The Transcription Factor T-Bet Is Regulated by MicroRNA-155 in Murine Anti-Viral CD8+ T Cells via SHIP-1

We report here that the expression of the transcription factor T-bet, which is known to be required for effector cytotoxic CD8+ T lymphocytes (CTL) generation and effector memory cell formation, is regulated in CTL by microRNA-155 (miR-155). Importantly, we show that the proliferative effect of miR-155 on CD8+ T cells is mediated by T-bet. T-bet levels in CTL were controlled in vivo by miR-155 via SH2 (Src homology 2)-containing inositol phosphatase-1 (SHIP-1), a known direct target of miR-155, and SHIP-1 directly downregulated T-bet. Our studies reveal an important and unexpected signaling axis between miR-155, T-bet, and SHIP-1 in in vivo CTL responses and suggest an important signaling module that regulates effector CTL immunity.

We report here that the expression of the transcription factor T-bet, which is known to be required for effector cytotoxic CD8 + T lymphocytes (CTL) generation and effector memory cell formation, is regulated in CTL by microRNA-155 (miR-155). Importantly, we show that the proliferative effect of miR-155 on CD8 + T cells is mediated by T-bet. T-bet levels in CTL were controlled in vivo by miR-155 via SH2 (Src homology 2)containing inositol phosphatase-1 (SHIP-1), a known direct target of miR-155, and SHIP-1 directly downregulated T-bet. Our studies reveal an important and unexpected signaling axis between miR-155, T-bet, and SHIP-1 in in vivo CTL responses and suggest an important signaling module that regulates effector CTL immunity.
Keywords: Micrornas, cD8 + T cells, T-bet, influenza, src homology 2-containing inositol phosphatase-1, mir-155 inTrODUcTiOn Cytotoxic CD8 + T cells (CTLs) are an essential component of the adaptive immune system, responsible for the clearance or control of cells infected with viruses or intracellular bacteria and tumors (1,2). Identifying the mechanisms that control the formation of effector and memory CTL is essential for rational vaccine design and immunotherapeutic approaches. Transcription factors have been recognized as crucial elements that control and direct the fate of CTL toward effector or memory differentiation states (3,4). In particular, the transcription factors T-bet, Eomes, Blimp-1, and Bcl-6 have been shown to regulate the differentiation and function of CTL (5). T-bet-and Eomes-deficient mice have shown that the "graded expression" of these transcription factors regulate the early effector and memory phenotype of CTL in response to infection with lymphocytic choriomeningitis virus (LCMV) (3). Specifically, T-bet has been linked to the generation of effector and effector memory CD8 + T cells (6)(7)(8) and the production of the pro-inflammatory cytokine IFNγ by CTL (6,7). Therefore, understanding how T-bet expression is regulated in CTL is critical for our understanding of the pathways that control effector and memory CTL formation.
MicroRNAs (miRNAs) are ~22 nucleotides small, single-stranded non-coding RNAs that are important post-transcriptional regulators of gene expression and play essential roles in modulating key cellular processes (9)(10)(11). miRNAs have been shown to control CD8 + T cell formation, differentiation, and function. In the absence of miRNA's, CD8 + T cells fail to develop normally (12) or aberrantly respond in vivo to Listeria monocytogenes infection (13). However, the regulatory role of specific miRNAs in generating effective CTL responses to viral infection and tumors is only now being elucidated. Indeed, the regulatory effect of miRNAs is not only cell-type specific, but also context dependent and influenced by the activation state of the cell and in vivo inflammatory environment (14). Thus, miRNA expression is expected to change depending on the differentiation and activation state of the CTL and this could also be influenced by the milieu and location in which the CTL response is elicited. Previous in vitro studies have indicated that miRNA expression in CTL changes with differentiation (15) and we have shown that in vivo miR-155 expression levels dynamically change during differentiation from naïve to effector to memory CTL (16).
We, and others, have demonstrated that, in the absence of miR-155, effector CTL responses against acute infections with influenza A virus or L. monocytogenes, and tumors are severely diminished (16)(17)(18)(19). Furthermore, in the absence of miR-155, the generation of memory CTL is decreased (16). This suggested that overexpression of miR-155 could enhance CTL responses to infections and cancer; indeed, we (16) and others (20) have found that overexpressing miR-155 in CD8 + T cells causes significant expansion. Multiple mechanisms and targets by which miR-155-deficiency can affect CTL responses have been proposed including type-I IFN and STAT1/2 signaling (16), γc chain cytokine signaling (20), and SOCS-1 (14,17), which suggests that miR-155 regulation of CTL is context dependent (14). This also raises the question of whether the overexpression of a specific miRNA targets the same mRNA as steady-state endogenous miRNA. Because miRNA and target mRNA availability is a dynamic process (21, 22), it is likely that supranormal levels of a miRNA may allow the targeting of additional mRNA which are normally outcompeted by other targets.
Increasing miR-155 expression in CTL augmented anti-viral effector CTL and skewed memory CD8 + T cells toward an effector memory phenotype. miR-155 overexpression induced enhanced T-bet expression and downregulated the inhibitory phosphatase SH2 (Src homology 2)-containing inositol phosphatase-1 (SHIP-1). T-bet upregulation was necessary for effector CTL augmentation by miR-155. Importantly, we show that SHIP-1 regulated T-bet expression and promoted the effector responses in miR-155-overexpressing CTL. Thus, T-bet expression can be controlled by miR-155 via SHIP-1 signaling and we have revealed a novel regulatory pathway for T-bet expression as well as effector and memory CTL generation.

