PD-1-induced T cell exhaustion is controlled by a Drp1-dependent mechanism

PD-1 signalling downregulates the T cell response, promoting an exhausted state in tumorinfiltrating T cells, through mostly unveiled molecular mechanisms. Drp1-dependent mitochondrial fission plays a crucial role to sustain T cell motility, proliferation, survival and glycolytic engagement and, interestingly, such processes are exactly those inhibited by PD-1 in tumorinfiltrating T cells. Here we show that pD1positive CD8+ T cells infiltrating MC38-derived murine tumor mass show downregulated Drp1 activity and more fused mitochondria compared to PD1negative counterparts. Also, PD1positive lymphocytic elements infiltrating human colon cancer almost never express active Drp1. Mechanistically, PD-1 signaling directly prevents mitochondria fragmentation following T cell stimulation by downregulating Drp1 phosphorylation on Ser616, via regulation of the ERK1/2 pathway and, to a lesser extent, of mTOR. In addition, downregulation of Drp1 activity in tumor-infiltrating PD1positive CD8+ T cells seems to be a mechanism exploited by PD-1 signaling to reduce motility and proliferation of these cells. Overall, our data indicate that the modulation of Drp1 activity in tumor-infiltrating T cells may become a valuable target to ameliorate the anti-cancer immune response in future immunotherapy approaches.


INTRODUCTION
Programmed Cell Death-1 (PD-1) is a T cell surface receptor that down-regulates T cell activation and the immune response (1). PD-1 signaling is activated by PD-1 interaction with its ligands PD-L1 and PD-L2, expressed on adjacent cells (2), and it dampens signals originating from T Cell Receptor (TCR) and CD28, such as the activation of mammalian-Target-of-Rapamycin (mTOR) and Mitogen-Activated-Protein-Kinases (MAPK) pathways (3,4). Besides being frequently observed in T cells during chronic infections (5), activation of PD-1 signaling has also been widely reported in tumor-infiltrating T cells, contributing to their functional exhaustion and poor anti-tumor response (6). Consistently, antibodies targeting the interaction between PD-1 and its ligands efficiently reinvigorate tumor-infiltrating T cells, thereby ameliorating anti-tumor response (7,8). On the other side, impaired PD-1 function plays an important role in a variety of autoimmune diseases (9).
Mitochondria are central modulators of cellular bioenergetics, and their dynamic morphology (i.e. a fragmented or fused network) is tightly linked to cell functions (10). Among the mitochondria-shaping proteins, Mitofusin-1 and -2 (Mfn1 and Mfn2) and Optic-Athropy-1 (Opa1) are the main regulators of mitochondria fusion (11,12), while Dynamin-Related Protein-1 (Drp1) is the main pro-fission protein and it is recruited from the cytosol to mitochondria thanks to several receptors, such as Mff and Fis1 (13,14). These mitochondria-shaping proteins are regulated at both transcriptional and post-translational levels (10).
Interestingly, mitochondria morphology is tightly linked to an optimal T cell functionality. Particularly, Drp1-dependent mitochondria fragmentation sustains T cell motility and proliferation (15), and effector T (Teff) cell apoptosis following TCR engagement (16). Also, Drp1-dependent mitochondria relocation at the immunological synapse controls the influx of calcium upon T cell activation (17), sustaining the cMyc-dependent upregulation of glycolytic enzymes (15,18), thus allowing the metabolic reprogramming required to cope with the increased bioenergetic demand of an activated T cell (19,20). All these processes contribute to an optimal anti-tumor T cell response, which is indeed defective in T cells lacking Drp1 (15).
