Peripheral blood CD8⁺ T lymphocytes were purified from healthy volunteers in agreement with local ethics committees (Cantonal Commission of Ethics in Research, State of Vaud, Switzerland, authorization #87/06) as described 4 . Briefly, total CD8⁺ T cell preparations were obtained from leukapheresis by magnetic bead enrichment (Myltenyi Biotech, Bergish Gladbach, Germany) and stained with a combination of antibodies to CD3, CD8, CD45RA, CCR7, CD27 and CD28. All antibodies were from BD Pharmingen (San Diego, CA) except anti-CD45RA purchased from Beckman Coulter (Paris, France). The different subsets were then separated on a FACS Aria device (BD Bioscience, Allschwil, Switzerland) to routinely over 99% purity cell suspensions. Cells were either sorted directly to RNA lysis buffer (RNAlater, Qiagen) and frozen (for single specific qPCR assays) or into cold medium (for TaqMan Low Density Array experiments), before washings in ice-cold PBS and lysis (miRVana kit, Ambion).

The TaqMan^® Low-density arrays (TLDA, Applied Biosystems, Foster City, CA) were used following manufacturer's instructions to simultaneously detect expression of 364 individual microRNAs. Briefly, total RNA was extracted with the miRVana kit (Ambion), quantified with a Nanodrop Spectrophotometer, and microRNA mature forms were reverse transcribed (30 min at 16°C, 30 min at 42°C, 5 min at 85°C) with Multiplex RT human primers and 100 mM dNTPs, 50 U MutliScribe reverse transcriptase, 20 U RNase inhibitor, 10 × RT buffer (8 reactions with pools of 48 stem-loop RT primers, 100 ng RNA per reaction). cDNA were then amplified on microfluidic cards in an ABI Prism 7900 HT with single microRNA specific primers (40 cycles of 15 s at 95°C and 1 min at 60°C). The same protocol was applied for single microRNA specific qRT-PCR (miR-17-3p, miR-17-5p, miR-19b, miR-20a, miR-92, miR-21, miR-155, miR-142-3p, miR-142-5p, and RNU44) from 10 ng total RNA, reverse transcribed with TaqMan Reverse transcription MicroRNA kit (Applied Biosystems) and amplified using Universal Fast Start Rox Probe Master Mix (Roche) and microRNA Assay kits in 384 well plates (Applied Biosystems) on a ABI Prism 7900 HT device (Applied Biosystems). Ct (RNU44) was subtracted to Ct (microRNA) to calculate relative expression (ΔCt).

Statistical significance was evaluated with the GraphPad Prism software on non normalized expression data using the Friedman test (for non-parametric, paired data set) with Dunn's multiple comparison correction applied. Results were considered significant for p values < 0.05.

Different subsets of human CD8⁺ T lymphocytes can be defined based on the expression of CCR7, CD45RA and CD28 4 . Naïve cells are CCR7⁺ (and CD45RA⁺ CD28⁺) while the effector memory subsets are CCR7^-. Among the latter, the CD45RA⁺ CD28^- cells representing the end stage of differentiation (late effector memory, L-EM), while CD45RA^- cells can be further divided in two functionally distinct subsets, with CD28⁺ cells (early effector memory (E-EM) closer to central memory cells) being less differentiated than CD28^- subsets (I-EM), which share features with fully differentiated effectors (L-EM).