MaTerials anD MeThODs animals and infections
C57BL/6 Tg(TcraTcrb)1100Mjb/J (OT-I) were backcrossed with B6.SJL-Ptprca Pepcb/BoyJ (CD45.1 + ) mice (both from the Jackson Laboratory) to generate OT-I CD45.1 + mice on the C57BL6/J background. C57BL/6J mice, miR-155-deficient OT-I mice, and T-bet +/-OT-I mice (on the C57BL/6J background) were kept in a barrier facility (certified by the Association for the Assessment and Accreditation of Laboratory Animal Care) at Drexel University College of Medicine, or in a barrier facility at Erasmus University Medical Center. This study was carried out in accordance with the recommendations of the Institutional Animal Care and Use Committee (IACUC) or the Instantie voor Dierenwelzijn (IvD). The protocols were approved by the IACUC or IvD. Female mice 8-10 weeks old were anesthetized with 2.5% isoflurane gas and were infected intranasally with influenza virus strain A/WSN/33-expressing OVA(257-264) (WSN-OVA, a gift from DJ Topham, University of Rochester Medical Center).

Flow cytometry
Cells were stained as previously described (24 After staining, cells were washed twice with Perm/Wash buffer (eBioscience) and fixed with 1% PFA. Anti-T-bet antibody staining was titrated using wild-type (T-bet +/+ ) and T-bet +/splenocytes to achieve clear distinction between heterozygote and homozygote T-bet expression ( Figure S3 in Supplementary Material). For staining of SHIP-1, cells were surface stained as above, then fixed using Fixation Buffer (BioLegend) for 20 min at RT, washed with Perm/Wash Buffer (BioLegend), and stained for 20 min at RT. The following antibodies were used in combination with intracellular flow cytometry: anti-SHIP-1 antibody (clone P1C1-A5). After staining, cells were washed twice with Perm/Wash buffer (BioLegend) and fixed with 1% PFA. All samples were collected with an LSR-Fortessa (BD Biosciences) and analyzed with FlowJo software (Treestar, Ashland, OR, USA).

statistics
For flow cytometry and qRT-PCR data analysis, the normality of the population distribution was assessed using the D' Agostino-Pearson (n = ~8 mice per group) or Shapiro-Wilk (n = <8 mice per group) normality test by GraphPad Prism 7. Significant differences between normally distributed populations were assessed using a two-tailed, unpaired t-test; significant differences between non-normally distributed populations were assessed using a two-tailed Mann-Whitney exact test. The subsequent p-values are denoted in each figure.