Of note, most of those Drp1-dependent processes are the same that appear to be downregulated by PD-1 inhibitory signaling, especially when considering tumor-infiltrating T cells. Indeed, PD-1 signaling reduces T cell proliferation and motility (both of them requiring Drp1) (4,15,21), and it also promotes a shift from a glycolysis-based metabolism (supported by Drp1) toward an OXPHOS-based metabolism (requiring mitochondrial fusion) (22). Given this striking inverse correlation, we asked whether PD-1 signaling may modulate Drp1 activity, and to what extent this modulation may down-regulate several processes in T cells. This point is of extreme importance, since the molecular mechanisms by which PD-1 regulates the aforementioned processes in T cells are not yet completely understood.
We here show that tumor-derived PD-1 positive CD8+ T cells exhibit a significant downregulation of Drp1 activity and a more fused mitochondrial network. Mechanistically, PD-1 signaling prevents Drp1 activation following T cell stimulation by regulating its phosphorylation on Ser616 through the modulation of Extracellular-Regulated Kinase 1/2 (ERK1/2) proteins. Also, we provide evidence that Drp1 down-regulation contributes to the reduced proliferation and motility of PD-1 positive tumor-infiltrating T cells and, as a consequence, identify Drp1 as a possible target for future therapeutic approaches aiming at restoring anti-tumor response in PD1 positive exhausted CD8+ T cells. 3

PD1 positive CD8+ T cells from MC38-derived murine tumors show a reduced mitochondrial fission.
To investigate if PD1 signaling modulated the morphology of the mitochondrial network, we decided to look at PD1 negative (PD1 neg ) and PD1 positive (PD1 pos ) CD8+ T cells infiltrating a 18 daysold solid tumor mass derived from s.c. inoculation of MC38 cells (murine adenocarcinoma) in c57BL/6 WT mice. We decided to take advantage of this tumor model, since we previously reported it as being characterized by a high level of T cell infiltration (15). By comparing expression levels of different mitochondria-shaping proteins in infiltrating CD8+ T cells (TILs) by intracellular flow cytometry, we observed that PD1 pos CD8+ TILs show a specific down-regulation of Drp1 phosphorylation on its activating residue Ser616 (Fig. 1A) compared to PD1 neg CD8+ TILs, while total Drp1 levels and the inhibitory phosphorylation on Ser637 did not vary (Fig. 1A). Also, we did not observe any differences in the expression of other main pro-fusion (Mfn1, Mfn2, Opa1) or pro-fission (Fis1, Mff) proteins (Fig. 1A).
Consistently, such a reduction in active Drp1 levels in PD1 pos CD8+ TILs correlates with an altered morphology of mitochondria. In fact, we found that sorted PD1 pos CD8+ TILs show a tendency toward a more fused mitochondrial network when compared to PD1 neg CD8+ TILs ( Fig.  1B and Supplemental Figure S1). Interestingly, this is observed even if sorted CD8+ TILs were stimulated for 2h with beads coated with anti-CD3/28 Abs plus PD-L1 (PD-1 ligand) before fixation (Fig. 1C), suggesting that such a fusion-prone mitochondria morphology may also influence T cell activation upon antigen encounter in a PD-L1-rich environment (such as a tumor mass).
Last, we extended these observations to a corresponding human tumor context, by staining sections of human colon carcinoma moderately differentiated (G2) cases with anti-PD-1 and anti-Drp1-pSer616 antibodies. Of note, we found that tumor-infiltrating lymphocyte elements almost never co-express PD-1 and active Drp1 (indeed, only less than 5% of cases were double positive) (Fig. 1D), with this suggesting that also in a human tumor context PD1 pos T cells downregulate Drp1 activity.
In sum, tumor-infiltrating PD1 pos CD8+ T cells show a tendency toward a more fused mitochondria morphology, associated with a reduced activation of Drp1.

PD1 signaling prevents Drp1 activation and mitochondria fragmentation in both murine and human T cells, upon in vitro stimulation.
