zfp36 KO T cells sorted from chimeras showed greater proliferation than WT 24 hr post-activation, indicating cell-intrinsic effects (Figure 5B). To assess the potential impact of secreted factors, chimera-derived WT and Zfp36 KO cells were re-mixed 1:1 ex vivo. Here, differences between Zfp36 KO and WT cells were still significant, but blunted compared to separate cultures. This result indicates cross-regulatory effects between WT and KO cells through secreted or surface factors, but these do not fully account for the observed differences. Interestingly, the reduced proliferation of Zfp36 KO cells in mixed (Figure 5B, right panel) versus separate (left panel) cultures indicate that WT cells can exert a restraining effect on KO cells. Thus, accelerated activation in Zfp36 KO cells may in part reflect compromised autoregulatory and/or suppressive functions. Three days after activation, mixed cultures remained skewed in favor of Zfp36 KO cells, again confirming accelerated expansion (Figure 5C). These results show that ZFP36 regulation of T-cell activation is cell-intrinsic, and that ZFP36 normally functions to restrain T -cell activation.

ZFP36 regulation of T cell activation kinetics cell-intrinsic.

(A) Lethally irradiated mice were reconstituted with congenically marked WT and Zfp36 KO BM to generate mixed chimeras. 10–12 weeks after reconstitution, naïve CD4 +WT and Zfp36 KO T cells were sorted, then activated ex vivo separately or mixed 1:1. (B) Proliferating Ki67 +cells were measured 24 hr after activating naïve CD4 +T cells under Th0 or Th1 conditions. (C) Cultures with a 1:1 starting ratio of naïve WT and Zfp36 KO CD4 +T cells were examined 3 days post-activation. Data from one experiment of two performed are shown.

Downstream effects of ZFP36 regulation

The efficient in vitro responses of Zfp36 KO T cells suggest they are functional, but respond with altered kinetics. To examine the downstream consequences of this differential regulation, HITS-CLIP and RNAseq analyses were done in Th1-skewed CD4 +T cells 3 days after activation. ZFP36 binding site features in cells activated for 3 days were similar to ones identified at 4 hr (Figure 6—figure supplement 1A–B supplementary file 3), but results from transcriptome profiling were strikingly different at these two time points (Figure 6A). First, in contrast to subtle effects observed at the 4 hr time point, many transcripts showed highly divergent expression in Zfp36 KO versus WT T cells 3 days after activation (Figure 6A). However, these differences were not correlated to ZFP36 HITS-CLIP binding at 72 hr (Figure 6B). Thus, the absence of ZFP36 in the early phases of T -cell activation can lead to significant secondary effects downstream.

Accelerated signs of in vitro T cell exhaustion in absence of ZFP36.

(A) Log2-transformed RPKM values from Zfp36 KO versus WT CD4 +Th1 cell RNAseq 72 hr post-activation, with red indicating differential expression (FDR < 0.05). Lines mark 2-fold changes. RNAseq data represent one experiment with three biological replicates per condition. (B) Log2-transformed fold-changes (KO/WT) plotted as a CDF, for mRNAs with 3’UTR, CDS, or no significant ZFP36 HITS-CLIP. Numbers of mRNAs in each category (n) and p-values from KS tests are indicated. (C) The gene expression profile in Zfp36 KO CD4 +T cells 72 hr post-activation was compared to reported profiles of CD4 +T cell exhaustion using GSEA. Upregulated (orange) and downregulated (gray) gene sets in exhausted T cells showed strong overlap with corresponding sets from Zfp36 KO T cells (FDR < 0.001, hypergeometric test). (D) IFN-γ and TNF-α measured by ICS 3 and 5 days after activation of naïve CD4 +T cells. (E) IFN-γ and TNF-α in culture supernatants 3 and 5 days after activation of naïve CD4 +T cells. (F) PD-1, ICOS, and LAG-3 expression 5 and 13 days after activation under Th0 or Th1 conditions. (D–F) show mean ± S.E.M. circles are individual mice (n = 3–5 per genotype). (G) Measurements as in (F) for Zfp36 KO and WT CD4 +T cells derived from mixed BM chimeras. Cells were activated under Th1 conditions for 13 days, either separately or mixed 1:1. For (D–G), one representative experiment of two performed is shown. Results of two-tailed t-tests: *=p < 0.05 **=p < 0.01 ***=p < 0.001 ****=p < 0.0001.

