TGF-β blockade drives a transitional effector phenotype in T cells reversing SIV latency and decreasing SIV reservoirs in vivo

HIV-1 persistence during ART is due to the establishment of long-lived viral reservoirs in resting immune cells. Using an NHP model of barcoded SIVmac239 intravenous infection and therapeutic dosing of anti-TGFBR1 inhibitor galunisertib (LY2157299), we confirm the latency reversal properties of in vivo TGF-β blockade, decrease viral reservoirs and stimulate immune responses. Treatment of eight female, SIV-infected macaques on ART with four 2-weeks cycles of galunisertib leads to viral reactivation as indicated by plasma viral load and immunoPET/CT with a 64Cu-DOTA-F(ab’)2-p7D3-probe. Post-galunisertib, lymph nodes, gut and PBMC exhibit lower cell-associated (CA-)SIV DNA and lower intact pro-virus (PBMC). Galunisertib does not lead to systemic increase in inflammatory cytokines. High-dimensional cytometry, bulk, and single-cell (sc)RNAseq reveal a galunisertib-driven shift toward an effector phenotype in T and NK cells characterized by a progressive downregulation in TCF1. In summary, we demonstrate that galunisertib, a clinical stage TGF-β inhibitor, reverses SIV latency and decreases SIV reservoirs by driving T cells toward an effector phenotype, enhancing immune responses in vivo in absence of toxicity.


Introduction
Interruption of antiretroviral therapy (ART) leads to rapid rebound of viremia in the vast majority of people living with HIV-1 (PLWH) due to the establishment of a persistent HIV-1 reservoir early after infection (1,2).A key mechanism of this persistence is the ability of HIV-1 to enter a state of virological latency characterized by the silencing of viral gene expression and/or lack of viral proteins translation (3,4).This allows the virus to remain invisible to the immune system and latently infected cells to survive and proliferate by homeostatic or antigen-driven proliferation (5,6).Of note, the viral reservoir was initially thought to be stable.However, recent evidence suggests that stochastic HIV reactivation under ART occurs, and selective killing is favored in cells bearing replication competent virus integrated in transcriptionally active sites within the genome (7)(8)(9).Hence, integration site, but especially the activation status of the infected cells profoundly influences HIV-1 persistence.
Ongoing efforts to achieve a functional cure for HIV-1 are directed towards supplementing ART with immunotherapies targeting the viral reservoir.Such strategies, under the umbrella of "shock and kill", aim to reactivate the replication competent reservoir and eliminate latently infected cells by viral cytopathic effects or immune-mediated killing (10,11).However, these strategies have thus far failed to achieve a reduction of the viral reservoir or post-ART virologic control in either the clinic or preclinical models.This is mainly due to the low inducibility of latent proviruses and the heterogeneity of the mechanisms of persistence (12,13) (14,15).In contrast, therapeutic vaccination strategies focused on enhancing HIV/SIV-specific responses have had some discreet measure of success (16).However, currently, no single strategy leads simultaneously to latency reversal and stimulation of effective immune responses.
The HIV life cycle and HIV's ability to replicate efficiently are especially dependent on the activation status of the infected cells.In this context, significant advances have been made to directly activate immune cells through non-canonical pathways in order to promote HIV latency reversal (17).However, the activation and differentiation of immune cells is intrinsically linked to (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint cellular metabolism (18).Indeed, signaling pathways that govern immune cell differentiation and activation such as mTOR and Wnt/β-catenin, not only have been implicated in regulating HIV latency (19)(20)(21), but are also critical regulators of cell metabolism (22).The metabolic status of an infected cell, in turn, plays a critical role in its ability to support latency reactivation and viral production (19,23).
In this context, recent evidence suggests that TGF-β plays a critical role in the regulation of immune cell activation and metabolic reprogramming (24)(25)(26).Specifically, in the context of CD8 + T cells, TGF-β has been shown to suppress mTOR signaling preserving the metabolic fitness of memory CD8 + T cells (25) and stem-like antigen specific CD8 + T cells through its modulation of Wnt/β-catenin factor TCF1 (26,27).TCF1 and the mTOR pathway are also critical to T cell differentiation and responsible for the transition from activated effector cells to resting memory cells during LCMV infection (22,26).The regulation of quiescence in CD8 + T cells that follows continuous TGF-β stimulation is critical to the transition to a memory phenotype and it is driven by specific metabolic changes that are linked to decreased glycolytic activity, more efficient mitochondrial respiration, and long-term survival (25,26).
Similarly, in the context of NK cells, TGF-β has been implicated in decreasing their baseline metabolism driving lower expression of markers of NK cytotoxic activity (28).
While TGF-β-mediated suppression of TCR and IL-2 signaling were shown to lead to lower CD4 + T cell activation following cognate antigen recognition in older studies (24,29), more recent work using CD4 + T cell-specific deletion of the TGF-β receptor demonstrated an even more profound effect of TGF-β on all stages of CD4 + T cell activation, proliferation and cytotoxic response to LCMV than in CD8 + T cells (30).However, little is known on the specific role of TGF-β in regulating the transition to and from memory and effector phenotypes in CD4 + T cells and how this may be associated with TGF-β-driven changes in CD4 + T cells metabolism.
Moreover, TGF-β regulates the expression of CD103 and other surface and intracellular factors (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
We and others have recently demonstrated that TGF-β regulates HIV-1 latency in primary CD4 + T cells ex vivo and in vivo (35)(36)(37).Latency reversal was detected in a non-human primate (NHP) model of HIV infection following a short treatment with a clinical stage TGF-β inhibitor, galunisertib (LY2157299) (38).In that previous study, we documented latency reversal particularly at the level of the gut mucosal tissue using the 64 Cu-anti-gp120 Fab 2 (7D3) probe and immuno-PET/CT (35).In that study, we further validated the ability of immune-PET/CT to identify sites of viral reactivation and replication in the gut by performing tissue resection in hot areas of the gut identified by PET followed by confirmatory PCR for vDNA/RNA and vRNAscope (35).
Herein, we demonstrate how treatment with galunisertib with a 2-week on, 2-week off regiment that mimics the therapeutic regimen employed in the clinic in phase 1 and 2 trials of solid tumors (39,40), leads to profound transcriptional and functional changes in immune cells in the absence of overt toxicity or increased systemic inflammation.Importantly, we observed a shift toward a transitional effector phenotype in CD4 + T cells and other immune cells both systemically and in the lymph nodes.This shift was accompanied by, and likely responsible for, increased viral reactivation in SIV-infected, ART treated macaques documented by molecular techniques and PET/CT images.At the end of the treatment with galunisertib, we detected lower viral reservoir levels, including total and intact proviral DNA in both PBMC, gut and lymph nodes and significantly higher immune responses.