To investigate whether such altered Drp1 expression in PD1 pos CD8+ TILs is directly caused by PD-1 activation, we switched to an in vitro system to specifically modulate PD-1 signaling. To this aim, we stimulated in vitro T cells isolated from spleen of WT mice with beads coated with anti-CD3/28 plus BSA (thereafter aCD3/28-beads) or anti-CD3/28 Abs plus PD-L1 (thereafter aCD3/28-PDL1-beads) for 48h. As expected, concomitant activation of PD-L1/PD-1 axis during T cell stimulation dampens activation of both mTOR and ERK1/2 pathways (as assessed by monitoring the corresponding activating phosphorylated residues) ( Fig. 2A) (3,4). Interestingly, we also observed that Drp1 phosphorylation on Ser616 (activating residue) is strongly reduced by the engagement of PD-1 signaling ( Fig. 2A), while no significant differences were observed for other mitochondria-shaping proteins (Fig. 2B). In line with this, while T cells fragment mitochondria upon activation (15,16,23), T cells stimulated in the presence of PD-L1 retain a more fused 4 network (Fig. 2C). Of note, mitochondria of murine activated T cells might characteristically appear slightly swollen when fragmented (Fig. 2C), still being fully functional (Supplemental Figure 2A-B).
In addition, such an absence of fragmented mitochondria in PD-L1-stimulated T cells is not due to ongoing mitophagy (which could hypothetically promote the clearance of the small and fragmented mitochondria). Indeed, we did not observe any significant sign of mitophagy (by checking MnSOD levels) in T cells stimulated with either aCD3/28-or aCD3/28-PDL1-beads (Supplemental Figure S2C), although autophagy is normally active.
Last, a PD-1-dependent Drp1 modulation was also observed in human T cells isolated from peripheral blood (hPBT). Similar to their murine counterpart, hPBT cells stimulated for 48h with aCD3/28-PDL1-beads do not show Drp1 phosphorylation on Ser616 (Fig. 2D) and do not fragment their mitochondria upon activation (Fig. 2E) at the same level as hPBT cells stimulated with aCD3/28-beads, which also display higher mTOR and ERK1/2 activation (Fig. 2D).
These data indicate that co-stimulation of PD-1 signaling during T cell activation prevents Drp1 phosphorylation and mitochondria fragmentation both in mice and human.

Both naïve and memory murine CD8+ T cells show reduced Drp1-dependent mitochondria fragmentation upon engagement of PD1 signaling during activation.
Next, we decided to look more closely at naïve (Tnv) and memory (Tmem) CD8+ T cell subpopulations, which, once activated, may undergo exhaustion within the tumor microenvironment. Of note, although naïve CD8+ T cells do not express PD-1 on their cell surface, they rapidly acquire PD-1 expression after 12h stimulation with aCD3/28-beads (as also observed for naïve CD4+ T cells), and the concomitant presence of PD-L1 ligand does not affect such upregulation (Supplemental Figure S3A). We thus isolated both subpopulations from the spleen of WT mice and stimulated them in vitro with aCD3/28-or aCD3/28-PDL1-beads for 48h, which is the optimal time point to detect Drp1 phosphorylation, especially in naïve CD8+ T cells (Supplemental Figure S3B). Interestingly, we found that engagement of PD-1 signaling during activation dampens Drp1 phosphorylation and mitochondria fragmentation in both CD8+ Tnv and Tmem cells (Fig. 3A-B), in parallel with a reduced activation of ERK and mTOR pathways (Fig.  3A).
