Argonaute 2 immunoprecipitation revealed large tumor suppressor kinase 1 as a novel proapoptotic target of miR‐21 in T cells

MicroRNA (miR)‐21 is an important suppressor of T‐cell apoptosis that is also overexpressed in many types of cancers. The exact mechanisms underlying the antiapoptotic effects of miR‐21 are not well understood. In this study, we used the Jurkat T‐cell line as a model to identify apoptosis‐associated miR‐21 target genes. We showed that expression of miR‐21 rapidly increases upon αCD3/αCD28 activation of Jurkat cells. Inhibition of miR‐21 reduced cell growth which could be explained by an increase in apoptosis. MicroRNA target gene identification by AGO2 RNA‐immunoprecipitation followed by gene expression microarray (RIP‐Chip) resulted in the identification of 72 predicted miR‐21 target genes that were at least twofold enriched in the AGO2‐IP fraction of miR‐21 overexpressing cells. Of these, 71 were at least twofold more enriched in the AGO2‐IP fraction of miR‐21 overexpressing cells as compared to AGO2‐IP fraction of control cells. The target gene for which the AGO2‐IP enrichment was most prominently increased upon miR‐21 overexpression was the proapoptotic protein LATS1. Luciferase reporter assays and western blot analysis confirmed targeting of LATS1 by miR‐21. qRT‐PCR analysis in primary T cells showed an inverse expression pattern between LATS1 transcript levels and miR‐21 upon T‐cell stimulation. Finally, LATS1 knockdown partially rescued the miR‐21 inhibition‐induced impaired cell growth. Collectively, these data identify LATS1 as a miR‐21 target important for the antiapoptotic function of miR‐21 in T cells and likely also in many types of cancer.

In addition, we showed that activation-induced miR-21 provides critical antiapoptotic signals in memory T cells allowing long-term survival [11,14]. However, the miR-21 target genes responsible for the protection against activation-induced apoptosis of T cells remain unknown.
In this study, we set out to investigate miR-21 target genes related to its antiapoptotic effects on T cells. We employed an experimental RNA-immunoprecipitation followed by gene expression microarray (RIP-Chip)based approach [15] in Jurkat cells. These cells are a commonly used model to study regulatory pathways involved in T-cell activation and apoptosis. We identified the proapoptotic large tumor suppressor kinase 1 (LATS1) as the miR-21 target gene whose enrichment in the argonaute 2 immunoprecipitated (AGO2-IP) increased the most upon miR-21 overexpression and showed its role in the antiapoptotic effect of miR-21.

Results and discussion
Jurkat is a suitable model to study the antiapoptotic role of miR-21 To assess if the Jurkat cell line is a suitable model to study the function of miR-21 in relation to apoptosis, we determined miR-21 expression levels in unstimulated cells and after stimulation with aCD3/aCD28. In comparison to other miRNAs known to be expressed in T cells at high (miR-17) or low (miR-146a) levels, miR-21 levels were moderate in unstimulated cells (Fig. 1A). Activation of Jurkat cells with aCD3/ aCD28 for 3 days revealed a marked induction of miR-21 expression (≥25-fold, P ≤ 0.001; Fig. 1B) consistent with previous studies on aCD3/aCD28stimulated primary T cells [11,13]. To determine whether loss of miR-21 resulted in a growth defect, we infected Jurkat cells with a miR-21 inhibitor vector which coexpresses GFP. GFP analysis over time of a mixture of transduced and nontransduced cells (GFP competition assay) revealed a significant decrease of miR-21 inhibitor-transduced (GFP-positive) cells when compared to nontransduced (GFP-negative) cells (Fig. 1C). Cells transduced with three nontargeting (NT) control inhibitors showed no effect on cell growth in the GFP competition assay ( Fig. 1C and data not shown). Jurkat cells transduced with miR-21 inhibitor showed a significant decrease of viable cells starting at day 4, which was not observed with control transduced cells (Fig. 1D). This effect was paralleled by an increase of apoptotic cells reaching > 80% at day 6 ( Fig. 1E,F). These findings are consistent with our reported findings in primary T cells [11] and indicate that endogenous levels of miR-21 in Jurkat cells provide an essential antiapoptotic signal. Together, these data show that the Jurkat cell line is a suitable model to study the antiapoptotic properties of miR-21 in T cells.