2-weeks on-off therapeutic regiment with Galunisertib leads to viral reactivation SIV infected, ART-treated macaques.
(which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.To confirm galunisertib-driven HIV/SIV latency reversal and investigate the underlying mechanisms, 8 Indian origin rhesus macaques (Macaca Mulatta, Mamu-A01-, -B08, -B17-, all females) were infected intravenously with 300 TCID50 of the barcoded SIVmac239M2.We  S1).ART was discontinued 3 weeks after the last galunisertib dose, and the macaques were followed for 6 weeks after ART discontinuation.Median peak plasma viral load (pVL) was 10 8 copies/ml at week 2pi.A rhesus anti-PD1 antibody was administered at 5mg/kg before the 3 rd and 4 th cycle to 2 macaques (08M156 and A6X003).However, no differences were noted for these 2 macaques in any of the parameters we measured, and the data were pooled.
Full suppression to undetectable levels (pVL LOD 15 copies/ml) was achieved in 3 out of the 8 macaques at week 10pi.In the other 5 macaques, pVL fell below 65 copies/ml by week 22pi with a single blip of 400 copies/ml in A8T010 at week 29pi (Fig 1B).Following the start of the galunisertib treatment, pVLs increased in 7 out of 8 macaques from a single peak over undetectable in 08M171 and A8L057 to several peaks and up to 10 3 copies/ml in the other macaques.Of note pVLs in A8T010 and A8R095 was undetectable for over 5 weeks before, respectively, blips of up to 10 2 copies/ml were detected following galunisertib treatment initiation (Fig 1B).More frequent blips were noted during the first 2 cycles with galunisertib compared to cycles 3 and 4 (Fig 1C).However, 08M171 and A8L057 did not experience a pVL increase until the 4 th cycle.
Importantly, we documented viral reactivation also using immunoPET/CT.The 64 Cu-anti-gp120 Fab 2 (7D3) probe was injected 24hrs before and after, respectively, the beginning and the end of the first 3 galunisertib cycles and scans performed before and after the first and the last dose.

A 4 cycle treatment with galunisertib decreases viral reservoir in absence of systemic inflammation.
To determine the impact of galunsiertib on SIV reservoir, we measured CA-vDNA in PBMC, colorectal biopsies and lymph nodes (LN).A significant decrease in CA-vDNA was detected in all tissues between week 35pi (beginning of cycle 1, BC1) and week 49pi (after/end of cycle 4, AC4) for gut and LN, and between week 35pi (BC1) and end of cycle 3 (AC3) for PBMC (Fig 3A , AC4 not measured for PBMC and Fig S4A).In the gut and LN (right axillary), decreases ranged from a Log to 1/3 of a Log (gut: median 0.77; range:0.33-0.94folds LN: median: 0.93; range:0.33-0.98 folds decrease).In the PBMC the decrease was slightly less pronounced with a (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint median half Log decrease (median: 0.58; range 0.28-0.88folds decrease).However, the comparison for the PBMCs was between week ~35 and week ~45 (end of cycle 3, AC3) instead of the end of all 4 cycles, because a snap frozen pellet was not available at the end of cycle 4 for PBMCs.The SIV reservoir in PBMCs was also monitored by SIV-IPDA comparing before cycle 1 (BC1) to the end of cycle 4 (AC4).Of note, we observed significant decreases of both total and intact provirus by SIV-IPDA.Intact provirus declined similarly to the CA-vDNA with a median of half Log (median: 0.53; range:0-0.71folds decrease).
In contrast, no decline in CA-vDNA was detected in the PBMCs of a group of 4 macaques infected intravenously with the same stock of SIVmac239M2 for a separate study.These 4 macaques were placed on ART on week 6pi as in our study.However, no decline in CA-vDNA was detected under ART, between weeks 28 and 52pi (untreated group, Fig S4B).Thus, we conclude that the decline in CA-vDNA in PBMC was not due to ART-alone.Importantly, we found no significant changes in any of the clinical variables (chemistry and hematology, Tables S2 and S3) measured before, during and after the galunisertib therapy.Moreover, we observed no changes in the concentrations of inflammatory chemokines and cytokines measured in plasma before and after the first 2 treatment cycles and at the end of the last cycle (Fig 3C).The only difference in cytokine and chemokine levels after galunisertib treatment was a small increase in IL-10 detected at the end of the last cycle compared to before treatment (Fig S4C ).

