Background
During an acute intracellular bacterial or viral infection antigen specific CD8+ T cells rapidly proliferate and expand into effector T cells which clear the pathogen. Depending on the strength of stimulating signals received the cells differentiate into specific subsets of memory cells (for a recent review see [
1]). The expression of cell surface makers allows these subsets to be identified using polychromatic flow cytometry [
2]. Although the exact lineage of the relationship between these subsets remains controversial [
3] they can be classified along a progressive path of differentiation based on their phenotype, function, expression of specific transcription factors and more recently their miRNA profile [
1],[
2],[
4]. However, the molecular circuitry that underlies this differentiation process has only recently begun to be elucidated.
miRNAs are small, 21–23 nucleotide long single-stranded molecules of RNA that function to inhibit gene expression post-transcriptionally [
5]. They bind by partial base-pair complementarity primarily to the 3′ UTR of messenger RNAs as part of the RNA induced silencing complex (RISC). It has been predicted that up to 90% of the human transcriptome is regulated by miRNAs [
6]. Furthermore it has been proposed that they function to fine tune gene expression [
7]. Therefore it is not surprising that the complex system of T cell development which requires careful regulation to achieve immune homeostasis is subject to control by miRNAs, as reviewed by Jeker and Bluestone [
4]. An early indication of their involvement in this process was provided by experiments in mice with a conditional deletion of Dicer, a key enzyme in microRNA biogenesis, in the T cell lineage. Impaired peripheral CD8+ T cell development was observed [
8]. More recently deletion of Dicer in mature CD8+ T cells in a mouse model suggested a role for miRNAs in the activation, migration and survival of these cells [
9].
Roles are also emerging for miRNAs in the differentiation in human CD8+ T cells [
10]-[
12]. Such functions for miRNAs have particular relevance to adoptive transfer of T cells for cancer therapy or immune reconstitution following bone marrow and haematopoietic stem cell transplantation. Experiments in mice indicate that the use of less differentiated T cells yields better anti-tumour properties compared to their more differentiated counterparts [
13]-[
16]. Clinical grade protocols designed to obtain and maintain these optimum cells are still being defined although it is clear that cytokines used in culture can affect the differentiation status of cells [
17]. Typically T cells are cultured in IL-2 which tends to drive cells to a more differentiated phenotype [
11],[
18]. Conversely cytokines such as IL-7 and IL-15 appear to conserve or promote a less differentiated phenotype [
19],[
20]. We set out to understand whether human CD8+ T cells cultured in such cytokines showed changes in specific miRNA molecules that have already been linked to differentiation status. Our hypothesis was that expression of specific miRNAs might vary under different cytokine culture regimes, and that this expression might correlate with – and potentially be responsible for – changes in cell surface phenotype and function. If so, miRNA expression might be informative to the design of protocols used to expand T cells
in vitro for use in immunotherapy. In addition a memory subset specific miRNA profile could aid identification of the prime T cells for therapeutic use and potentially identify miRNAs that could be used to genetically modify T cells
in vitro for use in adoptive immunotherapy.
Methods
Cell cultures and stimulation
Peripheral blood mononuclear cells (PBMCs) were harvested from healthy donors after informed consent in accordance with methods approved by the local ethics committee (University of Auckland Ethics Committee, NZ). PBMCs were isolated by gradient separation using Lymphoprep™ (Axis-Shield). CD8+ cells were enriched from PBMCs using the CD8+ T Cell Isolation Kit (MiltenyiBiotech) following manufacturer’s instructions. For initial microarray experiments cells were then labelled with anti-CD4-PE, anti-CD45RO-PECy7 and anti-CD28-APC (all from BD Biosciences), anti- CD45RA-PE-TR (Invitrogen), anti-CD8-APC Cy7 (Biolegend) and anti-CCR7 FITC (R&D) fluorescent antibodies. They were then FACS-purified into naïve (CD8+, CCR7+, CD45RA+, CD45RO-), central memory (CD8+, CCR7+, CD45RA-, CD45RO+) and effector memory subsets (CD8+, CCR7-, CD45RA-, CD45RO+) on FACS Aria™ II (BD Biosciences). Post-sorting analysis of purified subsets revealed greater than 98% purity. Cell surface memory phenotyping was performed using the antibodies listed above and CD62L-PerCpCy5.5 (Biolegend).