miR-21 target genes involved in apoptosis
To identify the antiapoptotic miR-21 target genes, we performed AGO2-RIP-Chip on Jurkat cells overexpressing miR-21 (Jurkat-miR-21) and used cells transduced with an empty vector construct (Jurkat-EV) as a control. The miR-21 levels showed an increase of 22-fold in Jurkat-miR-21 compared to Jurkat-EV ( Fig. 2A). Overexpression of miR-21 in Jurkat cells did not cause any obvious effects on the percentage of live cells (data not shown). The efficiency of the AGO2 immunoprecipitation (AGO2-IP) as determined by western blot was comparable between Jurkat-EV and Jurkat-miR-21 cells (Fig. 2B). As expected, miR-21 was strongly enriched in the Jurkat-miR-21 AGO2-IP fraction in comparison to the Jurkat-miR-21 total fraction and the Jurkat-miR-21 and Jurkat-EV IgG1 control IP fractions. Some miR-21 enrichment could also be observed in the Jurkat-EV AGO2-IP fraction as Jurkat cells endogenously express moderate miR-21 levels (Figs. 2C and 1A).
Gene set enrichment analysis (GSEA) [16] revealed a strong enrichment of multiple microRNA (miRNA)binding motifs in both Jurkat-miR-21 (8 of the top 10) and Jurkat-EV (9 of the top 10, Table 1) further validating the efficiency of the AGO2-RIP. The miR-21binding motif increased from the 46th position of most enriched gene sets in Jurkat-EV (false discovery rate (FDR) = 0.0013) to the 28th position in Jurkat-miR-21 cells (FDR ≤ 0.001). Comparison of both top-10 enriched gene sets revealed a marked overlap between Jurkat-miR-21 and Jurkat-EV cells with eight shared gene sets (Table 1). A marked difference was observed for two apoptosis-related gene sets. Genes regulated upon treatment with the growth and survival factor, IL-6, were among the top-10 most enriched gene sets in Jurkat-EV but not in Jurkat-miR-21 (position 258). Genes involved in sensitivity to TRAIL-induced apoptosis were found among the top-10 most enriched in Jurkat-miR-21 but not in Jurkat-EV cells (position 1237). The expression levels of genes represented by the latter gene set showed an overall decrease in the total fraction of Jurkat-miR-21 as compared to Jurkat-EV (not shown). These differences can be explained by either direct or indirect effects of miR-21 and fit with the observed antiapoptotic role of miR-21 in Jurkat cells. Comparison of the expression levels of all predicted miR-21 target genes (209 of 14 514 unique genes) between the total fractions of Jurkat-EV and Jurkat-miR-21 revealed a systematic decrease in transcript levels in Jurkat-miR-21 cells (Fig. 3A). This indicates that transcript levels of predicted miR-21 genes were decreased upon miR-21 overexpression. As a control, we also analyzed differences in the expression levels of predicted miR-146a targets (162 of 14 514) and predicted miR-17 targets (929 of 14 514).
No difference was observed for the predicted miR-146a target genes, while a mild decrease was observed for predicted miR-17 target genes (Fig. 3A). The latter observation can be explained by the marked overlap between the miR-21 and miR-17 predicted target genes, i.e., 62 shared predicted target genes. In line with these observations, we noted a specific enrichment of predicted miR-21 target genes among the top-1500 genes enriched in the IP of Jurkat-miR-21 (n = 72) as compared to the IP of Jurkat-EV cells (n = 48). The enrichment was even more pronounced among the top-250 enriched genes (n = 20 versus n = 10; Fig 3B).
As a control, we also analyzed enrichment of the predicted target genes of miR-146a and miR-17, which revealed no differences between Jurkat-EV and Jurkat-miR-21 (Fig. 3B). Together, these data show an efficient enrichment of miRNA target genes in the IP fractions of both conditions and a marked enrichment of miR-21 predicted target genes in Jurkat cells overexpressing miR-21.
To determine the relevance of the increase of LATS1 for the miR-21 inhibition-induced phenotype (Fig. 1C), we studied the effect of miR-21 inhibition in LATS1-knockdown (KD) cells. Stable LATS1-KD cells were generated by infection of Jurkat cells with lentiviral LATS1-shRNA constructs or NT control shRNA vectors. Western blotting for LATS1 in sorted LATS1-KD cells showed that the efficiency of the shRNAs ranged between 70% and 90% (Fig. 5A). Next, we infected the LATS1-KD and control cells with miR-21 inhibitor and control inhibitor virus, containing GFP and monitored the GFP percentage over time within the LATS1-KD cells. As expected, miR-21 inhibition caused a strong reduction in the percentage of GFP+ cells in the NT shRNA-infected and wild-type cells with on average 15% of GFP+ cells left after    22 days (Fig. 5B). Knockdown of LATS1 was found to partially rescue this effect as a more than 2.5-fold higher percentage of GFP+ cells was left (average of the three shRNAs, 38%) after 22 days of miR-21 inhibition (P-value ≤ 0.05, Fig. 5B). These results show that the proapoptotic effect observed upon miR-21 inhibition is at least in part mediated by downregulation of LATS1.
To study the relevance of the miR-21-LATS1 axis in primary T cells, we analyzed LATS1 and miR-21 levels of primary sorted na€ ıve T cells (CD4+CD45RO-) stimulated for 3 days with a-CD3 and a-CD28. This revealed an inverse pattern, i.e., increased levels of miR-21 and decreased levels of LATS1 upon stimulation (Fig. 6A,  B). The increase of miR-21 upon stimulation supports our previous findings and those of others [11][12][13]20]. The inverse expression pattern of LATS1 suggests that targeting of LATS1 by miR-21 is highly relevant for survival of primary T cells. We also studied miR-21 and LATS1 expression in sorted na€ ıve (CD4+CD45RO-) and memory (CD4+CD45RO+) T cells. In line with what we and others have previously shown, we observed higher levels of miR-21 in memory T cells compared to na€ ıve T cells (Fig. 6C) [11,12,20]. However, we did not observe decreased LATS1 transcript levels in memory T cells as compared to na€ ıve T cells (Fig. 6D). This suggests that miR-21 does not lower the LATS1 RNA levels in memory cells, but regulates LATS1 protein level by post-transcriptional repression of protein translation. To confirm regulation of LATS1 protein levels by miR-21 specifically in memory T cells, LATS1 protein analysis in na€ ıve and memory T cells as well as AGO2-RIP experiments should be performed.
In summary, we showed that the Jurkat cell line is a suitable model to study the role of miR-21 in the regulation of T-cell apoptosis. We experimentally identified multiple miR-21 target genes via employing AGO2-RIP-Chip, including the proapoptotic LATS1 gene. We showed that LATS1 is a bona fide miR-21 target whose knockdown can at least partially rescue the proapoptotic effect of miR-21 inhibition. Thus, LATS1 is likely to be an important target for the antiapoptotic role of miR-21 in (activated) T cells, possibly in combination with other targets such as PDCD4. As miR-21 is widely overexpressed in a variety of cancers, it is also of interest to further study the relevance of the miR-21 target LATS1 in relation to cancer.