Overexpression of mir-155 Promoting enhanced cD8 + T-cell responses
Previously, we demonstrated that the adoptive transfer of low numbers of miR-155-overexpressing OT-I cells resulted in increased expansion and increased viral clearance (16). To further understand the molecular basis of miR-155's impact on CTL, we retrovirally transduced CD45.1 + OT-I CD8 + T cells to overexpress miR-155 (miR-155-OE OT-I). As a control, a retrovirus expressing a scrambled insert was used (scrambled ctrl OT-I). Transduced OT-I cells were sorted and then 10 4 cells were adoptively transferred into CD45.2 + recipient mice that were then infected with WSN-OVA influenza virus. We We did not observe any differences in the activation state (as assessed by CD25 and CD69 expression frequency and MFI) of donor OT-I when miR-155 was overexpressed ( Figure S1D in Supplementary Material). Ki-67 staining of donor OT-I cells in the lungs at day 10 post-influenza virus infection showed that a greater frequency of miR-155-OE OT-I cells were Ki-67 + compared with scrambled ctrl OT-I cells, indicating that miR-155 overexpression enhanced the proliferation of anti-viral CD8 + T cells (p = 0.022) (Figures 1E,F). A higher frequency of miR-155-OE OT-I donor cells produced the effector cytokines IFNγ (60.3 ± 2.0% versus 67.7 ± 1.6%, mean ± SE donor scrambled ctrl OT-I versus miR-155-OE OT-I, respectively, p = 0.008) and TNFα (19.4 ± 2.5% versus 31.9 ± 4.4%, for donor scrambled ctrl OT-I versus miR-155-OE OT-I, respectively, p = 0.02) after peptide stimulation for 6 h (Figures 1G,H). No significant difference was observed in the MFI of cytokine production (data not shown). qRT-PCR analysis of the cytotoxic molecules granzyme A, granzyme B, and perforin assessed in lung donor OT-I cells 9 days post-influenza virus infection showed no difference between scrambled ctrl OT-I and miR-155-OE OT-I ( Figure S1E in Supplementary Material). Furthermore, western blot analysis of granzyme B protein levels confirmed no difference ( Figure S1F in Supplementary Material). These findings demonstrate that increased miR-155 expression is sufficient to promote significant expansion and increased effector cytokine production of CTL during an acute influenza virus infection.

Overexpression of mir-155 skewing cD8 + T cells Toward an effector Memory Phenotype
Because miR-155 affected effector CTL responses, we also assessed its effect on memory CTL formation. We did not observe any differences in the frequencies of SLEC (short-lived effector cell) or MPEC (memory precursor effector cell) populations in donor OT-I cells in the lung or spleen 10 days post-infection (Figures 2A,B). We also did not observe differences in the frequencies or absolute numbers of donor memory OT-I CD8 + T cells in the lungs, spleens, or MLN ( Figure S2A in Supplementary Material) when comparing miR-155-OE OT-I CD8 + T cells to scrambled ctrl OT-I CD8 + T cells. However, upon examination of the memory phenotype, there was skewing of spleen and MLN memory miR-155-OE OT-I CD8 + T cells toward the effector memory (TEM) phenotype (CD44 + CD62L − ) compared with the central memory (TCM) phenotype (CD44 + CD62L + ) ( Figure 2C). The ratios of the numbers of TEM/TCM donor OT-I CD8 + T cells in the spleen (p = 0.02) and MLN (p = 0.005) were skewed to the TEM phenotype ( Figure 2D) and a significantly lower frequency of TCM donor cells in the spleen was observed (p = 0.004) ( Figure S2B in Supplementary Material). We observed no difference in the frequency of CD44 + CD62L − CD69 + CD103 + tissue resident memory (TRM) OT-I populations in the lungs of recipient mice 60 days post-influenza infection ( Figure S2C in Supplementary Material). Unlike at day 10, we did not observe any difference in the frequency of IFNγ-or TNFα-producing donor memory miR-155-OE OT-I CD8 + T cells at day 60 ( Figure S2D in Supplementary Material). However, we did observe an increased expression of CD107a (LAMP-1) ( Figure S2D in Supplementary Material), which indicated increased degranulation of memory miR-155-OE OT-I CD8 + T cells.

