|Home | About | Journals | Submit | Contact Us | Français|
Distinct CD4+ T cell subsets are critical for host defense and immunoregulation. While these subsets can behave as terminally differentiated lineages, elements of plasticity are increasingly recognized. MicroRNAs are one factor that controls stability and plasticity. Herein, we report that miR-10a was highly expressed in naturally-occurring regulatory T (Treg) cells and induced by retinoic acid and TGF-β in inducible Treg cells. By simultaneously targeting Bcl-6 and a co-repressor, Ncor2, miR-10a attenuated phenotypic conversion of inducible Treg cells to follicular helper T cells. miR-10a also limited TH17 differentiation and therefore represents a new factor that can fine tune plasticity and fate decision of helper T cells.
Appropriate differentiation of CD4+ T cell subsets is important for proper adaptive immune responses and is achieved by alteration of genetic programs controlled by transcription factors and/or repressors induced by extracellular cues1. Cytokines in the milieu of differentiating T cells are the key exogenous factors in driving specification, such that multiple, distinct fates are possible for naïve T cells including T helper (TH)1, TH2 and TH17 cells. Regulatory T (Treg) cells are yet another subset of CD4+ T cells that serve to suppress activity of effector T cells, and thus maintain immune homeostasis. Treg cells can be further divided into two populations, naturally-occurring Treg (nTreg) that arise in the thymus and inducible Treg (iTreg), which can arise in the periphery from naïve CD4+ precursors. Both of these subsets express Foxp3, a master transcription factor that defines this lineage and factors that promote expression of Foxp3 include: TGF-β, IL-2 and retinoic acid (RA)2–4. T follicular helper (TFH) cells are another functional subset, which is critical for providing help to B cells in germinal centers5. TFH can arise from naïve CD4+ T cells; however, the relationship between TFH cells and other subsets is the subject of considerable debate and ongoing investigation5.
There is evidence that these subsets behave as lineages with respect to expression of specific cytokines and lineage-defining master regulator transcription factors. However, there is also much emerging information on functional plasticity of helper T cells6. One example of plasticity is the generation of TH1 cells from TH17 cells7. Another particularly dramatic example is the conversion of polarized TH2 cells to IFN-γ producers in the setting of viral infection8. Unlike nTreg cells, iTreg cells exhibit incomplete demethylation of the Foxp3 locus9. While the latter may be intrinsically less stable, nTreg cells are generally viewed as a more stable subset10. However, even for nTreg cells, there is evidence that they can alter their phenotype. For instance, nTreg cells can convert to TFH cells in the environment in Peyer's patches11 or become TH17 cells upon stimulation with IL-6 in vitro12. There are still other examples of conversion from other subsets such as TH2 to TFH cells13 and conversely conversion of TFH to other effector subsets including TH1, TH2 and TH1714. However, the molecular mechanisms underlying phenotype switch involving Treg and other cells are incompletely understood.
MicroRNAs (miRNAs) are small non-coding RNAs, which regulate gene expression at post-transcriptional level by directly binding to mRNA of target genes. miRNAs are critically involved in a wide variety of biological processes of the cells from embryonic stem cell pluripotency to cancer tumorigenicity15. Evidence showing critical roles of miRNAs in immune cells continue to mount16–24. Drosha and Dicer are two key components of the machinery responsible for miRNA generation and loss of these factors is associated with defects in lymphocyte differentiation and autoimmunity25–27. With respect to the issue of CD4+ T cell plasticity, deficiency of Dicer resulted in helper T cell instability, unstable expression of Foxp3 in Treg cells and skewing to a TH1 phenotype28. Similarly, CD4+ T cells deficient in Drosha showed accelerated TH25 differentiation and reduced iTreg differentiation25. Collectively, these data argue that miRNAs are important factors in preserving phenotypic stability of helper T cells. Furthermore, the data raise the question as to which specific miRNA(s) are responsible for maintaining lineage phenotype.
In this study, we investigated the function of miR-10a, which is highly expressed in nTreg cells and induced by TGF-β and RA. We found that miR-10a controls the levels of expression of two key repressors, Bcl-6 and Ncor2, limits the conversion of iTreg cells to TFH cells. Under conditions in which naïve CD4+ T cells are exposed to RA, miR-10a inhibits TH17 differentiation. This effect is dependent upon the induction of T-bet by RA. Our data demonstrate that miR-10a is one factor that preserves Treg phenotype by targeting and constraining transcription factor pathways that promote alternate fates.
To identify miRNAs with preferential expression in different T cell subsets, we obtained genome-wide miRNA expression profiles via massive parallel sequencing (miRNA-seq)29. Among miRNAs that were differentially expressed and abundant in T cells, miR-10a-5p, referred to as miR-10a hereafter, was identified as the most selective marker of nTreg cells versus other T cell subsets (Supplementary Fig. 1a). By contrast, another paralogous sequence, miR-10b-5p, was not detected in any of TH subsets (data not shown). We next confirmed the selective expression of miR-10a by quantitative RT-PCR and found that TH1, TH2, TH17, and naïve CD4+ T cells showed very low miR-10a expression, consistent with sequencing data (Fig. 1a and Supplementary Fig. 1b). By contrast, nTreg cells expressed high levels of miR-10a. Because in vitro generated iTreg cells share many of the functional properties of nTreg cells, we next compared the expression of miR-10a in these two types of cells. Interestingly, we found that iTreg cells expressed modest levels of miR-10a compared to nTreg cells; nonetheless, the levels were significantly higher in iTreg cells compared to other TH subsets.