GSEA with RNAseq data from Th1 Zfp36 KO cells 3 days post-activation showed reduced activity of transcription factors driving proliferation (e.g. Myc and E2F Figure 6—figure supplement 2A) and strong overlap with previously described signatures of T cell exhaustion (Figure 6C [Crawford et al., 2014]). Consistent with a late activated and exhausted phenotype in Zfp36 KO cells, the enhanced production of IFN-γ at early time points (Figures 2A and 3D) gave way to comparable production by day three and reduced production by day 5 (Figure 6D). Moreover, TNF-α production was not significantly different 3 or 5 days post-activation (Figure 6D), despite large differences early on (Figures 2A and 3D). In summary, relieved translational control drives elevated cytokine production in Zfp36 KO cells early post-activation, but enhanced production dissipates downstream due to more rapid expansion and exhaustion. The net result, reflecting both dysregulated cytokine production and more rapid culture expansion (Figures 4 and 5), is a higher accumulation of IFN-γ and TNF-αin Zfp36 KO culture supernatants 72 hr post-activation (Figure 6E).

We examined co-inhibitory and co-stimulatory checkpoint proteins that are linked to T cell exhaustion, and found elevated expression of PD-1 and ICOS at late time points in Zfp36 KO cells, and more rapid peaking of LAG-3 (Figure 6F). Interestingly, these effects were observed in Th1 but not Th0 conditions, suggesting a dependence on Th1 cytokines. To test this dependence, and to examine whether elevated receptor expression was T cell-intrinsic, we analyzed cells derived from mixed BM chimeras. This analysis confirmed differential, T cell-intrinsic expression of these receptors (Figure 6G). However, re-mixing WT and KO cells ex vivo neutralized these differences, indicating they are driven by secreted factors. We tested whether recombinant IFN-γ, supplemented at levels measured in KO cultures, could cause elevated receptor expression in WT Th1 cells, and found it promoted ICOS but not PD-1 upregulation (Figure 6—figure supplement 2B). These results indicate that Th1 cytokines, including but not only IFN-γ, can drive an exhaustion-like phenotype. The absence of ZFP36 promoted this phenotype in vitro, due to more rapid activation and expansion coupled with greater accumulation of Th1 cytokines.

ZFP36 regulates antiviral immunity

The accelerated response of Zfp36 KO T cells, and the potential for accelerated exhaustion, led us to examine the effects of ZFP36 regulation in vivo. We first determined that naïve Zfp36 KO mice had normal T cell levels in peripheral blood and no defects in thymocyte development (Figure 7—figure supplement 1A–B). Total splenocytes, including T cells, were slightly reduced in Zfp36 KO versus WT mice (Figure 7—figure supplement 1C), but proportions of total CD4 + and CD8+T cells were normal (Figure 7—figure supplement 1D). Proportions of naïve CD4 +T cells were also normal in KO mice, and naïve CD8 +T cells only slightly decreased. (Figure 7—figure supplement 1E). Levels of CD25-hi CD4 +cells were not significantly different in spleens of WT and KO mice, consistent with similar levels of natural Tregs (Figure 7—figure supplement 1F). FoxP3 expression was not examined directly ex vivo, but in vitro induction of Tregs from naïve cells CD4 +T cells, enumerated in FoxP3-GFP mice, was not different between WT and KO (Figure 7—figure supplement 1G [Haribhai et al., 2007]).

To examine the effector T cell populations present in spleen, splenocytes were stimulated directly ex vivo with phorbol myristate acetate (PMA) and ionomycin. More CD4 + and CD8+T cells produced IFN-γin KO versus WT splenocytes, but levels of IL-4 and IL17A production were comparable (Figure 7—figure supplement 1H). Therefore, greater numbers of Th1 cells are present at steady state in Zfp36 KO mice. To examine whether loss of ZFP36 causes an intrinsic disposition to the Th1 fate, skewing of sorted, naïve CD4 +T cells was examined. These analyses showed indistinguishable differentiation of Th1 and Th17 subsets ex vivo in WT and Zfp36 KO cells (Figure 7—figure supplement 1I). The greater accumulation of Th1 cells in vivo may therefore reflect a response to factors not replicated in vitro, or the effects of additional cell types.