Galunisertib treatment drives an effector phenotype in T and NK cells.
The phenotype of PBMCs before and after Galunisertib treatment was monitored by highparameter flow cytometry of T and NK cell subsets and phenotype (Table S4).Classical subsets and single-color analysis of MFI revealed a substantial increase in the expression of CD95 and a profound consistent decrease in TCF1 expression in CD4 + T cells that continued throughout the treatment (Fig 4A Hence, the frequency of central memory and effector memory as defined by CD95 and CD28 or CCR7 did not change (Table S5).The levels of T-bet did not change significantly (Fig S6A).
However, we detected a downregulation of the gut homing receptor integrin α4β7 and a decrease in the levels of granzyme B (GRZB, Fig S6A).Of note, the expression of activation markers CD69, HLA-DR and Ki67 remained mostly unchanged, with the exception of an increase in HLA-DR at the beginning of cycle 2 compared to before galunisertib (Fig 4C ).
Interestingly, the effect of Galunisertib on CD8 + T cells was not as pronounced as it was on CD4 + T cells.No significant increase was detected in CD95 expression and TCF1 downregulation reached significance only at the end of the treatment (Fig 4D).Markers of cell activation like HLA-DR, Ki67 and CD69 did not change (Fig 4D and Fig S6B).However, we detected a significant decrease in T-Bet at the beginning of cycle 2 (BC2) compared to before treatment (BC1).Finally, in contrast to CD4 + T cells, the decrease in GRZB was more pronounced at several time points during treatment and there was a sustained significant increase in CCR7 expression in memory CD8 + T cells (Fig 4E).Interestingly, the expression of PD1 on both CD4 + T cells and CD8 + T cells either did not change or was slightly downregulated.

4F).
High-dimensional data visualization with tSNE and clustering analysis with FlowSOM(44) confirmed the results of the classical analysis.We performed tSNE and FlowSOM after data clean up with FlowClean and normalization with the SwiftReg algorithm (45).The analysis was performed on before (BC1) and after the last cycle (AC4) only (Fig 4G

Galunisertib treatment in vivo increases pTreg while decreasing pTfh
The frequencies of circulating Tregs and Tfh cells were monitored with an established flow cytometry panel (46,47)  The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint diphosphohydrolase 1), which tracks within extracellular adenosine and immunosuppressive effects, was lower on total and central memory CD4 + T cells (Table S5 and Fig  In our previous studies, we determined that 6hrs after galunisertib treatment in naïve macaques there was an upregulation of the AP1 complex (JUN and FOS) and several genes encoding ribosomal proteins in CD4 + T cells (42).To understand early and later effect of galunisertib in the context of SIV infection, in the current study we performed bulk RNAseq of PBMC isolated 1hr after the first administration of galunisertib in cycle 1 and at the end of the 1 st 2-weeks cycle.We confirming an increase in translation and metabolism within these cells.
(which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint Since we noticed substantial differences in the impact of galunisertib on CD4 + T cells compared to CD8 + T cells by flow, we focused the analysis on these subsets.In CD4 + T cells we found only 25 DEGs with a log 2 FC=0.15, while 34 DEGs were in CD8 + T cells with more downregulated genes in the CD8 + T cells compared to the CD4 + T cells.The AP1 complex and

Galunisertib increases SIV-specific responses and changes barcode distribution
In order to understand how galunisertib affected immune cell function and SIV-specific responses, we stimulated PBMC with 15-mer SIV peptides from SIVmac239 gag, pol and env (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint for 24hrs on antibody coated Elispot plates.Because of sample availability, we probed before and after the first cycle and after the 3 rd galunisertib cycle only.Interestingly, by the end of the 3 rd cycle there was a significant increase in IFN-γ secretion both SIV-specific (particularly against env) and non-specific (DMSO control).TNF-α release increased particularly in response to Gag and non-specifically (Fig 7).However, IL-2 release appeared to increase slightly after the first cycle (non-significant), but remained unchanged with a slight decrease by the end of the 3 rd cycle ( Fig 7).
In order to understand if these changes in immune responses in combination with latency reversal and switch toward an effector phenotype may have impacted viral population dynamics, we analyzed changes in numbers and distribution of the viral barcodes before and after the first 3 treatment cycles.There was no barcode amplification at several time points particularly in the lymph nodes.However, although there were no significant changes in the number or diversity  and S15B).Specifically, the relative proportion of each barcode changed in all monkeys in all cycles in the rectal biopsies (probably a consequence of different sampling area), but also for all LN analyzed (except for 08M134 in cycle 1).Of note, the same LN were sampled at the beginning and after cycle 3 (Table S6).(which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.