Activation of PD-1 signaling is one of the hallmarks of exhausted T cells. Our in vitro observation that the engagement of PD-1 signaling prevents Drp1 phosphorylation and mitochondria fragmentation (see Figs. 2-3) is consistent with our data in vivo, indicating that PD1 pos (exhausted) CD8+ TILs show reduced Drp1 activity and more fused mitochondria (see Fig. 1). In vivo, PD1 pos (exhausted) CD8+ TILs may also arise through repetitive cycles of TCR stimulation. Therefore, to investigate whether Drp1 downregulation correlates with PD1 upregulation also in this case, we took advantage of an alternative in vitro system to induce an exhausted-like state in CD8+ T cells. Indeed, subsequent cycles of in vitro stimulation of naïve CD8+ T cells (up to 4 stimulation for 24h, each one followed by 6 days in IL2-containing medium) induce a functional exhausted-like state characterized by diminished cytokine production, proliferative capacity, and cytotoxic potential (24). Of note, we previously reported that such protocol in vitro induces a robust and progressive PD-1 upregulation in CD8+ T cells (increasing its expression further at each stimulation), and eventually mimicking the PD1 pos exhausted-like state of tumor-engaged T cells (15). Interestingly, we observed that upon subsequent cycles of naïve CD8+ T cell stimulation, the level of Drp1 phosphorylation on Ser616 progressively decreases (Fig. 3C), in line with a progressively less fragmented morphology of the mitochondria (Fig. 3D). Therefore, increasing PD-1 expression in CD8+ naïve T cells upon multiple in vitro stimulations correlates with a progressive inhibition of the ability of these cells to fragment mitochondria.
In sum, these data indicate that PD-1 signaling prevents Drp1-dependent mitochondria fragmentation in both naïve and memory murine CD8+ T cells. Also, impaired Drp1 activity is observed in in vitro-induced exhausted-like CD8+ T cells, too.

PD1 signaling downregulates Drp1 activation by modulating mTOR and ERK pathways.
Next, we decided to better investigate the molecular pathways linking PD-1 activation to downregulation of Drp1 activity. PD-1 signaling dampens both PI3K/Akt/mTOR pathway downstream of CD28 activation (3) and MAPK/ERK pathway downstream of TCR engagement (4), as indeed confirmed by our results (see Fig. 2A, D). Of note, both these pathways have been reported to modulate Drp1-dependent mitochondria fragmentation. Indeed, ERK1/2 directly phosphorylate Drp1 on Ser616 (25), as we already showed in T cells, too (15,16,26). Similarly, the mTOR pathway promotes mitochondria fragmentation via Drp1 activation (27), even though it is supposed to not directly phosphorylate Drp1, and rather acting through the modulation of ERK activity, as suggested in murine embryonic fibroblasts (27). Also, mTOR and ERK pathways extensively crosstalk and support each other (28,29). Therefore, our working hypothesis is that concomitant activation of PD-1 signaling upon T cell stimulation dampens both mTOR and ERK pathways, these being required for Drp1 phosphorylation on Ser616.
To test this hypothesis, we first activated murine T cells in the presence of mTOR inhibitor RAD-001 (30), and observed that RAD-001 prevents both mTOR auto-phosphorylation and ERK phosphorylation (Fig. 4A). Also, it reduces Drp1 phosphorylation on Ser616 and mitochondria fragmentation ( Fig. 4A-B). Second, we have reported previously that ERK1/2 inhibitor FR180204 (ERKi) prevents Drp1 phosphorylation upon both chemokine stimulation and TCR engagement in T cells (31), and further confirmed the latter condition here again (Fig. 4C). Also, ERKi prevents mitochondria fragmentation upon T cell activation (Fig. 4D). Therefore, inhibition of mTOR and ERK pathways (as observed upon activation of PD-1 signaling) impacts Drp1 activity and mitochondria fragmentation upon activation. Last, to confirm that ERK1/2 proteins, which are inhibited by PD-1 pathways, are the kinases directly responsible for Drp1 modulation, we rescued ERK1/2 activity downstream of PD-1 signaling. To this aim, we activated hPBT cells with aCD3/28-or aCD3/28-PDL1-beads in presence of low doses of C6-ceramide (to avoid apoptosis induction), a known activator of the ERK pathway (32). Of note, ceramide, which slightly increase ERK phosphorylation, rescues Drp1 phosphorylation on Ser616 and mitochondria fragmentation in PD-1-engaged T cells during activation ( Fig. 4E-F).