Since the generation of iTreg in vitro requires TGF-β, we next asked if this was a relevant factor for miR-10a induction. As shown in Fig. 1b and c, TGF-β induced miR-10a in a dose-dependent manner whereas IL-2 had no effect. Because the induction of miR-10a by TGF-β was modest, we considered other factors that might regulate its expression. Another means of inducing iTreg cells is through exposure of T cells to RA3, 4. This was notable as miR-10a resides within Hox gene cluster on murine chromosome 11, a genomic segment that is highly evolutionarily conserved among mammals. Hox genes are highly regulated by RA and in non-T cells miR-10a has been reported to be induced by RA30. Consistent with these previous results, we noted that all-trans RA (ATRA) dramatically induced miR-10a expression in a dose-dependent manner. Importantly though, we also noted that the induction of miR-10a by ATRA was totally abolished by blocking TGF-β (Fig. 1d). This has not been appreciated previously and thus in contrast to previous findings, our data support the idea that RA and TGF-β are both required for induction of miR-10a in T cells. Along this line, we also found that the induction of miR-10a expression was blocked in the presence of the pan-RA receptor (RAR) inhibitor, LE540 (Fig. 1e).
To help explain the interaction of RA and TGF-β, we considered the possibility that these factors might be impinging upon one another's signaling pathways. Indeed, we found that TGF-β induced RARα expression in a dose-dependent manner (Fig. 1f). Accordingly, in conjunction with TGF-β, we found that a RARα agonist but not a RARγ agonist induced miR-10a expression to levels equivalent to those induced by ATRA (Fig. 1g). Collectively these results demonstrated that RA-induced miR-10a expression is mediated through RARα in a TGF-β-dependent manner, with the latter inducing expression of the former.
Since identification of target genes by miR-10a is indispensable to understanding how the miRNA functions in helper T cells, we next performed in silico analysis to predict target mRNAs based on nucleotide sequences using PicTar (http://pictar.mdc-berlin.de/). Among the predicted targets, mRNAs for Bcl-6 and Ncor2 were identified as carrying potential miR-10a target sequences in their 3' untranslated regions (3'UTR) (Supplementary Fig. 2a). The target genes were also predicted by another algorithm, TargetScan (http://www.targetscan.org), and the target sequences in 3'UTR of Bcl-6 and Ncor2 are highly conserved among various species (Supplementary Fig. 2b). We reasoned that if miR-10a were involved in fate decision of helper T cell phenotype, these targets would provide good potential mechanisms for transmitting its effects. That is, Bcl-6 is as a key transcriptional repressor thought to be a master regulator for TFH cells and Ncor2 is a co-repressor that complexes with Bcl-6 and RAR to suppress transcription of genes regulated by these factors31–35.
To investigate whether miR-10a regulates Bcl-6 and Ncor2 expression, we first generated luciferase reporter constructs carrying 3'UTR of Bcl-6 or Ncor2 that comprised potential miR-10a target sequences linked to a luciferase gene (Fig. 2a). We found that over-expression of miR-10a significantly reduced activity of constructs expressing the Bcl-6 and Ncor2 target sequences (Fig. 2a: middle bars). However, miR-10a had no effect on the expression of constructs lacking these target sequences (Fig. 2a: right bars). These results demonstrate that, as predicted, miR-10a can directly target sequences in the 3'UTR of Bcl-6 and Ncor2.
We next sought to evaluate whether miR-10a influences endogenous Bcl-6 and Ncor2 protein levels. To this end, we first used CH12, a murine B cell lymphoma cell line that constitutively expresses Bcl-6 and found that miR-10a over-expression decreased Bcl-6 protein levels as measured by flow cytometry and immunoblotting (Fig. 2b and c). To assess whether miR-10a regulates Bcl-6 in primary helper T cells, we over-expressed miR-10a in helper T cells using retroviral transduction and measured Bcl-6 protein levels. We found that miR-10a over-expression significantly decreased Bcl-6 expression as determined by flow cytometry and immunoblotting (Fig. 2d and e). Reduction in Bcl-6 protein levels was associated with compensatory increases in Bcl-6 mRNA levels (Fig. 2f). Conversely, over-expression of Bcl-6 protein using retroviral transduction suppressed endogenous Bcl-6 mRNA levels (Supplementary Fig. 3), consistent with previous data36.
To simplify the complex regulation of Bcl-6 expression, we expressed a tagged Bcl-6 construct under a heterologous promoter and manipulated miR-10a levels by over-expressing or by sequestering miR-10a with “sponge target sequence” (Supplementary Fig. 4)37. As shown in Fig. 2g, over-expression of miR-10 down-regulated Bcl-6 expression, whereas expression of sponge target sequence restored expression (Fig. 2g, red and green column). However, deletion of miR-10a target sequence from 3'UTR eliminated this regulation (Fig. 2g, blue column). Over-expression of miR-10a in helper T cells also reduced Ncor2 protein levels, but had little effect on mRNA levels (Fig. 2h and i).
Finally, to verify the effect of endogenous miR-10a in regulating its target proteins, we expressed the miR-10a sponge sequence in iTreg cells. We found that expression of the miR-10a sponge sequence significantly up-regulated expression of Bcl-6 and Ncor2 compared to expression of a control vector (Fig. 2j and k).