T-Bet expression in cTl regulated by mir-155
The transcription factor T-bet has previously been associated with IFNγ production, and effector and effector memory cell generation in T cells during LCMV or L. monocytogenes infection (6,25). We therefore examined whether miR-155 overexpression in CD8 + T cells affected T-bet levels. Tbx21 (T-bet) mRNA levels were assessed by qRT-PCR in total RNA from scrambled ctrl OT-I and miR-155-OE OT-I cells sorted from the lungs on day 9 post-infection. This demonstrated a trending increase in relative T-bet mRNA levels in miR-155-OE OT-I CD8 + T cells (Figure 3A). To assess T-bet expression at the protein level, T-bet expression was measured using intracellular flow cytometry that was first validated by staining T-bet +/+ and T-bet +/-OT-I CD8 + T cells, which as expected (3)     in the lungs (p = 0.01) (Figures 3B,C) and spleen (p = 0.005) (Figure 3D), indicating that T-bet expression in miR-155-OE OT-I CD8 + T cells may be regulated at the translational level as a significant difference was observed only at the protein level and not at the level of mRNA.
Since miR-155-overexpressing CTLs have increased amounts of T-bet, we reasoned that conversely miR-155-deficient CD8 + T cells would express decreased levels of T-bet. To assess this, splenic miR-155 −/− OT-I CD8 + T cells were isolated and activated in vitro with plate-bound α-CD3 and α-CD28 antibodies and cultured for 7 days. In parallel, wild-type T-bet-sufficient OT-I CD8 + T cells and T-bet heterozygote (T-bet +/-) OT-I CD8 + T cells were similarly cultured. We found that T-bet +/-OT-I cells exhibited a gene-dosage effect and expressed half the amount of T-bet compared with wild-type OT-I cells (Figures 3E,F). T-bet staining of miR-155 −/− OT-I CD8 + T cells revealed a 40% (p = 0.0001) reduction in T-bet expression relative to wild-type OT-I CD8 + T cells (Figure 3F). Adoptively transferred miR-155 −/− OT-I cells exhibited significantly decreased T-bet levels compared with wild-type OT-I cells 10 days post-influenza virus infection (p = 0.008) (Figures 3E,G,H). These data provide the first evidence that miR-155 regulates T-bet expression in CTL. Furthermore, miR-155-overexpressing anti-viral CD8 + T cells demonstrate increased levels of T-bet, and thus enhanced T-bet expression may be responsible for the augmented effector generation and effector memory phenotype observed.
To directly test if increased T-bet in miR-155-OE OT-I cells was responsible for augmenting effector generation of anti-viral CTL, T-bet +/-OT-I or wild-type T-bet +/+ OT-I CD8 + T cells were retrovirally transduced to overexpress miR-155 or the scrambled sequence. Transduced cells were then adoptively transferred into host mice followed by infection with WSN-OVA influenza virus. miR-155-overexpressing T-bet +/-OT-I CD8 + T cells (T-bet +/-miR-155-OE OT-I) were incapable of expanding to the same capacity as miR-155-OE OT-I at day 10 post-infection in the lungs of influenza virus-infected animals (p = 0.012) (Figures 4A,B). A similar effect was observed in the spleen (p = 0.006), but not the MLN (Figures 4C,D). These data demonstrate that T-bet is essential to drive the increased expansion of miR-155-overexpressing CD8 + T cells during influenza virus infection.

Overexpression of mir-155 altering the Transcription Factor Profile of Memory cD8 + T cells enhancing the expression of T-Bet and Blimp-1
Since transcription factors such as T-bet, Eomes, Blimp-1, and Bcl-6 are important regulators of CD8 + T-cell memory generation (4) and miR-155 overexpression promoted effector memory CTL formation, we examined the transcription factor profile of miR-155-overexpressing memory CTL. To assess this, TCM and TEM donor cell compartments were sorted 60 days postinfluenza virus infection and mRNA expression levels of Tbx21 (encoding T-bet), Eomes, Prdm1 (encoding Blimp-1), and Bcl6 were measured. We observed that Blimp-1 mRNA expression was modestly increased in miR-155-OE TEM OT-I cells, while Bcl-6 mRNA expression was decreased in miR-155-OE TEM OT-I cells compared with scrambled ctrl OT-I CD8 + T cells (Figure 5A,  top). Interestingly, we observed in both TEM and TCM compartments an increase in T-bet mRNA expression when miR-155 was overexpressed, with the greatest increase observed in the TCM compartment (Figure 5A, bottom). We did not observe any differences in Eomes mRNA expression. Intracellular staining and flow cytometry confirmed increased T-bet expression when miR-155 was overexpressed in both TEM and TCM compartments in the spleen (Figures 5B,C).