For subsequent validation experiments CD3+ or CD8+ cells were enriched from PBMCs using the Pan T Cell Isolation Kit II or CD8 Isolation Kit respectively (MiltenyiBiotech) following manufacturer’s instructions. To further enrich for the naïve cell fraction (CD45RO-) CD45RO microbeads were used (MiltenyiBiotech). For expansion experiments T cells were activated with anti-CD3/CD28-conjugated magnetic beads (Expander Beads, Invitrogen) in 1:1 bead/T-cell ratio in RPMI medium supplemented with 5% human serum and IL-2 and IL-12 at 10 ng/ml. After 48 hours beads were removed and cells were grown in either IL-2 at 10 ng/ml, IL-7 at 10 ng/ml or IL-15 at 10 ng/ml or concentrations as stated in the text. For culture in cytokine in the absence of a TCR stimulus naïve cells were rested in RS5 with either IL-2 or IL-15 +/− IL-7 or IL-7 alone. All cytokines were supplied by Peprotech. Cytokines and medium were replaced every 3–4 days.
Microarrays
FACS sorted cells were washed once in ice-cold PBS and total RNA was purified using the miRVANA kit (Ambion). RNA integrity was assessed using a bioanalyser (Agilent). 500 ngs of RNA were reverse transcribed and labelled using the Flash Tag Biotin HSR kit (Genisphere) and hybridised to Gene Chip miRNA Arrays 1.0 (Affymetrix) according to manufacturer’s protocols. Fluorescent signals were recorded by an Affymetrix scanner 3000 using Gene Chip Operating Software. The statistical software program R was used to analyse the results. The data was pre-processed and normalised using RMA from the R package [
21]. The R package limma was used to check for differential expression, and an empirical Bayes method was used to moderate the t-statistic. In order to adjust for multiple testing, the Benjamini-Hochberg method was used to correct the p-values, and adjusted p-values are reported in the text. The human subset of the results was extracted and all subsequent plots were made on the human transcripts. Microarray data has been deposited in the Gene Expression Omnibus (GEO) database and can be accessed via accession no. GSE54867.
qRT-PCR
Total RNA was purified using the miRVANA kit (Ambion) or where stated with RNA-GEM Tissue Plus (ZyGem). To analyse the expression of specific miRNAs individual cDNAs were prepared from 10 ngs miRVANA RNA using the Taqman miRNA RT kit (ABI) and specific Taqman small RNA primers (ABI) according to manufacturer’s instruction. Real-time PCR reactions were prepared with Taqman FAST mastermix and run on an ABI Prism HT 7900 machine (Applied Biosystems). RNU44 was used as an internal control and relative expression was calculated as ΔΔCt. To analyse the expression of specific mRNAs total cDNA was prepared using the Superscript first strand synthesis cDNA kit (Invitrogen). Real-time PCR reactions were prepared using 10 ngs total cDNA with Taqman probes (ABI) and run as above. HPRT and B2M were used as internal control genes and relative expression was calculated as ΔΔCt. RNA-GEM Tissue Plus was used according to manufacturer’s instructions. 500,000 cells were lysed in 50 uls lysis solution. 2uls of the resulting lysate was used directly in cDNA synthesis reactions for miRNA and total cDNA.
Nucleofection
Cells were electroporated using the Amaxa Human T Cell Nucleofector® Kit (Lonza) according to manufacturer’s instructions for unstimulated human T cells. Per transfection, 5 million cells were combined with 300nM siRNA or miRNA mimic (miRVANA miRNA mimic, Ambion) and pulsed in a Nucleofector II device using program V-024. Equivalent amounts of fluorescently labelled Block iT siRNA (Invitrogen) was used to assess electroporation efficiency in separate control samples. After 48 hours total RNA was purified using the miRVANA kit (Ambion).