Viral constructs
To generate the lentiviral miR-21 overexpression construct, pre-miR-21 with~150 nucleotides of the flanking sequence was amplified from genomic DNA using forward 5 0 -gtcagaatagaatagaattgggg-3 0 and reverse 5 0 -gctgcattatggcacaaaag-3 0 primers. NheI and XhoI restriction sites were added to the forward primer and an EcoRI site was added to the reverse primer to allow directional cloning into the retroviral MXW-PGK-IRES-GFP vector [22] using standard laboratory procedures. To stably inhibit miR-21 function, we used a lentiviral miR-21 inhibition vector (pmiRZip-21; Cat. Nr: MZIP21-PA-1) and three NT lentiviral inhibitor vectors (control inhibitor, Cat. Nr: MZIP000-PA-1, both from Systems Biosciences, Mountain View, USA and shNT1 and shNT2, see below) as controls. The sequences of shRNAs against LATS1 and controls used for cloning to the BamHI and EcoRI sites of the lentiviral pDsREDPuro vector (MZIP/pGreenPuro vector with the copGFP replaced by dsRED, Systems Biosciences) are as follows: LATS1-sh1-S: 5 0 GATCCGCTGCTCCT TCGTCATATACATTCAAGAGATGTATATGACGAAGG

Virus production and viral transduction
Lentiviral particles were produced with a third-generation lentiviral system in 293T cells by CaPO 4 transfection as described previously [23]. Lentiviral transduction of Jurkat cells was carried out for 24 h in the presence of 4 lgÁmL À1 polybrene (Sigma-Aldrich, St. Louis, USA). Retroviral particles were produced by calcium phosphate (CaPO 4 )-mediated transfection of Phoenix-Ampho packaging cells with 10 lg pMXW-PGK-IRES-GFP-miR-21 (miR-21 overexpression) or pMXW-PGK-IRES-GFP-EV (control) and 0.63 lg of pSuper-DGCR8 in T25 flask. Retroviral particles were collected 48 h after transfection, passed through a 0.45 lm Millex-HV filter (Millipore, Amsterdam, The Netherlands) and concentrated with Retro-X concentrator (Clontech, Saint-Germain-en-Laye, France) according to the manufacturer's protocol. Jurkat cells were transduced with the virus by spinning at 1200 g for 2 h.