These human CD8⁺ T lymphocyte subsets (naive, E-EM, I-EM and L-EM, ordered from least to most differentiated) were purified from peripheral blood leukocytes from 3 healthy donors as indicated (Figure 1A), and the relative expression of 365 microRNAs was analyzed in these subpopulations using TaqMan Low Density Arrays (TLDA). Unsupervised clustering of relative expression levels in all subsets identified 4 main groups of microRNAs (Figure 1B): highly expressed (group A, 22 microRNAs, see Table 1), intermediate (group B, 31 microRNAs, see Table 1), low (group C, 46 microRNAs) and rare/absent (group D, 267 microRNAs). In total, 97 microRNAs were expressed well over detection limit, a figure similar to the 113 microRNAs recently detected in B cells using the same TLDA 17 . This list is likely not comprehensive, since the set of microRNAs detected with TLDA does not cover the whole range of microRNAs cloned in human. However, the low number of microRNAs considered expressed at high levels (group A) is in agreement with previous cloning studies in mouse CD8⁺ lymphocytes 15 18 showing that a few highly expressed microRNAs prevailed in frequency. Interestingly, these microRNAs include miR16, miR-21, miR-142-3p and miR-142-5p, which were also found in the group A in human CD8⁺ T lymphocytes, with miR-142-3p showing the highest relative expression levels (Table 1). MiR-26a/b and miR-146a/b also clustered to this group. In addition, 7 microRNAs of the 17~92 and paralog 106b~25 clusters (namely miR-19a, miR-19b, miR-20a, miR-25, miR-92, miR-93 and miR-106b) were identified among the 53 most expressed microRNAs (groups A and B, see Table 1). MiR-17-5p and miR-17-3p are expressed at lower levels (belong to group C), indicating intra-cluster differential expression.

Transgenic overexpression of miR-17~92 cluster in mouse lymphocytes was shown to induce lymphoproliferative disease 16 . High expression levels are therefore likely to be tightly controlled. The high expression levels found for a large set of microRNAs from the 17~92 cluster in primary human CD8⁺ T lymphocytes had not been reported yet, and strongly suggests a role for this set of microRNAs in the biology of CD8⁺ T cells.

To investigate if defined microRNA profiles can be associated with CD8⁺ T cell differentiation status, the relative expression levels of the microRNAs from groups A, B and C were calculated for each CD8⁺ lymphocyte subset relative to those found in the naive cells. Global unsupervised clustering did not identify clear microRNA clusters, and grouped subpopulations from the same donor together, due to strong inter-donor heterogeneity. However, specific patterns could be highlighted when the analysis was focused on a more restricted set of microRNAs (Figure 2A), chosen for their suggested role in immunity. The well expressed miR-21, miR-155 and miR-146a clustered together as consistently upregulated, while the abundant microRNAs of the miR17~92 clusters (miR-19b, miR-20a and miR-92) showed a clear trend towards decreased expression in differentiated cells, as did miR-26a (Figure 2A).

MiR-21 has been shown to be upregulated in mouse CD8⁺ T cells upon in vitro activation 15 , but its function in these cells remains unknown. MiR-155, encoded by the BIC transcript, is induced upon antigen-receptor triggering in lymphocytes and was shown to be instrumental for the generation of adaptive immunity in mouse 19 20 . Our TLDA analysis suggested that both miR-21 and miR-155 were more expressed in in vivo differentiated human primary CD8⁺ T cells than in their naive counterparts (Figure 2A). These results were confirmed by single specific RT-qPCR on eight additional donors, which all clearly showed higher expression of both miR-21 and miR-155 in antigen-experienced cells (Figure 2B).

Altogether these results demonstrate that an increase of miR-21 and miR-155 expression occurs upon in vivo differentiation of human CD8⁺ T lymphocytes. Dynamic regulation of these microRNAs in mouse CD8⁺ T cells has been described upon in vitro activation 15 . By contrast, the data presented here on ex vivo sorted lymphocytes revealed for the first time modulations that occurred during in vivo differentiation of human peripheral CD8⁺ T cell subsets.