The lymphocytic choriomeningitis virus (LCMV) Armstrong strain causes an acute infection leading to massive T cell expansion and viral clearance in 8–10 days (Dutko and Oldstone, 1983). Using MHC-tetramers, we observed accelerated expansion and recession of virus-specific CD4+ (Figure 7A) and CD8+ (Figure 7B) T cells in Zfp36 KO versus WT mice in peripheral blood. This result was confirmed in independent experiments focused on early time points post-infection (p.i.), where virus-specific T cells in Zfp36 KO mice showed earlier expansion and more rapid upregulation of CD69 (Figure 7C–D). Enumeration of virus-specific T cells in spleen mirrored dynamics in blood levels were greater in Zfp36 KO animals 6 days p.i., but marginally lower by day 10, consistent with more rapid expansion and resolution (Figure 7E–F). Levels of memory T cells day 40 p.i. were similar in Zfp36 KO and WT mice. In summary, antigen-specific T cell response is clinically functional but accelerated in Zfp36 KO mice during viral infection.

ZFP36 regulates anti-viral immunity.

(A) Virus-specific CD4 + or (B) CD8 +T cells were tracked in peripheral blood using MHC-tetramers after LCMV Armstrong infection (n = 8–9 mice per genotype). (C) Virus-specific CD4 +T cells and CD69 expression on CD4 +T cells in peripheral blood at early time points post-infection (p.i.) (n = 7–8 mice per genotype). (D) Virus-specific CD8 +T cells and CD69 expression on CD8 +T cells in peripheral blood at early time points p.i. (n = 7–8 mice per genotype). (E) Virus-specific CD4 +and (F) CD8 +T cells in spleen after LCMV infection (n = 5–8 mice per genotype). (G) Fraction of CD4 +T cells producing IFN-γ and TNF-α in splenic CD4 +T cells 6 days p.i., after ex vivo stimulation with GP66-77 peptide (n = 7–8 mice per genotype). (H) Levels of IFN-γ and TNF-α(gated on cytokine-producing CD4 +cells) 6 days p.i. after ex vivo stimulation with GP66-77 (n = 7–8 mice per genotype). (I) Raw percentage of bifunctional IFN-γ+TNF-α+CD4+cells in spleen 6 days p.i. after ex vivo stimulation with GP66-77 (left), or normalized to percentage of GP66-77 tetramer +cells (n = 7–8 mice per genotype). (J) Levels of LCMV genomic RNA in spleen measured by RT-qPCR (n = 9–14 per group). For (A–J), mean values ± S.E.M. are shown, with circles as individual mice. Results of two-tailed t-tests: *=p < 0.05 **=p < 0.01 ***=p < 0.001 ****=p < 0.0001. In each panel, one representative experiment of two is shown.

Stimulation with LCMV peptides ex vivo revealed higher rates of IFN-γ and TNF-αproduction in Zfp36 KO versus WT CD4+ (Figure 7G) and CD8 +T cells 6 days p.i. (Figure 7—figure supplement 2A). Numbers of cytokine-producing cells were proportional to LCMV-specific tetramer +cells (Figure 7E–F). However, levels of IFN-γand TNF-α protein were significantly greater in CD4 +Zfp36 KO cytokine-producing cells versus WT (Figure 7H), and TNF-α levels were also higher for CD8 +cells (Figure 7—figure supplement 2B). In addition, ‘bifunctional’ IFN-γ+TNF-α+T cells were more frequent in Zfp36 KO mice, even when normalized to frequencies of tetramer +cells (Figure 7I and Figure 7—figure supplement 3C).