Considering that Ncor2 is a co-repressor of RARα, it seemed plausible to expect that reduction in Ncor2 levels would augment the ability of RA to induce miR-10a expression by relieving the inhibitory effects of Ncor2. To examine this possibility, we utilized retroviral expression of shRNA to knockdown Ncor2 expression (Supplementary Figure 5a and b) and cultured the cells with RA. Consistent with our previous results, RA induced expression of miR-10a in a dose-dependent manner; however, inhibiting Ncor2 shifted the dose-response curve to the left, making the cells more sensitive to RA and more effectively inducing miR-10a (Supplementary Figure 5c). Manipulating the levels of another miR-10a target, either by knocking down and over-expressing Bcl-6 had no effect on RA-induced miR-10a expression (Supplementary Figure 5c and Supplementary Figure 6). These results suggest that miR-10a amplifies its own expression during ATRA treatment by forming a positive-feedback loop through targeting Ncor2. A recent report that showed increase in miR-10a level by inhibition of Ncor2 in neuroblastoma cell line is in line with our results38.
Since Bcl-6 is pivotal for TFH differentiation31–33, we hypothesized that miR-10a might function to limit TFH generation by suppressing Bcl-6. Since Treg cells can convert to TFH cells in Peyer's patches (PP), this is a circumstance in which miR-10a might be functionally relevant11. In order to evaluate a potential effect of miR-10a on Treg-TFH conversion, we adapted this previously reported model, using iTreg cells rather than nTreg cells, since the former are more suited for effective retroviral transduction. To this end, naïve CD4+ and in vitro differentiated iTreg cells derived from Foxp3GFP mice were transferred into TCRa−/− mice and the appearance of TFH in PP was assessed 6 weeks after adoptive transfer (Supplementary Fig. 7a and b). Immunofluorescence staining showed that transferred naïve CD4+ T cells resided predominantly in interfollicular regions, although a proportion of the cells localized to germinal centers (GC) of PP suggesting differentiation of the naive T cells to TFH (Supplementary Fig. 7c left panel). To further confirm TFH differentiation of the naïve T cells, we used flow cytometry to assess expression of the characteristic surface markers PD-1 and CXCR5 (Supplementary Fig. 7d left panel). Consistent with the immunofluorescence staining, it was clear that a portion of the transferred naïve CD4+ T cells acquired characteristics of TFH cells. Similarly, transferred in vitro-generated iTreg cells localized to GC of PP (Supplementary Fig. 7c middle panel) and they also acquired expression of PD-1 and CXCR5; in fact, iTreg cells were more efficient in their ability to convert to TFH cells than were naive CD4+ T cells (Supplementary Fig. 7d). As a control, we also assessed GFP+ (Foxp3+) expression in transferred naïve and iTreg cells. GFP+ (Foxp3+) cells were barely detectable in PP and spleens when naïve CD4+ T cells were transferred (Supplementary Fig. 7e). On the other hand, in mice that received GFP+ iTreg cells, cells more easily lost GFP expression in PP compared to spleen, suggesting the PP environment promoted TFH conversion (Supplementary Fig. 7e).
We next evaluated the effect of inhibiting miR-10a function on the conversion of iTreg to TFH cells. To this end, we generated iTreg cells (97.7 ± 0.4 % Foxp3+, data not shown), which expressed the miR-10a sponge sequence or a control sequence. We then adoptively transferred the transduced iTreg cells into TCRa−/− mice and assessed the acquisition of TFH markers on this cell population. Compared to cells transduced with the control vector, cells transduced with the miR-10a sponge vector more readily acquired the TFH markers PD-1 and CXCR5 in PP (Fig. 3a). This was associated with significantly more germinal center B cells in this tissue (Fig. 3b).
To verify this finding, we next determined the effect of over-expressing miR-10a in iTreg cells, reasoning that enforced expression of miR-10a should limit TFH conversion by attenuating levels of Bcl-6. As shown in Fig. 3c, we observed that over-expression of miR-10a significantly reduced the efficiency of transferred iTreg cells to convert into TFH cells in PP, underscoring the crucial roles of miR-10a in regulating iTreg to TFH conversion.
We next measured the levels of miR-10a and Bcl-6 in TFH (CD4+Vβ+PD-1hiCXCR5hi) cells and non-TFH (CD4+Vβ+PD-1−CXCR5−) cells from PP (Supplementary Fig. 8). As expected, Bcl-6 was preferentially expressed in TFH cells compared to non-TFH cells (Fig. 3d), whereas the opposite was the case for miR-10a (Fig. 3e). Recently a new population of helper T cells designated follicular regulatory T cells (TFR) has been identified. These cells express both Foxp3 and Bcl-6, and exert a regulatory function in germinal center39–41. To assess miR-10a expression levels in this interesting subset, we isolated three populations: TFR (Foxp3+CD4+Vβ+PD-1hiCXCR5hi), Treg (Foxp3+CD4+Vβ+PD-1−CXCR5−) and Foxp3−TFH (Foxp3−CD4+Vβ+PD-1hiCXCR5hi) (Supplementary Fig. 9a). As shown in Supplementary Fig. 9b, TFR cells expressed higher levels of miR-10a than Treg cells. Taken together, these results argue that one function of miR-10a is to constrain iTreg to TFH conversion.
Deficiency of Bcl-6 has also been reported to influence IL-17A production32. Consequently, miR-10a might also regulate IL-17A production; however, effects of miR-10a would only come in to play in the presence of RA and in previous work the effect of RA was not considered. Additionally though, alterations in Ncor2 levels could affect sensitivity to RA through effects on RAR34. Thus, miR-10a might alter TH17 differentiation in the context of exposure to RA, but it was not obvious as to what that effect might be. Moreover, whereas RA was originally reported to induce iTreg and suppress TH17 differentiation3, more recent work suggests that RA is necessary for normal immune responses and can enhance inflammation42, 43.