In sum, activation of PD-1 signaling during T cell stimulation reduces Drp1 phosphorylation and mitochondria fragmentation through direct inhibition of the ERK pathway, and concomitant dampening of mTOR activity. The latter may support indirectly Drp1-dependent mitochondria fragmentation by sustaining the ERK pathway, even though there is the possibility that it could also modulate Drp1 independently from ERK.

Drp1 is required for an efficient reduction of tumor growth mediated by anti-PD-1 therapy.
Most processes downregulated by PD-1 inhibitory signaling in T cells (proliferation, motility, glycolytic engagement) are the same requiring Drp1 to function correctly (4,15,21). Therefore, we asked to what extent PD-1 signaling may down-regulate such processes in T cells via inhibition of Drp1 activity, especially in cancer. To answer this point, we analysed the s.c. growth of MC38-derived tumors in WT mice, which can be significantly reduced by treatment with anti-PD-1 Ab, also increasing the amount of infiltrating leukocytes (33). Also, we previously reported that MC38-derived tumors growth faster in mice lacking Drp1 specifically in T cells (15). Therefore, we s.c. inoculated MC38 cells into both control (Drp1 fl/fl ) and Drp1 conditional-KO (Drp1 fl/fl Lck:cre+, T cell-restricted Drp1 ablation, indicated as Drp1cKO) mice. After one week, we treated these mice with either anti-IgG (control) or anti-PD-1 Ab every 2/3 days for up to 10 days (Fig. 5A). Of note, we found that anti-PD-1 treatment is much more efficient in reducing tumor growth in control mice than in mice where Drp1 was absent from T cells (Drp1-cKO) (Fig. 5B). Since Drp1 ablation in T cells favors per se a faster tumor growth in this model (15), we also compared the relative percentage of reduction in tumor growth (anti-PD-1 / anti-IgG ratio) to correct for the larger tumor volume in Drp1-cKO mice compared to control mice. Even so, we still observed that while anti-PD-1 treatment reduces tumor growth by ca. 60% in control mice, this percentage drops to just ca. 20% in Drp1-cKO ones (Fig. 5C). In addition, we found that in control (Drp1 fl/fl ) mice, where Drp1 phosphorylation is reduced in PD1 pos CD8+ TILs when compared with PD1 neg TILs, anti-PD-1 treatment is able to increase and restore Drp1 phosphorylation in PD1 pos CD8+ T cells to a level comparable to PD1 neg CD8+ T cells (Fig. 5D).
Therefore, anti-PD-1 treatment restores Drp1 phosphorylation on Ser616 in control CD8+ TILs and this correlates with a significant reduction in tumor growth. On the contrary, in Drp1-cKO mice (where Drp1 is absent and therefore cannot be "rescued" by anti-PD-1 treatment), anti-PD-1 treatment is much less efficient in reducing tumor growth. Overall, these data suggest that, among the processes restored by anti-PD-1 treatment in CD8+ exhausted TILs, at least some of them requires a functional Drp1 (and are therefore not rescued in Drp1-cKO mice) and, consequently, that the mechanisms by which PD-1 signaling inhibits T cell activity may require Drp1 inhibition at least for some of the processes involved.
Next, we decided to get more insight into such processes mediated by Drp1 downregulation downstream of PD-1 signaling. Of note, while anti-PD-1 treatment increases the number of CD8+ TILs recovered from the tumor mass (per mm 3 ), this effect is not observed in Drp1-cKO mice (Fig.  5E), indicating that Drp1 is required to mediate such anti-PD-1-dependent increase in CD8+ TILs accumulation. Although Drp1 is required to sustain IFNγ production in CD8+ TILs (15), this does not seem to be related to the impaired accumulation of Drp1-KO CD8+ TILs after anti-PD-1 treatment. Indeed, upon anti-PD-1 treatment, Drp1-KO CD8+ TILs increase IFNγ production to a level similar to control CD8+ TILs (Fig. 5F), suggesting that PD-1-dependent regulation of IFNγ production does not involve Drp1. We previously reported that Drp1 is also required to sustain T cell clonal expansion after stimulation, both in vitro and in vivo (15). Therefore, we asked whether the inability of Drp1-KO CD8+ TILs to increase their number within the tumor mass upon anti-PD-1 treatment may depend or not on their impaired proliferation. To this aim, we isolated TILs from MC38-derived tumor masses grown in control or Drp1-cKO mice (treated with anti-IgG or anti-PD-1) and let them expand in vitro in the presence of IL-2, IL-7 and IL-15 cytokines. Interestingly, while the anti-PD-1 treatment significantly increases fold expansion of control CD8+ T cells, this effect is completely lost in Drp1-KO CD8+ T cells (Fig. 5G). These data suggest that Drp1 7 inhibition may be at least one of the mechanisms by which PD-1 signaling reduces proliferation of PD1 pos CD8+ TILs.