cells Modulating T-Bet levels
As T-bet is not predicted to be a direct target of miR-155, we reasoned that miR-155 was regulating T-bet expression indirectly. SHIP-1 has previously been identified as a direct target of miR-155 and has been reported to be affected by miR-155 in CTL (16,20,26). We assessed SHIP-1 expression at both the mRNA and protein levels when miR-155 is overexpressed in CTL. qRT-PCR of sorted day 9 post-influenza virus infection miR-155-OE OT-I cells showed a 25% decrease in SHIP-1 (Inpp5d) mRNA expression levels ( Figure 6A) compared with controls. Western blot analysis ( Figure 6B) and intracellular flow cytometry staining (Figures 6C,D) (Figures 7C,D). Finally, we tested if SHIP-1 directly mediates the effect of miR-155 on T-bet expression. OT-I cells were transduced with either an SHIP-1-overexpressing retrovirus (SHIP-1-OE OT-I) or a dominant-negative SHIP-1 resulting in the inhibition of SHIP-1 signaling (23) (SHIP-1DN OT-I) and adoptive transfers were performed. We show that SHIP-1 overexpression alone is sufficient to decrease T-bet relative to control levels ( Figures 7E,F), while SHIP-1-DN results in significantly increased T-bet expression levels in donor OT-I cells (p = 0.013) ( Figure 7G). Taken together, these data provide further support that miR-155 modulates T-bet expression via SHIP-1.

DiscUssiOn
The importance of miRNA in T cells, particularly CD8 + T cells, has been demonstrated with Dicer-deficient animals (12,13,27), where the lack of mature miRNA has profound effects on the differentiation and function of T cells. Since miR-155 is upregulated in effector CTL and we have previously shown that the absence of miR-155 significantly reduces effective CTL responses to infection (16), we argued that its forced overexpression would enhance the CTL response. Indeed, overexpression of miR-155 in CTL augmented the response, in agreement with previous reports from us and others (16,17,20). In exploring the mechanistic underpinnings of the effect of miR-155 overexpression, we have revealed an unexpected connection between miR-155 and T-bet expression in virus-specific CD8 + T cells, mediated through the SHIP-1-signaling pathway. miR-155 has been shown, by us and others, to control the CTL immune response (14,(16)(17)(18)(19)(20), while the direct targets by which  miR-155 affects CTL responses have been suggested to be context dependent (14). Previous studies have indicated that miR-155 in CD8 + T cells may regulate SOCS-1 and STAT5 in acute and chronic LCMV infection (17) and tumors (17,20), STAT1/ STAT2 and type I IFN signaling in influenza virus infection (16), and c-maf in CD4 + T cells (28) depending on the cell type and the environment in which the response is taking place. To the best of our knowledge, this is the first report that miR-155 can regulate T-bet expression in CTL and miR-155 overexpression leads to increased numbers of effector CTL and a skewing of CTL memory to an effector memory phenotype. This is consistent with T-bet's well-known capacity to promote effector CTL expansion and effector memory CD8 + T-cell generation. Indeed, we find that the increased effector CTL expansion is induced by miR-155 overexpression and is mediated by T-bet as miR-155-overexpressing T-bet +/-CTL fail to expand. In CD8 + T cells, T-bet promotes the differentiation of effector CTL and is associated with TEM CD8 + T cells (3). In the absence of T-bet during a viral infection such as LCMV, CTL differentiation was skewed toward a KLRG1 lo CD127 hi memory precursor phenotype with less formation of terminally differentiated effector cells (3), and memory CD8 + T cells are maintained (29).
Overexpression of T-bet sustains effector CTL during chronic LCMV infection (30). Because T-bet mRNA 3′UTR lacks a miR-155 seeding sequence and is not a known or predicted target of miR-155, we postulated that this is an indirect effect. Indeed, we report that the modulation of T-bet levels by miR-155 is mediated by miR-155's direct target SHIP-1. The relationship between SHIP-1 and T-bet has been previously demonstrated by Tarasenko et al. who observed that SHIP-1-deficient CD8 + T cells expressed 60% more T-bet than control CD8 + T cells (31). This raises the question of how SHIP-1 signaling negatively affects T-bet levels. One pathway that is well established to control T-bet expression is mTOR signaling (8), which is downstream of phosphatidylinositol-3-kinase (PI3K) signaling (32). SHIP-1 is known to inhibit PI3K signaling (33), which would result in reduced mTOR activity and decreased T-bet. miR-155, by suppressing SHIP-1 levels, may amplify PI3K and mTOR signaling, thus resulting in more T-bet. Future studies examining the PI3K pathway when miR-155 is modulated could address this question.