These findings led us to re-examine the effect of RA on TH17 differentiation to elucidate potential effects of miR-10a. Consistent with previous reports, we found that 10 nM ATRA inhibited IL-17A production and enhanced Foxp3 expression (Fig. 4a and b). However, physiologic levels of RA (1 nM)44 enhanced the proportion of IL-17A-producing T cells (Fig. 4b; blue triangles). Previous work has demonstrated that IL-2 inhibits TH17 differentiation45 and the positive effect of RA on IL-17A production was even more evident in situations in which IL-2 was neutralized (Fig. 4b; green circles). As a control, cells cultured under iTreg condition without IL-6 did not produce IL-17A even with low dose of RA (Fig. 4b; black diamonds). Interestingly, the magnitude of IL-17A production, assessed by MFI, increased at all doses of ATRA (Supplementary Fig. 10), despite the reduction in the proportions of IL-17A-producing cells. The discrepancy between MFI and percentage of IL-17A+ cells on ATRA response underscores the complex biological effects of ATRA on TH17. Under these culture conditions, ATRA significantly up-regulated miR-10a expression in a dose-dependent manner (Supplementary Fig. 11).
TGF-β is another factor that regulates IL-17A production, with low doses of TGF-β promoting IL-17A production and high doses inhibiting46. However, as indicated above, TGF-β also enhances T cell responsiveness to ATRA by up-regulating RARα expression (Fig. 1f). We therefore assessed the combined effects of RA and TGF-β on in vitro TH17 differentiation (Fig. 4c and d). Consistent with previous reports, TH17 differentiation was more efficient at lower concentrations of TGF-β and inhibited by higher concentrations. The ability of RA to enhance IL-17A production was more dramatic at high concentrations of TGF-β, but regardless of the concentration of TGF-β, the biphasic effects of RA were evident (Fig. 4c and d). These results indicate that TH17 differentiation can be fine-tuned by ATRA and TGF-β; inefficient induction of TH17 response that might occur with higher levels of TGF-β can be compensated by the presence of RA.
Given that the ability of RA to influence TH17 differentiation and induce miR-10a, we next sought to analyze the impact of miR-10a expression on TH17 differentiation in vivo using a disease model dependent upon these cells. To this end, we transduced naïve CD4+ T cells from 2D2 transgenic mice and assessed whether miR-10 over-expression influenced the severity of experimental autoimmune encephalomyelitis. Following immunization with MOG peptide, mice that received 2D2 CD4+ T cells that over-expressed miR-10a had significantly delayed onset of neurological disease (Fig. 5a). Consistent with this finding IL-17A production in LN, spleen and CNS was significantly reduced in miR-10a-expressing 2D2 cells compared to cells transduced with the control vector, whereas there was no significant effect of miR-10a on IFN-γ production (Fig. 5b). These results argue that miR-10a can function as a factor that limits TH17 responses in vivo. It is notable however, that although the onset of disease was delayed, disease was not abrogated. On the contrary, disease eventually occurred at the same severity as controls, suggesting that miR-10a fine tunes IL-17.
Having found that miR-10a over-expression can limit TH17 differentiation in vivo, we set out to better understand the circumstances in which miR-10a can influence TH17 differentiation. As shown in Figure 6a, when RA was used to generate TH17 cells that expressed miR-10a, (data not shown), sequestering miR-10a with the sponge vector significantly increased the proportion of IL-17A producing cells compared to a control vector. Conversely, when miR-10a was over-expressed under the same conditions, TH17 differentiation was inhibited (Fig. 6b). The alterations in IL-17A production were not due to changes in the levels of Foxp3 (Supplementary Figure 12). Interestingly though, when miR-10a was over-expressed in TH17 cells cultured without RA, there was no significant effect (Fig. 6c). Thus, miR-10a appears to be a regulator of TH17 differentiation, but only in circumstances in which RA is present.
The finding that miR-10a inhibited TH17 differentiation was surprising to us. We had anticipated that since miR-10a inhibits Bcl-6 expression and Bcl-6 has been reported to inhibit IL-17A production32, miR-10a expression would promote TH17 differentiation. We therefore next sought to clarify potential mechanisms by mimicking its effects and directly down-regulating expression of Bcl-6 and Ncor2. In contrast to previous reports, knocking down Bcl-6 resulted in reduced TH17 differentiation (Fig. 7a). The same result was observed with knocking down Ncor2 levels (Fig. 7a). Thus, inhibiting Bcl-6 and Ncor2 expression recapitulated the effect of over-expressing miR-10a. While this result may seem counter-intuitive, it needs to be borne in mind that TH17 differentiation occurred in the presence of RA in these experiments.
While modulating sensitivity of cells to RA by changing Ncor2 levels can alter TH17 differentiation, it was less obvious why down-regulation of Bcl-6 might inhibit TH17 differentiation. In this regard, a relevant Bcl-6 target is T-bet, an important negative regulator of TH17 differentiation32, 33, 47. Therefore we speculated that deletion of T-bet might reveal the positive action of Bcl-6 with regard to IL-17A production. In other words, if Bcl-6 antagonizes the action of T-bet, reducing the expression of the former would be expected to enhance the latter's function and thereby suppress TH17 differentiation. To test the possible involvement of T-bet in Bcl-6-dependent regulation of TH17, we evaluated the effect of Bcl6-KD in T-bet−/− CD4+ T cells on TH17 differentiation and found that decreasing Bcl-6 levels had no effect on TH17 differentiation in the absence of T-bet (Fig. 7b). Thus, the reduction in IL-17A production associated with reduced expression of Bcl-6 is entirely regulated upon T-bet.
In addition to forming a complex with RARα, Ncor2 forms a repressor complex with Bcl-635. We therefore wondered whether the effects of inhibiting Ncor2 might also be dependent upon T-bet. Of note, we found that the effect of knocking down Ncor2 on IL-17A production was also T-bet dependent (Fig. 7a and b; middle panels).