TCF1 downregulation in CD4 + T cells correlates with virological and immunological endpoints.
In order to explore a possible association between the various virological and immunological parameters and their changes, we built a correlation matrix with a curated set of variables of interest or their fold changes.This analysis revealed an association between the decrease in TCF1 expression in the CD4 + T cells and several virological and immunological variables (Fig 8E).Specifically, both the levels of CA-vDNA in the colorectal tissue and the fold increase in gut-SUV strongly inversely correlated with fold changes in TCF1.Since TCF1 decreased, a larger decrease in TCF1 was directly proportional to residual vDNA in the gut at the end of cycle 4 and to the increase in PET signal in the gut (Fig 8D and 8E).A weaker, but still significant association was also present with CA-vDNA in PBMC at the end of cycle 3 and with the levels of vRNA in the lymph nodes at the end of the 4 th cycle (Fig 8D and S16A).Finally, the decrease in TCF1 correlated with the increase in IFN-γ and poly-functional IFN-γ/TNF-α releasing cells (cumulative increase of SIV-specific responses to gag, pol and env; Fig 8D and 8E).Of note, the increase in PET signal in the gut also correlated with the levels of vDNA in PBMC at the end of

Discussion
(which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint HIV-1 latency in T cells is maintained through diverse mechanisms that include blocks in transcriptional elongation, completion, and splicing (13).A common characteristic shared by HIV-1 latently infected cells of both T and myeloid cell lineages is their "resting" phenotype (14,(52)(53)(54) In these cells, an inability to transcribe proviral DNA is linked to a generalized decrease in transcriptional activity which, in turn, is linked to their metabolic status (23).Cellular metabolism is in turn influenced by tissue location and environmental cues (55).TGF-β is released at high level in PLWH and its levels remain high during ART (56).The immunosuppressive activity of TGF-β is well-known.However, the effect of TGF-β signaling in immune cells is highly context dependent (57).Hence, TGF-β plays different roles according to a cell differentiation and activation status (57).In CD8 + T cells and NK cells, TGF-β was shown to decrease mTOR activity and preserve cellular metabolism (high mitochondrial activity and spare respiratory capacity, but reduced mTOR activity) preventing metabolic exhaustion (25,26,58,59).This effect was linked to survival of antigen-specific CD8 + T cells, preservation of their stemness and it was linked to higher expression of the TCF1 factor (25).
In contrast, in CD4 + T cells TGF-β is known to decrease TCR activation (29,60), restrict proliferation and inhibit cytotoxicity (including granzyme and perforin release) at different stage of infection in vivo (30,61).However, the role of TGF-β in the formation and preservation of CD4 + T cell memory is still unclear (57).Moreover, the link between TGF-β signaling and TCF1 expression in CD4 + T cells is unexplored.
Here, we used a clinical stage small drug, galunisertib, developed by Eli Lilly and used in several phase 1 and 1/2 clinical studies against solid cancer (39,62,63) to investigate the impact of TGF-β blockade on SIV latency, SIV reservoir and immune responses.Of note, Eli Lilly did not terminate galunisertib development program because of toxicity (64,65).Indeed, in our studies in macaques, we observed no adverse events nor changes in chemical or hematological variables.Moreover, there were no detectable changes in the levels of the 24 (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint inflammatory factors that we probed in plasma during the treatment.This suggests that this therapeutic approach may be safe in people living with HIV (PLWH).
The first important finding of our study was the confirmation of our previous report of the latency reversal properties of TGF-β blockade in vivo (35).Indeed, we found increase in pVL in 7 out of the 8 macaques upon initiation of galunisertib therapy.Although not all macaques were fully suppressed before treatment, substantial increases in pVL (>10 2 copies/ml) were noted also in fully suppressed macaques (08M171, A8R095 and A8L057).Moreover, extensive viral reactivation was documented in tissues by immunoPET/CT.The increase detected postgalunisertib in gut SUV correlated with CA-vRNA as in our previous studies (35) and, interestingly, it was associated with a decrease in TCF1.Galunisertib led to increased probe uptake in the gut and lymphoid tissues, as well as a generalized increase in whole-body uptake.
Yet, there was no change in the PET background signal in kidneys and spleen.This, in conjunction with our previous studies demonstrating the specificity of the PET signal for areas of enhanced SIV replication in gut and lymph nodes (35,41,66), demonstrates that the galunisertib-driven increases in SUV were mostly specific, and highlighted areas of SIV latency reversal at least in gut and lymphoid tissues.The extent to which the increased signal in the spine and bones recapitulates an increase in SIV replication at these sites remains to be determined.
Importantly, we observed a decrease in CA-vDNA in all the tissues that cannot be attributed to ART alone.Indeed, although we did not have a concurrent control group, this decrease was not present in similar studies conducted in SIVmac239M2 infected macaques on the same ART regimen, but not treated with galunisertib.Considering studies by other groups with different models (SIVmac251), they report 2 nd phase decay of SIV intact provirus (weeks ~32 to ~100pi, Fig 2B in (67)) with a t 1/2 of >8 months (67).In contrast, in our study, the intact provirus decreased by 3 folds (median) in a little over 3 months (from week 35pi, BC1 to week 49pi, AC4).
Interestingly, this decrease in intact pro-virus trended toward a direct correlation with IFN-γ (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint responses.However, IFN−γ levels also inversely correlated with the absolute value of residual intact provirus, suggesting that IFN-γ responses were driven by residual viral reservoir while, at the same time, were involved in clearing intact virus.Indeed, the increase in IFN-γ and TNF-α also correlated with residual CA-vRNA in the LN at the end of the treatment.The latter, in turn, was directly proportional to the increase in gut SUV.This is in line with increased latency reversal explaining residual viral RNA in lymphoid tissues.
Finally, one of our most intriguing results was the profound downregulation of TCF1 in CD4 + T cells at both the transcriptional and protein levels.Although TCF1 is conventionally viewed as an effector of the canonical Wnt pathway (68) and recently reached notoriety for its role in maintaining stemness of antigen-specific memory CD8 + T cells (26,69), TCF1 has a plethora of functions in T cell development and differentiation largely independent of Wnt signaling (22).In CD4 + T cells, TCF1 has been implicated in orchestrating all the major Th subsets, including Th1, Th2, Th17 and Tfh(70).TCF1 is known to control the bifurcation between Th1 and Tfh in favor of Tfh cells(70), while it negatively regulates Treg development (22).This is in line with our findings of increased Treg and decreased Tfh, in the midst of a profound downregulation of TCF1.Importantly, TCF1 is downregulated with increased cellular differentiation and progression toward effector functions in T cells.T cell activation leads to reduced levels of TCF1 and higher levels of TCF1 are present in T cells with higher stemness and low anabolic metabolism (68).These findings suggest that TCF1 has a critical role in maintaining quiescence in immune cells likely in concert with TGF-β (18,61).Our data reveal that this link may be even more prominent and important in CD4 + T cells than in CD8 + T cells.Importantly, the decrease in TCF1 was accompanied with enhancement in other effector markers such as CD95, CD16 and GRZB (at the transcriptional level) and an increase in the transcription of AP1 complex.
However, there was no clear upregulation of other classical markers of immune activation such as CD69 and no increase in T cell proliferation (Ki67 expression).Hence, galunisertib treatment (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint does not appear to lead to classical T cell activation.Instead, in vivo TGF-β blockade seems to primarily change the metabolic state of T cells (and likely other immune cells) increasing OXPHOS and mitochondrial function.Of note, although glycolysis is essential during cell activation, mitochondrial pathways are engaged and remodeled early after activation and OXPHOS upregulation has a pivotal role in the earliest stages of cell activation (18).
Hence, we propose a model in which TGF-β inhibition forces cells (particularly CD4 + T cells) out of quiescence to a transitional state where they reinitiate their transcriptional program and are metabolically ready to be activated and effector.In absence of direct TCR engagement or other activation stimuli, they do not undergo full/classical activation and proliferation, yet they can reinitiate viral transcription and respond to SIV antigenic stimulation.Indeed, functionally, we demonstrated that the PBMC after galunisertib treatment secrete higher levels of IFN-γ and TNF-α.Interestingly, there was no increase in IL-2 secretion.The link between TGF-β and IL-2(71) and the critical role of IL-2 in T cell proliferation again suggest that galunisertib enhances an effector phenotype uncoupled from cellular proliferation.scRNAseq analysis demonstrated a trend toward an effector phenotype also in other immune cells, such as B cells, NKs and macrophages.Future studies will need to uncover in depth the effect of TGF-β blockade on these other immune subsets.
This study has several limitations.The most important limitations are the relatively small number of macaques and the lack of a concurrent untreated control group.We also could not investigate in depth the viral kinetics after ART interruption because of the short follow up after ATI.Moreover, we did not explore changes in the phenotype or turnover of cells isolated from gut and lymph nodes and relied solely on transcriptional data for these tissues.Finally, because of sample availability, we could not dissect the cellular origin of increased IFN-γ and TNF-α.
In conclusion, we report that in vivo treatment with a clinical stage small molecule TGF-β inhibitor drives an effector phenotype in T cells that is likely responsible for increasing the (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint frequency of spontaneous latency reversal events, stimulating SIV-specific immune responses, and decreasing the viral reservoir.Future work will determine whether the galunisertib-driven enhanced antiviral responses and decreased viral reservoirs can significantly contribute to post-ART virological control.