Strikingly, LCMV genomic RNA in spleen was

10 fold lower day 6 p.i. in Zfp36 KO versus WT animals, consistent with more rapid clearance of LCMV infection (Figure 7J). Viral load correlated inversely with levels of tetramer +CD4+ and CD8+T cells in both Zfp36 KO and WT mice, consistent with the established role of T cell response in LCMV clearance (Figure 7—figure supplement 2D). To examine whether the accelerated LCMV-specific T cell response in Zfp36 KO mice may be T cell-intrinsic, infections were repeated in mixed BM chimeras. Irradiated recipient mice were re-constituted with a 1:1 mix of congenically marked WT or Zfp36 KO BM cells (Figure 7—figure supplement 3A). Ten weeks after re-constitution, pre-infection baseline measurements showed a significantly greater expansion of Zfp36 KO T cells versus WT in blood (Figure 7—figure supplement 3B). However, the kinetics of WT and KO T cell response in these animals upon LCMV infection were indistinguishable (Figure 7—figure supplement 3C–D). Therefore, in a mixed environment in vivo, Zfp36 KO and WT T cells show similar kinetics upon viral challenge. Notably, maximum T cell expansion was observed in mixed chimeras 7–8 days p.i., which was intermediate to the maxima observed in Zfp36 KO (6 days) and WT (10 days) mice.

Collectively, these data demonstrate a remarkable enhancement of anti-viral immunity in the setting of reduced ZFP36 family activity in vivo. This enhancement was marked by an accelerated T cell response and enhanced production of effector cytokines and, based on mixed chimera experiments, may involve immune cell types in addition to T cells.

Author information

Sandra Schoors and Ulrike Bruning: These authors contributed equally to this work.


Department of Oncology, Laboratory of Angiogenesis and Neurovascular link, KU Leuven, B-3000 Leuven, Belgium

Sandra Schoors, Ulrike Bruning, Rindert Missiaen, Karla C. S. Queiroz, Gitte Borgers, Annalisa Zecchin, Anna Rita Cantelmo, Jermaine Goveia, Lucica Goddé, Stefan Vinckier, Guy Eelen, Luc Schoonjans, Mieke Dewerchin, Katrien De Bock, Bart Ghesquière & Peter Carmeliet

Laboratory of Angiogenesis and Neurovascular Link, Vesalius Research Center, VIB, B-3000 Leuven, Belgium

Sandra Schoors, Ulrike Bruning, Rindert Missiaen, Karla C. S. Queiroz, Gitte Borgers, Annalisa Zecchin, Anna Rita Cantelmo, Jermaine Goveia, Lucica Goddé, Stefan Vinckier, Guy Eelen, Luc Schoonjans, Mieke Dewerchin, Katrien De Bock, Bart Ghesquière & Peter Carmeliet

Department of Oncology, Laboratory of Cellular Metabolism and Metabolic Regulation, KU Leuven, B-3000 Leuven, Belgium

Ilaria Elia, Stefan Christen & Sarah-Maria Fendt

Laboratory of Cellular Metabolism and Metabolic Regulation, Vesalius Research Center, VIB, B-3000 Leuven, Belgium

Ilaria Elia, Stefan Christen & Sarah-Maria Fendt

Department of Cardiovascular Research, Center for Molecular & Vascular Biology, KU Leuven

Division of Clinical Cardiology, UZ Leuven, B-3000 Leuven, Belgium

Laboratory of Lipid Biochemistry and Protein Interactions, KU Leuven, B-3000 Leuven, Belgium

Department of Oncology, Vascular Patterning Laboratory, KU Leuven, B-3000 Leuven, Belgium

Vascular Patterning Laboratory, Vesalius Research Center, VIB, B-3000 Leuven, Belgium

Integrative Vascular Biology Laboratory, Max Delbrück Center for Molecular Medicine, Berlin, 13125, Germany

Department of Pharmaceutical and Pharmacological Sciences, Laboratory of Cell Metabolism, KU Leuven, B-3000 Leuven, Belgium

Department of Kinesiology, Exercise Physiology Research Group, KU Leuven, B-3001 Leuven, Belgium

Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, 48824, Michigan, USA

Watch the video: cell proliferation assay services (February 2023).