The unexpected involvement of T-bet in mediating regulation of TH17 differentiation by miR-10a targets, Bcl-6 and Ncor2, prompted us next to evaluate the effect of RA on T-bet expression. As shown in Fig. 7c and d, RA was a very effective inducer of T-bet under TH17 conditions. This result then helps explain the negative effect of miR-10a on TH17 differentiation in the presence of RA and the lack of effect in the absence of RA (Fig. 6c).
In the present study, we investigated role of a microRNA that was found to be preferentially expressed in Treg cells. We found that miR-10a was induced by RA and TGF-β and targets Bcl-6 and Ncor2. Functionally, we found that miR-10a limited the ability of iTreg cells to convert to TFH cells in PP, which was confirmed by both loss-of-function and gain-of-function experiments. We also found that miR-10a limited TH17 differentiation in vitro and in vivo and that the effect of miR-10a could be phenocopied by reducing expression of its targets Bcl-6 and Ncor2. Strikingly though, the negative effects of Bcl-6 and Ncor2 knock-down on TH17 differentiation were dependent upon T-bet, a transcription factor that we also found to be induced by RA. Thus, miR-10a appears to be a factor that limits conversion to TFH and TH17 cells in the setting of RA and TGF-β by regulating levels of Bcl-6 and Ncor2.
The importance of miRNA in T cell biology is rapidly becoming apparent. It is clear that mouse and human TH subsets express distinct miRNAs16, 29. Although there is clear evidence that miRNAs in general are critically involved in T cell stability25–27, the precise roles of individual miRNAs have been less clear. However, this gap in knowledge is rapidly being filled and several studies reported the roles of individual miRNAs in helper T cells. Such reports include miR-125b in maintenance of the naïve helper T cells, miR-182 in clonal expansion16, 19, miR-326 in TH17 differentiation, miR-29 in TH differentiation17, 18, 21, miR-146a in Treg20, miR-155 in TH2 differentiation and Treg development22–24. It appears that miR-10a can be added to this list as a factor that influences TFH/Treg plasticity.
miR-10a is highly expressed in nTreg but does not seem to directly regulate Foxp3 or other factors involved in Treg homeostasis. Rather, it appears that by targeting Bcl-6, a relevant action of miR-10a with respect to T cell biology is to limit the ability of Treg cells to acquire features of TFH cells. That is, its function is to reduce the plasticity of Treg cells rather than directly preserving Foxp3 expression or promoting suppressive activity of this subset. Of note in this regard is that iTreg cells generated with RA were more likely to retain Foxp3 expression and were more resistant to conversion to other lineages compared to iTreg cells generated by other methods4. The differential expression of miR-10a among nTreg and RA-treated iTreg cells helps explain the stability of these subsets and is consistent with the functional data presented in the present study.
Very recently, it has been shown that a subset of GC T cells express Foxp3 and Bcl-639–41. Functionally, it appears that these cells exhibit suppressive function. Our data show that miR-10a is highly expressed in TFR cells compared to Treg cells (Supplementary Figure 9b). Consistent with our view that miR-10a can limit Treg/TFH conversion, it is possible that miR-10a might have particularly important functions in this subset.
In our hands, miR-10a did not reduce Bcl-6 mRNA levels. While a previous report demonstrated that many miRNAs function by destabilizing their target mRNAs48, it is also apparent that it is not always the case. However, the regulation of Bcl-6 expression is quite complex and subject to autoregulation36. Given this complexity, it is difficult to say with certainty that miR-10a does or does not affect Bcl-6 mRNA. Nevertheless, the preponderance of data clearly indicate that miR-10a can influence Bcl-6 protein levels. There is less information on Ncor2 mRNA regulation, but in any case it does appear that miR-10a has immunologically relevant functions that can be linked to altering the levels of Ncor2 and Bcl-6.
An unexpected finding in the present study was that miR-10a also controls IL-17A production and that this may also be related to its ability to modulate expression of Bcl-6. We did not anticipate this result as Bcl-6 has been previously reported to itself negatively regulate IL-17A32. What was not taken into account in the previous in vitro experiments were the complex actions of RA as it pertains to TH17 differentiation.
RA is a ubiquitous metabolite of vitamin A with broad actions on cellular differentiation. Vitamin A and RA are abundantly present in the liver and the gut and it is very clear that RA plays major roles in local homeostasis of mucosal immunity. It is important for trafficking and gut tropism of conventional T cells and Treg cells4, 49. RA can induce expression of Foxp3 and inhibit IL-17A production3. Herein, we confirmed that high doses of RA inhibit IL-17A production and found accordingly, that over-expression of miR-10a, which is induced by RA, likewise inhibits IL-17A. It is of interest in this regard to note that RA appears to directly regulate miR-10a. Recent work shows that RA receptor directly binds the miR-10a locus in ES cells50.
Recent work, however, has demonstrated that even though RA can promote Foxp3 expression, it is also necessary for normal gut immune responses. That is, vitamin A-deficient mice exhibited poor TH17 responses during T. gondii infection42. Our results showing the biphasic effects of RA on TH17 differentiation are consistent with these findings. At present, we do not have a precise explanation for how RA enhances IL-17 production. The induction of miR-10a also does not provide a ready explanation for the positive effects of RA, as over-expressing it inhibits IL-17A production. As Ncor2 is a regulator of RA signaling though, it is conceivable that fine-tuning the levels of Ncor2 could have divergent effects on IL-17A production since RA's effects are clearly not simple.