Study design and Ethics Statement
A total of 8 adult female Indian origin Rhesus macaques (Macaca mulatta; Mamu A*01, B*08 and B*17 negative) were used for the study described in the manuscript (Table S1).All the macaques were selected form the colonies bred and raised at the New Iberia Research Center (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
Blood viral load was monitored biweekly before and during ART and every 3-4 days during Galunisertib treatment.Colorectal biopsies and LN FNA were collected before and after galunisertib treatment.ART was terminated 3 weeks after the last galunisertib dose and euthanasia and necropsy to harvest tissues were performed at week 58 post infection, tissues samples were flash frozen, fixed in OCT or Z-fix.

Plasma and Tissue SIV Viral loads (VL)
Blood was collected in EDTA tubes and plasma was separated by density gradient centrifugation and used for the determination of plasma VL by SIVgag qRT-PCR at NIRC or at Leidos (Quantitative Molecular Diagnostics Core, AIDS and Cancer Virus Program Frederick National Laboratory).Tissue VL from snap frozen PBMC pellets, colorectal biopsies and LN FNA were performed as described in (72).SIV-TILDA was performed on freshly stored PBMC before cycle 1 and at the end of cycle 4 of galunisertib by Accelevir, Baltimore, MD.

ImmunoPET/CT
ImmunoPET/CT for mapping SIV signals in total body scans were conducted in part as reported previously (42).The probe consisted of primatized p7D3 anti-env F(ab)'2 coupled with the chelator DOTA and labeled with Cu 64 just prior to administration to the animals.For probe administration, the animals were sedated, and a venous catheter was placed into an arm of leg vein to minimize bleeding of the probe into the tissue surrounding the site of injection.The probe for each animal consisted of ~1mg of the p7D3 F(ab)'2 labeled with 2-3 mCi of 64 Cu.After injection, the animals were allowed to recuperate in their cage until the next day.At 24 hours, the animals were again anesthetized with Telazol and the macaque's body was immobilized in (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint dorsal recumbency on the scanner table.Scans were conducted in a Phillips Gemini TF64 scanner.The final CT image was compiled from 200 to 300 slices, depending on macaque size.
PET Image analysis was performed using the MIM software.PET/CT fusions were generated scaled according to calculated Standardized Uptake Values (SUV).The SUV scale for the PET scans was selected based on the overall signal intensity of the PET scans (whole body), and the CT scale was selected for optimal visibility of the tissues.All images and maximum image projections were set to the same scale for visual comparisons.Additional details on MIM analysis are described in Supplemental Methods.