Since central memory (CM) CD8⁺ T cells are present at extremely low frequency in peripheral blood, a new sorting strategy was then designed to include this subset in our analysis. Addition of anti-CD27 to the panel of cell surface molecule specific antibodies allowed the isolation of naïve (CCR7⁺ CD45RA⁺ CD27⁺ CD28⁺), CM (CCR7⁺ CD45RA^- CD27⁺ CD28⁺), E-EM (CCR7^- CD45RA^+/- CD27⁺ CD28⁺), I-EM (CCR7^- CD45RA^+/- CD27⁺ CD28^-) and L-EM (CCR7^- CD45RA^+/- CD27^- CD28^-) subsets. Consistent with our previous analysis (Figure 2B), upregulation of miR-155 in antigen experienced cells versus their naïve counterparts was confirmed in the new set of 9 independent donors, with a clear trend toward highest expression levels in the most differentiated CD8⁺ T cells (L-EM) (Figure 3A). Since the TLDA analysis showed that both miR-142-3p and miR-142-5p were among the most expressed microRNAs in CD8⁺ T cells (Table 1), their expression levels were assessed in the different subsets. No significant regulation of expression could be found, likely suggesting a constitutive role rather than a differentiation induced function for these microRNAs (Figure 3B).

Seven members of the 17~92 cluster, which recently were shown to be critical regulators of lymphocyte development and proliferation 16 , were identified among the 53 most expressed microRNAs in human CD8⁺ T cells (Table 1), with a trend toward lower expression in more differentiated cells (Figure 2A). Expression levels of miR-17-3p, miR-17-5p, miR-19b, miR-20a and miR-92 were therefore determined by single specific qPCR in differentiated CD8⁺ T cell subsets, and compared to the levels found in naïve cells. In agreement with the TLDA data, and despite a relatively important inter-donor variability, the expression of some of these miRNAs was found to be preferentially associated with specific subsets of CD8⁺ T cell differentiation. Interestingly, the expression level of miR-20a was significantly increased in the central memory subset, in contrast to more differentiated subsets. Along the same lines, miR-17-3p expression was significantly decreased in late effector memory cells. There were also non significant trends towards preferential expression of miR-19b and miR-92 in the central memory cells. MiR-17-5p expression showed no association with CD8⁺ T cell differentiation. On the whole, expression of several members of the miR-17~92 cluster appeared to be found preferentially during early memory differentiation. Interestingly, not all members are regulated concomitantly, indicating a differential intra-cluster regulation in agreement with the TLDA results (Table 1).

Altogether, the data presented here demonstrate for the first time that human CD8⁺ T cell subsets express a limited set of microRNAs, which include several members of the 17~92 and related clusters. Moreover, this study shows that the in vivo differentiation from naive cells is associated with a stage-specific regulation of distinct microRNA expression, with miR-21 and miR-155 being upregulated in antigen-experienced subsets, the more differentiated subsets expressing higher levels of these microRNA (Figures 2B and 3A). MiR-155 has been suggested to be part of a negative feedback loop downstream the Toll-like receptor/Interleukin 1 receptor pathway 21 . The results presented here may suggest similar functions in human CD8⁺ T cells downstream of the TCR. Moreover, miR-155 has been recently shown to target SOCS-1 expression, both in tumor cells 22 and in mouse CD4⁺ regulatory T cells, in which it modulates the cellular responses to IL-2. Whether miR-155 plays a similar role in CD8⁺ T cells remains an interesting issue to investigate, as it might shed new light on the relationship between cytokine signaling and lymphocytes differentiation.

In contrast to the clear upregulation observed for miR-155, the expression of the miR-17~92 cluster (especially miR-17-3p and miR-20a) tended to decrease along differentiation. Our results are in line with a recent report by Hackl and colleagues 23 that shows downregulation of this cluster in CD28 negative CD8+ T cells. Considering the role of the miR-19~92 cluster in lymphocyte development and proliferation, it is tempting to speculate that its expression may be related to the greater proliferative potential and memory precursor-like capacity, characteristic of central memory cells compared to more differentiated subsets (in particular late effector memory cells). Interestingly, expression of this cluster is induced by the transcription factor c-Myc 24 , which has been shown in the mouse to be a component of the IL-15-dependent pathway controlling homeostatic proliferation of memory CD8⁺ T cells 25 . In addition, the 17~92 cluster has been experimentally shown to inhibit Bim expression 16 , which plays a key role in activation induced cell death in lymphocytes 26 . Whether 17~92 induced regulation of Bim plays a function in CD8⁺ T cells survival potential (greater in less differentiated cells) is an interesting issue to be further investigated.