Total RNA was purified using the miRVANA kit (Ambion) and CD45 isoform expression was assessed by PCR as described by ten Dam
et al.[
22]. In brief, first strand cDNA was prepared using primer LCA9 (5′-GTAATCCACAGTGATGTTTGC-3′) with 250 ng total RNA and the Superscript first strand synthesis cDNA kit (Invitrogen). cDNA was then amplified using primers LCA2 (5′-ATTGGATCCGCTGACTTCCAGATATGACC-3′) and LCA7 (5′-CCGAGATCTTCAGAGGCATTAAGGTAGGC-3′). PCR products were visualised on a 2% agarose gel.
Discussion
We find that miR-146a is highly expressed in cells cultured in vitro in the presence of IL-2 or IL-15, even in the absence of a TCR stimulus. In addition, expression of this microRNA correlates with a cell surface memory phenotype in vitro and in ex vivo cells. Interestingly, miR-146a expression is suppressed in the presence of IL-7. Our results therefore show that miRNA expression can be significantly affected by the presence of cytokines alone. This has implications when designing expansion protocols for T cells for use in immunotherapy.
There have been some recent reports indicating that the expression of other microRNAs appears to be influenced by cytokines [
11],[
26]-[
28]. However, miR-182 in mouse CD4+ T cells for example, requires both a TCR and IL-2 to be induced; IL-2 alone was insufficient [
26]. Expression of miR-182 drops 4 days post stimulation. In concordance with that result we did not observe up-regulation of miR-182 in our array experiments on stimulated cells exposed to IL-2 for 28 days. Rapid induction of miR-146a has been shown in human monocytes in response to LPS and various microbial components (expression was induced within hours and assessed up to 14 hours post stimulation) [
29]. miR-146a has also been shown to be rapidly induced in response to IL-1β in human lung alveolar epithelial cells (expression was induced within hours and assessed up to 24 hours post stimulation) [
30]. However, to our knowledge this paper represents the first report to show significant modulation of expression of a microRNA in T cells by prolonged exposure to cytokine alone. miR-146a has also been reported to be induced post a TCR stimulus [
23]-[
25] in T cells but we did not observe this in our experiments (data not shown). However we note that our data agrees with data reported for primary human CD4+ cells [
24] where up-regulation of miR-146a was only observed at between 12–14 days post TCR stimulation. Our data suggest that TCR signalling is not causative for miR-146a expression and that it is possibly the IL-2 produced by T cells following TCR stimulation that up-regulates miR-146a.
miR-146a is emerging as a critical regulator of both the innate and adaptive immune systems [
31]. Its function in T cells appears to be to regulate TCR-driven NFkB activation by targeting the signal transducers TRAF6 and IRAK1, such that T cells in mice lacking miR-146a are hyperactive [
25]. More recently it has been implicated in T cell differentiation [
23] which is in concordance with our data. Although nucleofection utilising a miR-146a mimic and human naïve CD8+ cells was not sufficient to drive a full memory phenotype we did observe down-regulation of CCR7 and TRAF6 mRNA comparable to other recent observations [
23]. As CCR7 is not predicted to be a direct target of miR-146a this is likely to be an indirect effect. In addition we note that
ex vivo Tcm which are CCR7+ and Tem which are CCR7- appear to express similar levels of miR-146a. Nevertheless the reductionist approach of nucleofection suggests that one component of CCR7 modulation is miR-146a expression. TRAF6 has been implicated in the generation of memory cells in mice [
32]. As TRAF6 has been shown to be a direct target for miR-146a [
23],[
25] this supports the idea that miR-146a is involved in T cell memory formation or function. However, further experimentation with sustained overexpression of miR-146a will be required to more fully examine the role of this microRNA in T cell memory.
Our nucleofection results also validated FADD as another potential direct target for miR-146a in line with other recent observations [
24]. FADD is an adaptor protein involved in FAS-mediated apoptosis pathway. However, we found no correlation between miR-146a expression and a reduction in apoptosis in our naïve T cells cultured in IL-2 (data not shown), which is in concordance with the findings of others [
23].