RA has also been found to enhance TH1 responses42, 43. This was attributed to enhanced IL-12 production from dendritic cells; however, our results indicate that under certain circumstances RA can also induce expression of T-bet in helper T cells. This is also a mechanism by which RA can inhibit IL-17A production, as T-bet inhibits RORγt function47.
What was initially very perplexing to us was how miR-10a's ability to down-regulate Bcl-6 would result in decreased IL-17 production. This seemed at odds with the reported ability of Bcl-6 to inhibit IL-17A32 but in the context of RA's ability to induce T-bet, a potent inhibitor of IL-17A, the result now makes more sense. While in some circumstances reducing the level of Bcl-6 might amplify IL-17A production, in circumstances in which T-bet is present, reduction in Bcl-6 levels has the opposite effect. That is, inhibition of Bcl-6 expression results in an opposed action of T-bet to more efficiently inhibit IL-17A production. By modulating Bcl-6 levels, under the conditions employed in the present study, this is how we believe miR-10a inhibits IL-17A production. Although the effect of miR-10a on in vitro TH17 differentiation was only evident in the presence of RA, we were able to discern an effect in vivo when miR-10a was over-expressed in adoptively transferred 2D2 T cells. However, our data also show that miR-10a modulated but did not abrogate disease. Thus, like other miRs, miR-10a is best thought of as “tuning” responses. The data also make clear that the effects of RA on TH17 differentiation are complex and its effects are influenced by another ubiquitous regulator of T cell differentiation, namely TGF-β. Thus, the action of RA is by no means simple, as it can affect the levels of T-bet, Foxp3 and Bcl-6, depending upon the context.
In this study, we sought to dissect the function of miR-10a, a miRNA that is highly expressed in nTreg cells and inducible by RA and TGF-β. By targeting Bcl-6, the miRNA appears to have functionally attenuated TFH and TH17 cell differentiation. As such, it is one factor that appears to contribute to maintenance of cell specific phenotype in Treg cells by targeting factors that could lead to conversion to other fates (Fig. 8). It is quite conceivable that the regulatory role of miR-10a would be most effective under the circumstances where cells are poised for phenotypic transition, the gut being one place where this occurs. Thus, regulation of miR-10a levels would appear to be a reasonable mechanism for fine-tuning factors that influence fate decision versus plasticity in TH subsets.
Foxp3-IRES-GFP knock-in mice on C57BL/6 background, referred to as Foxp3GFP mice, were described previously1. B6.129S2-Tcratm1Mom/J mice (TCRα−/− mouse), B6.129S6-Tbx21tm1Glm/J mice (Tbx21−/− mice) and C57BL/6-Tg (Tcra2D2, Tcrb2D2)1Kuch/J mice (2D2 mice)2 were purchased from Jackson Laboratory. C57BL/6J mice, B6.CD45.1 mice and B6.129S6-Rag2tm1Fwa N12 mice (Rag2−/− mice) were purchased from Taconic. All animals were handled and housed in facilities in accordance with National Institutes of Health Animal Care and Use Committee guidelines.
CD44-FITC (IM7), B220-PE (RA3-6B2), CD25-PE (PC61), CD45.1-PE (A20), IL-17A-PE (TC11-18H10), CD62L-APC (MEL-14), TCRβ-APC (H57-597), B220-Alexa488, GL-7-Alexa647, hNGFR-biotin (C40-1457), and streptavidin-PE/Cy7 were purchased from BD Bioscience. hNGFR-FITC (ME20.4), hNGFR-PE, hNGFR-APC, CD4-APC/Cy7 (RM4-5), TCRβ-Pacific blue (H57-597) were purchased from Biolegend. Foxp3-FITC or APC (FJK-16s), AID (mAID-2), T-bet-eFlouor660 (4B10) and Bcl6-PE (GI191E) were purchased from eBioscience. Rabbit anti-Bcl-6 antibody (#4242) and anti-Flag-Alexa647 antibody (#3916S) were purchased from Cell Signaling Technology. Rabbit anti-Ncor2 antibody (06-891) and mouse anti-actin antibody (C4) were purchased from Millipore. Anti-rat IgG-Alexa647, anti-rabbit IgG-Alexa488 and streptavidin-Alexa594 were purchased from Invitrogen. Anti-rabbit IgG-IRDye800 and anti-mouse IgG-IRDye700 were purchased from Rockland.
Peripheral T cells were obtained from spleen and LNs of 8- to 12-week-old mice. Naïve CD4+ T cells (CD4+CD25−CD62LhiCD44lo) were isolated by FACSAria II (BD) or MoFlo (Beckman Coulter) after enrichment of CD4+ T cells by using AutoMACS with mouse CD4+ T Cell Isolation Kit (Miltenyi Biotec). For some experiments, CD4+CD25−CD11b−CD11c−CD62LhiCD44lo population was isolated as naïve CD4+ T cells as mentioned. To isolate naïve CD4+ T cells and nTreg cells from Foxp3EGFP mouse, CD4+GFP−CD62LhiCD44lo cells and CD4+GFP+ cells were isolated, respectively. To isolate iTreg cells, GFP+ cells were differentiated from naïve T cells of Foxp3EGFP mouse and isolated. To analyze lymphocytes from Peyer's patches (PP), PP were enucleated from the small bowels and subjected to 15-minute incubation at 37 °C in HBSS containing 10% FBS, 5mM EDTA, 15mM HEPES, and 1mM DTT. Supernatant containing cell debris was removed at every vigorous vortex and wash with HBSS containing 5% FBS, 25mM HEPES until supernatant becomes clear. Then Peyer's patch was mechanically smashed to make single cell suspension. For some experiments, TFH cells (CD4+Vβ+PD-1hiCXCR5hi) and non-TFH cells (CD4+Vβ+PD-1−CXCR5−) were isolated from PP by FACS. For some experiments, TFR cells (CD4+Vβ+Foxp3+PD-1hiCXCR5hi), Treg cells (CD4+Vβ+Foxp3+PD-1−CXCR5−), TFH cells (CD4+Vβ+Foxp3−PD-1hiCXCR5hi) were isolated by FACS from spleen of Foxp3GFP mice which were immunized with SRBC as previously described3.