GFP competition assay
GFP percentage of pmiRZip-21-or pmiRZip-scrambledinfected Jurkat cells was followed over a period of 22 days. The starting GFP percentage varied between 30% and 40%. Data were acquired on FACS Calibur flow cytometer (BD Biosciences, San Jose, CA, USA) and analyzed using FLOWJO software (version 7.6, Treestar, Ashland, OR, USA). The GFP percentage analyzed at the first day of measurement (day 4) was set to 1. The GFP competition assay was performed 3 times.

Apoptosis measurement
Percentages of apoptotic cells were assessed in Jurkat cells transduced with miR-21 or control inhibitor in > 95% of the cells on day 4, 6, and 8 following viral transduction by FACS-based measurement of mitochondrial transmembrane potential loss. Briefly, cells were stained for 20 min at 37°C in cell culture medium containing 50 nM DiLC1 (Enzo Life Sciences, NY, USA), which was followed by a washing step with PBS. Cells were kept on ice and DiLC1 staining was measured at the FACS Calibur flow cytometer using Cell Quest software (BD Biosciences). Data were analyzed using Kaluza Flow Analysis Software (Beckman Coulter).

Quantitative RT-PCR
Total cellular RNA was extracted using the miRNeasy Mini Kit (Qiagen) following the manufacturer's instructions. The RNA quantity was measured on a NanoDrop ND-1000 Spectrophotometer (Thermo Scientific).
cDNA synthesis for mRNA was performed using Superscript III RTase (Thermo Scientific). The qPCR reaction was performed using qPCR MasterMix Plus. Taqman gene assay was used for detection of LATS1: Hs01125523_m1; Thermo Scientific). Primers and probe (Integrated DNA Technologies, Coralville, USA) used for detection of TBP were as follows: forward 5 0 -GCCCGAAACGCCGAAT AT-3 0 , reverse 5 0 -CCGTGGTTCGTGGCTCTCT-3 0 . Mean cycle threshold (C t ) values were quantified with the Sequence Detection Software (SDS, version 2.3, Thermo Scientific), using ABI7900HT thermo cycler (Thermo Scientific). Relative expression levels were determined using the 2 ÀDCt formula, where DCt = Ct gene À Ct ref.gene .

Functional annotation analysis
The functional annotation of genes was performed using the DAVID BIOINFORMATIC RESOURCES 6.7 (https://david.ncifcrf. gov/), based on the following GO categories: GOTERM_BP_-FAT, GOTERM_CC_FAT, GOTERM_MF_FAT, KEGG_-PATHWAY, and SP_PIR_KEYWORDS. About 1-4 GO terms were considered for description of each gene.

Gene set enrichment analysis
Gene sets significantly enriched in the AGO2-IP in comparison to T fraction of Jurkat-EV and Jurkat-miR-21 were determined by the GSEA using the Molecular Signatures Database (GSEA; http://software.broadinstitute.org/gsea/ index.jsp) [16]. Lists containing the expression values of 14 415 genes detected in IP and total fractions of Jurkat-EV or Jurkat-miR-21 were uploaded for the analysis.

Statistical analysis
For comparison of qRT-PCR data of nonstimulated and stimulated Jurkat cells, we applied the Friedman repeated measurements nonparametric test. Data from day 0 were compared to other days of stimulation (days 1, 2, 3). For comparison of viable and apoptotic Jurkat cells upon transduction with control or miR-21 inhibitor, we applied two-way repeated measures ANOVA with a Bonferroni post-test. To determine whether miR-21-inhibited cells have a significant impaired cell growth as compared to control inhibitor-infected cells, we performed mixed model analysis as described previously [25]. Significance for (RL/ FL) luciferase ratios between control and miR-21 inhibitor was calculated using unpaired t-test. The same test was used to determine whether the remaining percentages of GFP+ cells in LATS1-KD group (LATS1 sh1-3) infected with a miR-21 inhibitor were significantly different from the remaining GFP percentages within the controls group (wild-type, NT1 and NT2) infected with a miR-21 inhibitor. For comparisons of LATS1 and miR-21 in nonstimulated and stimulated CD4+ T cells, we applied the Mann-Whitney test. Statistical analysis were performed with GRAPH-PAD Prism version 5.0 (GRAPHPAD Software, San Diego, CA, USA) or SPSS Statistics version 22.0 (IBM Corp. Armonk, NY, USA).