Cell isolation, flow cytometry staining, classical and high-dimensional analysis
Colorectal biopsy tissues were isolated by enzymatic digestion while LN biopsies were passed through a 70μm cell strainer as described in (42).Isolated cells were phenotyped with panels listed in Supplemental Table S4.FlowJo vs 10.8 was use for both classical and high-dimesional analysis.PeacoQC, FlowtSNE and FlowSOM plug-ins were used with default settings.SwiftReg was used for normalization (45).More details on the staining procedures and analysis pipeline in Supplemental Methods.

Bulk and scRNAseq analysis
For bulk RNAseq, snap frozen PBMC pellets from BC1, 1hr after first Gal dose and AC1 were used for RNA extraction with the RNeasy kit with on column DNA digestion (Qiagen).Library preparation was performed using TruSeq Stranded Total RNA with Ribo-Zero Globin and sequencing was done with an Illumina HiSeq4000 with >20M reads/samples.Sequencing data was demultiplexed and trimmed using Trimmomatic v0.36 to remove adapters and low-quality reads.Trimmed reads were aligned to the Mmul10 reference genome and transcripts quantified using the Hisat2-StringTie pipeline (73).Differential gene expression analysis using the quantified gene transcripts was performed with DESeq2 R package (74) comparing the samples (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
Differentially expressed genes (DEGs) were analyzed by functional enrichment analysis and gene set enrichment analysis (GSEA) to identify specific pathways and molecular processes altered by galunisertib.
The Parse pipeline and Partek software were.For bulk RNAseq features with <100 counts were removed, data normalized and DESeq2 was used to obtain a DEG list.Genes with a false discovery rate (FDR)-adjusted p-value ≤0.05 and absolute log 2 fold-change (FC) (compared to BC1) above 2 were defined as significantly differentially expressed (DEG).For scRNAseq analysis cells isolated from lymph nodes before (BC1) and after (AC1) galuniertib were fixes with the Parse fixation kit, barcoded and sequenced at the NUseq Core.The Partek software was used for scRNAseq analysis.Cells with 400-8000 features, excluding features with 0 reads in >99.99 cells, were included.Scran deconvolution was used for normalization and cell classification was based on gene expression.Hurdle models were used to compare DEGs in each cell subset before and after galunisertib.See supplemental methods for a detailed description of scRNAseq analysis and additional control analysis using Seurat R package with more stringent QC cut-offs and SCTransform normalization.

Plasma cytokines and T cell responses
Cytokines in plasma (at 1:2 dilution) were measured using the NHP Cytokine 24-Plex kit by Meso Scale Diagnostics (MSD) according to manufacturer instructions.Frozen PBMC collected at week 6 post-infection (pre-ART), right before the first galunisertib administration (BC1), at the end of cycle 1 (AC1) and at the end of cycle 3 (AC3) were thawed in AIM V medium (Thermo Fisher) with benzonase (Sigma) and plated on a FluoroSpot (CTL) plate pre-activated with 70% Ethanol and IFN-γ, TNF-α and IL-2 capture solution.Gag, pol, and env 15-mer peptides (NIH AIDS Reagents program) were prepared at two times the final concentration of 2.5μg/ml with co-stimulatory reagents anti-CD28 10μg/ml and anti-CD49d 10μg/ml and added to the cells in (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint CTL-Test TM Medium.Parallel positive control of PMA (20ng/mL) and ionomycin (200ng/mL) or mock DMSO solution was also plated with the stimulatory reagents.PBMCs were added at 300,000 cells per well.After 24hrs, the plate was washed, and incubated with detection and tertiary solutions and shipped to CTL for scanning and QC.

Statistics
GraphPad Prism v10, R and Python were used for statistical analysis and data visualization.
Wilcoxon matched-pairs test and repeated measures ANOVA or mixed effect analysis (when the data set had missing data) were used to compare the different virological and immunological variables between baseline (BC1) and a single or multiple post-galunisertib time points.In Prism the mixed model uses a compound symmetry covariance matrix, and is fit using Restricted Maximum Likelihood (REML).Cytokine data from MSD assay were first Log transformed, then normalized by subcolumn/factor in percentage with 0% as smallest value and 100% larger value in each dataset.Each factor was analyzed separately with ANOVA for repeated measures and Dunn's multiple comparison's post-hoc test and together by principal component analysis (PCA).Holm-Sidak test was used for multiple comparison correction in all cases, but FlowSOM cluster comparison where the FDR method was used.Pearson r coefficient was calculated for pairwise tests of association in a correlation matrix with selected variables.For RNAseq analysis see above and supplemental methods.For viral population analysis, Shannon Entropy of the viral barcodes present in each sample was used to measure the diversity of the viral populations (see supplemental Methods).Chi-squared tests were used in R to compare barcode composition between before and after cycle time points within a tissue/macaque using paired barcode relative frequencies.Unless otherwise specified p-value<0.05was considered statistically significant.
(which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.