All culture for T cells were performed in RPMI 1640 (Invitrogen) supplemented with 10% fetal calf serum, 2 mM glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin, 10 mM HEPES, 1 mM sodium pyruvate and non-essential amino acid, and 2 βM β-mercaptoethanol. Cells were activated with plate-bound anti-CD3 (10 μg/ml; 145-2C11) and CD28 (10 μg/ml; 37.51, both from eBioscience) Ab. The following culture conditions were used unless mentioned elsewhere. TH17 condition contained 2 ng/ml TGF-β (R&D systems), 10 ng/ml IL-6 (R&D), 10 μg/ml anti-IFN-γ (BD) and 10 μg/ml anti-IL-2 (BD). For Figure 7 a and b, anti-IL-4 antibody was additionally used. iTreg condition contained 5 ng/ml TGF-β, 50 U/ml hIL-2, and 10 μg/ml anti-IFN-γ. ATRA, A7980, LE540 and AM580 were used as described previously4. NIH3T3 cells were maintained in DMEM (Invitrogen) supplemented with 10% calf serum, 2 mM glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin, 1 mM sodium pyruvate and non-essential amino acid. For HEK293T cells, DMEM supplemented with 10% FBS, 2 mM glutamine, 100 IU/mL penicillin, 0.1 mg/mL streptomycin, 1 mM sodium pyruvate and non-essential amino acid was used.
Naïve CD4+Vβ11+CD25−CD62Lhi CD44lo cells isolated from 2D2 mice were stimulated with plate-coated anti-CD3 and CD28 antibodies for 16 hours and retrovirally transduced with miR-10a over-expression vector or control vector and then stimulated under the neutral condition (anti-IFN-γ and anti-IL-4 antibodies; 10 μg/ml for each) for further 2 days. After that, cells were maintained in the RPMI medium supplemented with 50 U/ml hIL-2 for 2 days. Transduced hNGFR+ cells which hardly expressed IFN-γ (< 0.5 %) and IL-17A (undetectable; data not shown), were isolated by magnetic beads and 1.0 × 106 cells were adoptively transferred to Rag2−/− mice followed by subcutaneous immunization with 100 βg MOG35-55 peptide (Peptides international) in CFA on day 0. These mice were scored daily according to the criteria as previously mentioned5. For analysis of cytokine production in transferred cells, mononuclear cells were isolated from the spinal cord, lymph nodes and spleen and analyzed by intracellular staining with flow cytometry.
To express miR-10a-5p, genomic DNA of mmu-mir-10a were cloned by using primers, 5'-ACCCACAGTGACTTTTCTGCTCC-3' and 5'-GGACACCTCAGGTAGATGAGATTT-3' and inserted between CD19 exon 10 and 11 sequences following Orange2 sequence on pLenti vector (Invitrogen), referred to as pLenti-Orange2-miR-10a, so that miR-10a locus will be transcribed and spliced out as a mirtron (Supplemental Figure 4a). Then miR-10a-5p is expected to be produced from mirtron through endogenous processing mechanism of mature miRNA. pLenti-Orange2-Control that transcribes only a part of CD19 sequence was generated as a control. Orange2 was replaced by truncated hNGFR and whole essential component of interest was transferred into pMYs-puro retroviral vector (Cell Biolabs) generating pMY-hNGFR-miR-10a and pMY-hNGFR-miR-NC (Supplemental Figure 4d). Target sequence for miR-10a-5p, 5'-cacaaattcggtaaacagggta-3', sequence was repeated 8 times with linker sequence between them following Δ4GFP sequence on LV-SFFV vector6, referred to as LV-SFFV-Δ4GFP-miR-10a-5pT (Supplemental Figure 4a). LV-SFFV-Δ4GFP-scT was generated by using 8-time repeated scrambled target sequence as a control. Δ4GFP was replaced by truncated hNGFR and whole essential component of interest was transferred into pMY-puro vector generating pMY-hNGFR-miR10a-5pT and pMY-hNGFR-scT from corresponding lentiviral vectors (Supplemental Figure 4d). To express shRNA specific for Bcl-6 and Ncor2, BLOCK-iT Lentiviral Pol II miR RNAi Expression System (Invigrogen) was utilized. shRNA sequences for Bcl-6 (Mmi505152) and Ncor2 (Mmi520051) were purchased from Invitrogen. pcDNA6.2-GW/EmGFP-miR-neg Control was used to make negative control vectors. Gene segments including EmGFP and shRNA were transferred into pMYs-puro vector generating pMY-EmGFP-iBcl6, pMY-EmGFP-iNcor2 and pMY-EmGFP-Control (negative control) from corresponding original vectors. EmGFP was replaced by truncated hNGFR to generate pMY-hNGFR-iBcl6, pMY-hNGFR-iNcor2 and pMY-hNGFR-Conrol. To over-express Bcl-6, we first made pMY-IRES-hNGFR vector as a control vector by replacing EGFP of pMYs-IRES-GFP vector (Cell Biolabs) with hNGFR. Then Bcl-6 cDNA was appropriately subcloned into pMY-IRES-hNGFR vector to generate pMY-Bcl-6-IRES-hNGFR vector (RV-Bcl-6) for Bcl-6 over-expression. To express Flag-tagged Bcl-6 protein from mRNA including 3'UTR of Bcl-6 or that lacking miR-10a-seed sequence, we generated pMYs-based expression vectors, Bcl-6 and Bcl-6-del described in Figure 2g, which also express hNGFR, using the 3'UTR sequences derived from luciferase reporter construct.