Our results suggest that both IL-15 and IL-2 can drive miR-146a expression whilst IL-7 does not. IL-2, IL-7 and IL-15 share the CD132 subunit in their receptors but only IL-2 and IL-15 share the CD122 receptor subunit. This suggests that signalling via CD122, the shared receptor chain common to both of these cytokines, may be driving expression of miR-146a. Concomitant with increased expression of miR-146a we also observed a move towards a cell-surface memory phenotype even in the absence of a TCR stimulus. The effect is less marked than in the presence of a TCR and we hypothesise that both miR-146a expression and cell division are required to achieve a memory phenotype. It is interesting to note that signalling via CD122 has been implicated in T cell memory development [
33] and further supports the hypothesis that miR-146a is involved in formation of a memory phenotype. The fact that IL-7 appears to inhibit the IL-2 or IL-15 driven upregulation of miR-146a indicates that there may be competition for the CD132 subunit shared by all three cytokines, as has been suggested previously [
34]. We note that miR-146a expression is somewhat delayed in cells exposed to IL-2 in the presence of a TCR stimulus compared to its absence (compare Figure
1E with Figure
2). CD25 is the alpha chain of the IL-2 receptor, expression of which has been shown to be up-regulated for a number of days following a TCR stimulus [
35]. CD25 up-regulation may interfere with the CD122 driven signalling which is most likely driving expression of miR-146a. This up-regulation would not occur in cells exposed to IL-2 in the absence of a TCR stimulus, and this may explain the more rapid expression of miRNA-146a in these cells.
A new T cell subset, stem cell-like memory T cells (Tscm), has recently been identified [
13]. It was reported that naïve T cells expanded in the presence of IL-15 and IL-7 retain an early stem cell memory phenotype
i.e. CD45RA + CD45R0 + CD62L + CCR7 + CD95+ [
19]. The presence of IL-7 was found to be unique in its ability to instruct the T cells to a Tscm phenotype, while IL-15 and CD28 co-stimulation appeared to be critical to optimum expansion of the cells. Similar to our observations it was reported that expansion in the presence of IL-2 or IL-15 drive a more memory-like phenotype [
19]. In combination with our results we propose that the presence of IL-7 is required to inhibit the expression of miR-146a and therefore promote an early stem cell phenotype. Therefore expression of miR-146a may be a critical molecular modulator underlying the generation of the optimum T cell for use in immunotherapy.
Our microarray studies on
ex vivo CD8+ T cell memory subsets confirms a correlation between miR-146a expression and a cell surface memory phenotype as has been observed by others [
23]-[
25]. The array data also showed that there is differential expression of 112 miRNAs in the three CD8+ T cell subsets we examined. As each miRNA can affect multiple target genes and a single gene can be targeted by multiple miRNAs [
36] this is likely to affect subsequent gene expression and affect subset function. The number of miRNAs we identified is similar to that observed in another study using a RT-PCR array approach [
10]. Our data expands this earlier data set and helps to define the miRNA fingerprint present in human CD8+ memory subtypes. Up or down-regulation of individual miRNA expression was confirmed in each case where RT-PCR was used to validate microarray results. Although results were not always statistically significant this reflects high levels of inter-donor variability as has been observed previously [
10]. In general the trend of expression was consistent across the 4 subsets tested
i.e. increasing or decreasing progressively from naive through to Tem subsets.
Competing interests
The authors declare that they have no competing interests.
Authors’ contribution
HS designed the study, was involved in performing all of the experiments and wrote the manuscript. DV was involved in the design and execution of experiments and helped to produce figures. AB developed in vitro culture protocols, was involved in the design and execution of sorting cells by FACS. YJH developed in vitro culture protocols and supplied cells. VFe helped with microarray analysis. NL and NB helped with nucleofector experiments. VFa helped with microarray analysis. AD was involved in flow cytometry experiments. RD contributed to manuscript drafting and provided critical input. All authors read and approved the final manuscript.