Here, we report that SHIP-1 can be downregulated in CTL as a result of miR-155 overexpression, while miR-155 deficiency leads to increased SHIP-1 levels. SHIP-1 is a bona fide direct target of miR-155 and this downregulation has been shown to play an important functional role in many immune cell subsets including macrophages (26), NK cells (34), and dendritic cells (35) and SHIP-1 levels have been reported to be modulated in CTL by miR-155 overexpression (20). In our current study, we demonstrate that this intrinsic SHIP-1 modulation impacts CTL responses. Although the role of SHIP-1 in immune cells has been extensively studied, the role of SHIP-1 in T cells is less clear. There is conflicting data regarding how SHIP-1 expression impacts the ratio of peripheral CD4 + and CD8 + T-lymphocyte populations (31,36), and the role of SHIP-1 in in vivo CTL responses is questionable. In a previous study by Tarasenko et al., the role of SHIP-1 in T cells was explored using a CD4 cre SHIP fl/fl mice in which all T cells lacked SHIP-1 expression. Although a Th1 skewing of CD4 + T cells was observed, no differences were found in the in vivo expansion of SHIP-deficient CD4 + or CD8 + T cells in mice immunized with ovalbumin (31). However, SHIP-1 may play an important role in CTL in the inflammatory milieu of a viral infection, where dampening inflammatory signaling pathways may be critical for the CTL responses. Alternatively, the effect of SHIP-1 on CTL that overexpress miR-155 may be due to additional miR-155-mediated perturbations in other signaling pathways that change the balance between signaling pathways and raise SHIP-1 to a more prominent functional role.
We find that miR-155 overexpression also affects memory CTL responses. We found that TEM CTL formation was favored by miR-155 overexpression. Indeed, we previously observed that miR-155 was more highly expressed in anti-viral TEM versus TCM (16). This most likely is due to miR-155 overexpression increasing T-bet expression as T-bet is known to promote TEM CTL formation (8,37). However, miR-155 overexpression also increased Blimp-1 and decreased Bcl-6 in memory cells. Thus, the promotion of TEM CTL formation could be due to changes in several transcriptional factors that affect TEM CTL. Interestingly, we found that miR-155-overexpressing TCM CTL also increased T-bet expression in the spleen. Since the numbers of TEM and TCM CTL were not increased in spleens, it is possible that increased T-bet levels in TCM promotes their conversion to TEM, a dynamic process previously observed in vesicular stomatitis virus (VSV)specific and L. monocytogenes-specific memory CTL (38,39), and leads to their preferential migration to sites of inflammation such as the lung or MLN. This is consistent with T-bet's ability to drive TEM formation but also suggests that miR-155 overexpression could regulate plasticity of CTL memory phenotypes.
We report here that miR-155 overexpression leads to increased effector CTL responses and a skewing of memory CTL toward an effector memory phenotype. To the best of our knowledge, our studies conclusively show for the first time a novel connection between miR-155 and T-bet in CD8 + T cells. The link between miR-155 overexpression and T-bet upregulation is essential for the expansion of effector CTL. We show that T-bet is indirectly targeted by miR-155 via miR-155's repression of its direct target SHIP-1. Our studies reveal an unexpected miR-155/SHIP-1/Tbet axis in CTL immunity to viral infection that may play a pivotal role in CTL immunity in the context of infection and cancer.

eThics sTaTeMenT
This study was carried out in accordance with the recommendations of the Institutional Animal Care and Use Committee (IACUC), Drexel University College of Medicine or the Instantie voor Dierenwelzijn (IvD), Erasmus University Medical Center. The protocols were approved by IACUC or IvD.
aUThOr cOnTriBUTiOns JH performed infections, adoptive transfers, retroviral transduction, and flow cytometry. CS, PS, AC, DG, AF, MM, IB-H, LR, and JF performed infections, flow cytometry, RT-PCR, in vitro assays, retroviral production, and mouse breeding. RK and SE cloned and designed vectors. YM performed adoptive transfers and cell sorting. ES, EW, JH, and PK designed and performed T-bet experiments. JH and PK were responsible for study design, data analysis, and manuscript authorship. All authors discussed the results and commented on manuscript.

FUnDing
This work was supported in part by NIH (grant number R01 AI117718, grant number AI105343, grant number P01 AI108545, grant number U19 AI082630, and grant number U19 AI117950 to EW), the Parker Institute (to EW), Worldwide Cancer Research, and Erasmus University Medical Center (to PK).