Retroviral vector was transfected into PlatE cells to generate recombinant retrovirus. To perform retroviral transduction of CD4+ T cells, naive CD4+ T cells were stimulated with plate-bound anti-CD3 and CD28 Ab + soluble anti-IFN-γ Ab for 16 hours. Culture medium was replaced with retroviral soup and 4 μg/ml polybrene, followed by centrifugation at 2500 rpm for 2 hours. After 4-hour incubation at 37 °C, viral supernatant was replaced with T cell culture medium containing cytokines and antibodies of interest. 3 days later, T cell differentiation was evaluated by flow cytometry. For retroviral transduction to HEK293T cells and CH12, we spinoculated cells with retroviral soup at 1800 rpm or 2500 rpm, respectively for 90 min and incubate them at 37 °C for 4 hours.
Naïve CD4+GFP−CD62Lhi CD44lo cells were isolated from Foxp3EGFP mice and transferred to TCRa−/− mice. Some naïve T cells were stimulated with 10 ng/ml TGF-β and 50 U/ml rhIL-2 for 6 days and then Foxp3+ (GFP+) cells were isolated by FACS and transferred to TCRα−/− mice. 6 – 7 days after retroviral transduction of naïve CD4+ T cells, hNGFR+ cells were isolated by FACS sorting or magnetic beads and transferred into TCRa−/− mice. 5 or 6 weeks later, spleen and Peyer's patches were collected and analyzed by flow cytometry or immunohistocmistry.
Cytokines, transcriptional factors, and surface markers were evaluated by FACS Canto, Calibur or Verse (BD). In brief, for cytokine detection, cells were stimulated for 2 h with PMA and ionomycin with the addition of GolgiPlug (BD). To exclude dead cells, cells were stained 7-AAD (BD) and washed with PBS twice before fixation. Then cells were fixed and permeabilized with Foxp3 Staining Buffer Set (eBioscience) or BD Cytofix/Cytoperm (BD) and stained with fluorescent antibodies. Events were collected and analyzed with FlowJo software (Tree Star).
Total RNA was isolated by mirVana miRNA Isolation Kit (Applied Biosystems). cDNA synthesis was performed with TaqMan Reverse Transcription Reagents (Applied Biosystems). Quantitative PCR was performed with ABI 7500 Fast Real-Time PCR System using Taqman site-specific primers and probes (Applied Biosystems). For reverse transcription and quantification of miRNA, TaqMan Reverse Transcription Kit was used in combination with TaqMan miRNA assays for snoRNA202 and hsa-mir-10a. Results were properly normalized to β-actin or snoRNA202 levels.
Cell lysate was fractionated on 4–12 % Bis-Tris gel (Invitrogen) SDS-polyacrylamide gel electrophoresis, followed by transfer to a nitrocellulose membrane. The membrane was blocked in blocking buffer (Odyssey) for 30 min and subjected to immunoblots with appropriate antibodies. Bands were visualized using Odyssey infrared imaging system (LI-COR Biosciences). Band intensities were analyzed by ImageJ (available at http://rsbweb.nih.gov/ij/) and normalized by the action intensity and shown as in ratio.
3'UTR of Bcl-6 and Ncor2 were cloned into pmirGLO vector (Promega) from genomic DNA by using the following primers, referred to as Luc-Bcl6 and Luc-Ncor2, respectively: forward; 5'-atgaagcatggagtgttcctcgccctt-3' and reverse; 5'-atctgcaggcagacacggatctgaga-3' for Bcl-6, forward; 5'-tgagacactctcggacagcgagtga-3'and reverse; 5'-aaatgacagaatgccgccgtgcaca-3' for Ncor2. Predicted target sequences of both 3'UTR were removed from each vector by using QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies) and referred to as Luc-Bcl6-del and Luc-Ncor2-del. These constructs were transfected in NIH3T3 cells with expression vector for miR-10a or a control vector. Then firefly luciferase and Renilla luciferase activity were measured with Dual-Glo Luciferase Assay System (Promega).
Peyer's patches were embedded in OCT compound (Sakura Finetechnical). Frozen sections were made by cryostat at 20 μm and stored at −80 °C until use. After fixation with 4 % PFA or acetone, sections were stained with antibodies and analyzed by Leica SP5 NLO Confocal Microscope.
Three vectors to express miR-10a, miR-10a sponge sequence, and Flag-tagged Bcl-6 were transfected to NIH3T3 cells by Lipofectamine 2000 (Invitrogen) as described in the manufacturer's instruction. Two days later, the cells were collected with PBS + 2 mM EDTA and analyzed by flow cytometry.
Appropriate analyses were used and stated in each figure legend. P-value less than 0.05 was considered as significant difference. All error bars show SEM.
We thank Y. Belkaid, W. Chen and A. Villarino for their careful reading of this manuscript; K. Moro (Keio University, Tokyo, Japan) and T. Tamachi (NIH) for their excellent advice on manipulating PP; L. Naldini (San Raffaele Institute) for the LV-SFFV lentiviral vector; and J. Simone (NIH) and J. Lay (NIH) for cell sorting. This study was supported by JSPS Research Fellowship for Japanese Biomedical and Behavioral Researchers at NIH (H.T.) and the NIAMS Intramural Program.