Fig 2 ,
Fig S1 and Movies S1-S8, the PET signal visibly increased in different areas (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint after cycle 2 (in A8R095, 08M156, A8L014 and A8T010) or at the beginning of cycle 3 (in A6X003, 08M134 and A8L057).In 08M171 we observed an increase in the gut area only at the beginning of cycle 2.A corresponding increase in standard uptake values (SUV) was detected in the gastrointestinal area and axillary lymph nodes (Fig 2B) and was significant in cycle 3 compared to before cycle 1 (BC1).In these anatomical areas, SUV increases correspond to increases in viral replication as shown in previous studies(41, 42).A PET signal increase was also noted in the area of the vertebral column (spine).Spine SUV already increased significantly with the first cycle (Fig 2B).However, neither cerebrospinal fluid (CSF) nor bone marrow (BM) was collected during the study, and we have no prior validation of the specificity of the signal in the spine.Hence, whether this signal corresponds to increased viral replication and whether this occurs in the vertebral bones or cerebrospinal fluids remains to be determined.No SUV increase was present at the level of the spleen or kidney, where probe accumulation and background signal likely masked any specific signal.However, a significant increase in SUV was detected in the liver (Fig 2B) and whole-body (Fig S2).Of note, the SUV increase in the gut area in most cases followed an increase in cell-associated vRNA detected in colorectal biopsies at the same time points during treatment (Fig S3).

Fig
Fig S7A).The remaining populations were likely monocytes and other minor subsets.Direct after the first and third cycle of galunisertib.The frequency of all CD4 + Tregs (CD25 high FoxP3 + ; gating in Fig S8) and CD8 + Tregs increased after the first cycle and remained higher until the end of the 3 rd cycle, while CCR4 + Treg were proportionally higher at the end of the treatment compared to before (Fig 4J).In contrast, circulating Tfh (CXCR5 + PD1 + ; gating in Fig S8) were lower at the end of the Galunisertib treatment both within total and central memory CD4 + T cells (Fig 4J).Finally, the expression of CD39 (ecto-nucleotide triphosphate (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
4J, respectively).Bulk RNAseq of PBMC and single-cell (sc)RNAseq analysis of LN confirm a profound shift toward effector phenotype.
found 640 genes significantly modulated (FDR q<0.05; log 2 FC=2) in PBMC just 1hr after the first dose of galunisertib.The majority (457 genes) were downregulated (Fig 5A).Gene set enrichment analysis (GSEA) revealed an upregulation of the oxidative phosphorylation (OXPHOS; ES 0.58 FDR q=0.048)) and the reactive oxygen (ES 0.55 FDR q=0.075) pathways among the Hallmark genet sets (Fig 5A and S9).Among the Biocarta sets, there was an enrichment in the electron transport chain (ETC; ES 0.84 FDR q=0.132) and a downregulation of the circadian (ES -0.81 FDR q=0.163) pathways (Fig 5A and S9).Finally, among the top modulated genes there were several genes encoding for soluble transporters, while classical activation markers like CD69 and CD38 were downregulated (Fig S9B).An early engagement of metabolic pathways was confirmed by enrichment analysis of significant DEG with Metascape(48) with an upregulation of adipogenesis, OXPHOS and fatty acid metabolism (Fig S9C).Of note, 2 weeks after the beginning of galunisertib, metabolic pathways were still among the most enriched upregulated pathways in PBMCs (Fig 5B and Fig S10A).Among the top upregulated genes there were CD44, CCR5, several integrins and GRZA and GRZB (Fig S10B).Finally, we performed bulk RNAseq of rectal biopsy tissue before and after the first cycle with (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.The copyright holder for this preprint this version posted September 8, 2023.; https://doi.org/10.1101/2023.09.05.556422 doi: bioRxiv preprint galunisertib (Fig 5C).There were only 51 differentially expressed genes (DEGs; Adjp<0.05;log 2 FC=2).GSEA analysis of all DEGs (Adjp<0.05)revealed the G2_M_DNA replication pathway highly enriched (ES 0.97 FDR q=0.012) within the Hallmark set.Among the most interesting changes, we observed a pronounced downregulation of integrin β7, in contrast to an increase in integrin αE (CD103; Fig 6C), suggesting that TGF-β-driven increase in αE(49) may be driven by non-canonical TGF-β pathway signaling not blocked by galunisertib.To clarify the impact of galunisertib at the single cell level and in lymphoid tissues, we also performed scRNAseq analysis of cells isolated from LNs before (right axillar) and after (right or left inguinal) the first cycle (Fig 6).Dimensionality reduction and clustering analysis with PCA and uniform manifold approximation and projection (UMAP)(50) was performed to visualize the data (Fig 6A).However, cells were classified based only on gene expression (Fig 6B and C).We first classified major subsets: T cells (38,705 cells), B cells (31,385 cells), NK cells (1,628 cells), macrophages (285) and cells expressing both CD19 and CD3 (942 T/B cells) (Fig 6A shows this classification over UMAP).Then we classified only CD4 + and CD8 + T cell subsets (Fig 6C and S11A) and B/T subsets including naïve and germinal center (GC) B cells and Tfh cells (Fig S11B and S11C).Cell number for all these subsets did not change with treatment (Fig S12).However, differential gene expression analysis of T cells revealed an upregulation of members of the AP1 complex, CD69, β2 macroglobulin and RPL13 among the most upregulated genes (Fig 6D).Moreover, it confirmed a downregulation of TCF1 at the transcriptional level (TCF7 gene; Fig 6D).Gene enrichment analysis revealed again Myc_targets_V1, OXPHOS and mTORC1 as the most enriched hallmark pathways (Fig 6E).Of note, among gene ontology cellular processes, RNA processing was the most highly enriched pathway with intracellular transport and catabolic processes following right after (Fig S13A) TCF7 were again upregulated and downregulated respectively in both CD4 + (Fig 6F) and CD8 + T cells (Fig 6G).However, STAT1 was more strongly downregulated in CD4 + T cells.Enrichment analysis showed once again upregulation of Myc_targets_V1, OXPHOS and mTORC1 pathways in both CD4 + and CD8 + T cells (Fig 6H and Fig S13B).Interestingly, the most enriched set among in Biocarta was the HIV-Nef pathway (Fig 6J) demonstrating the relevance of these galunisertib-driven changes to HIV cell cycle and transcription (highly enriched KREB pathway as well).The 2 nd most enriched Biocarta pathway was TCR signaling, linking galunisertib to cell activation.Next, we analyzed changes in gene expression in Tfh cells.In this subset we obtained a similar number of DEG as in other T cells and myc_targets_V1 was still the most enriched hallmark pathway (Fig S13C and D).In B cells, the AP1 complex was again prominently upregulated, together with CD83, CD69 and MAMU-DR.Of note, more genes were modulated in B cells (44 genes with a log 2 FC=0.15 and 18 with logFC0.2) than in T cells with some differences in enriched pathways (Fig S13F).Similar genes were modulated in GC B cells and naïve B cells with several more DEGs in naïve B cells than in GC cells (Fig S14A and B).

Finally, 144
genes were modulated in NK cells and 39 genes in macrophages (FigS14C and D; log 2 FC=0.15).GRZB was prominently upregulated, but CD44 downregulated.In macrophages genes were mostly downregulated including TCF7L2 and KLF4 suggesting an increase in inflammatory phenotype and decrease in M2 polarization(51) (FigS14D).

(
measured as Shannon Entropy; Sh) of barcodes before and after each galunisertib cycle (Fig S15A and Fig 8A), we found significant changes in barcode frequency distribution in most tissues and cycles after galunisertib treatment (Fig 8B Hence, sampling location does not explain the changes barcode distribution.Changes in the proportion of barcodes were also detected in at least half of the PBMCs after cycle 1 and 2 (cycle 3 not analyzed).Finally, barcode diversity decreased in plasma post-ATI compared to the time point right before ART (Fig 8C) and barcode distribution significantly changed post-ATI compared to pre-ART in 2 of the 7 macaques that rebounded (Fig S15C).

cycle 3 ,
vDNA in gut biopsies and residual CA-vRNA in LN at the end of cycle 4 (Fig S16B).Interestingly, the levels of residual CA-vRNA in LN also correlated with the increase in IFN-γ and poly-functional IFN-γ/TNF-α releasing cells (Pearson r=0.85 and 0.76, respectively).Finally, the residual intact pro-virus (IPDA) directly correlated with the levels of IFN-γ and IFN-γ/TNF-α produced in response to SIV peptides at the end of cycle 3 (Fig 8D; r= 0.76 and 0.88, respectively).In contrast, the change in intact provirus trented to correlate inversely with the levels of IFN-γ, so that a larger decrease directly correlated with more IFN−γ responses.However, this did not reach significance (p=0.117,Fig S16C).

(
NIRC), University of Louisiana at Lafayette.Treatments were in accordance with NIRC standards which comply with the Guide for the Use and Care of Laboratory Animals.Any abnormal observations and/or signs of illness or distress were reported per NIRC SOPs.Appropriate procedures were performed to ensure that potential distress, pain, discomfort were limited to that unavoidable in the conduct of the research plan.Ketamine (10mg/kg) and/or telazol (4mg/kg) sedation were used for sample administration, collection of samples and conducting image collections.Analgesics were used when determined appropriate by veterinary medical staff.Monkeys were fed Teklad monkey chow supplemented with fresh fruit or vegetables and water ad libitum.Rhesus macaques (n=8 main study + 4 separate non-concomitant study) were infected with 300 TCID 50 of the barcoded SIVmac239M2 stock intravenously and ART (Tenofovir [PMPA], Emtricitabine [FTC] and Dolutegravir [DTG]) was initiated on week 6 pi.Galunisertib treatment was initiated on week 35 p.i. Powder (MedChemExpress -MCE, NJ, USA) was dissolved in water and given orally in a treat twice daily at 20mg/kg. 4 cycles of 2 weeks daily treatment with 2 weeks wash out period were performed.Macaques 08M156 and A6X003 were given the rhesus recombinant antibody (rhesus/human chimeric) anti-PD1 antibody [NIVOR4LALA;

Figure 4 .Figure 5 .
Figure 4. Galunisertib leads toward effector in T and NK cells, increasing Treg and

(Figure 6 .
Figure 6.scRNAseq of lymph node before and after cycle 1 confirms a switch toward

Figure 7 .Figure 8 .
Figure 7. Galunisertib increases SIV-specific responses.Average spots (from triplicates) per 10 6 PBMC at the time of ART initiation (pre-ART), before cycle 1 (BC1), after cycle 1 (AC1) and ).In contrast, a small decrease in CD62L after the first cycle, reverted to (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
and Fig S7A)and on all (which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.
(which was not certified by peer review) is the author/funder.All rights reserved.No reuse allowed without permission.