In sum, a functional Drp1 is required for the efficacy of the anti-PD-1 treatment in reducing MC38-derived tumor growth in mice. Also, PD-1 signaling may mediate the reduction in CD8+ TILs proliferative potential via Drp1 downregulation.

Reduced motility of PD1 pos and in vitro exhausted-like T cells correlates with altered Drp1dependent mitochondria remodeling.
Besides controlling T cell proliferation, Drp1 is required to sustain T cell motility by favoring mitochondria repositioning at the uropod and it is directly phosphorylated on Ser616 in response to chemokine stimulation (15,34). Also, T cell motility is another process dampened by PD-1 signaling (21). Therefore, in addition to controlling proliferation of PD1 pos CD8+ TILs, the PD-1-dependent downregulation of Drp1 activity may also account for their reduced motility. To address this point, we isolated TILs from MC38-derived tumor masses grown in control or Drp1-cKO mice (treated with anti-IgG or anti-PD-1) and let them starve from serum in vitro. Then, we assayed their migratory ability in response to serum by using the transwell migration assay. Interestingly, we found that anti-PD-1 treatment significantly increases motility of control PD1 pos CD8+ TILs, when compared to the motility of the same cells from anti-IgG-treated control mice (Fig. 6A), while it does not improve the motility of PD1 neg CD8+ T cells, as expected (Fig. 6A). However, this effect is completely lost when looking at PD1 pos CD8+ TILs from Drp1-cKO mice, whose T cells lack Drp1 (Fig. 6A). These data suggest that Drp1 inhibition may be one of the mechanisms by which PD-1 signaling reduces motility of PD1 pos CD8+ TILs.
To analyze in more details the modulation of Drp1 in functional and exhausted T cells during T cell migration, we switched to our in vitro model to induce an exhaustion-like state in CD8+ T cells (as already described in Fig. 3C-D). We thus compared the migratory ability of effector (Teff) CD8+ T cells (stimulated only once) and exhausted-like (Tex) CD8+ T cells (stimulated 4 times). As expected, Tex cells migrate much less than Teff cells (Fig. 6B). Interestingly, we also observed that while functional Teff cells correctly polarize the mitochondria at the uropod (Fig. 6C), a process requiring Drp1 (15), and show both Drp1 and ERK1/2 phosphorylation upon chemokine stimulation (Fig. 6D), in Tex cells the mitochondrial network fails to rearrange (Fig. 6C), and Drp1 and ERK1/2 are unresponsive to chemokine stimulation ( Fig. 6D), this implying the lack of key steps normally required for T cell motility.
Overall, these data indicate that PD-1 signaling may mediate the reduction in CD8+ TILs motility via Drp1 downregulation. Also, in vitro-induced exhausted-like T cells are unable to respond to chemokine stimulation by activating Drp1 and by rearranging their mitochondrial network, this providing a mechanistic basis for their impaired motility. 8

DISCUSSION
It has been reported that tumor-infiltrating T cell show altered mitochondria functionality and morphology when compared with T cells from peripheral blood (35). However, modulation of mitochondria morphology in different subpopulation of TILs (such as PD1 pos and PD1 neg cells) has never been investigated before. We here report that PD1 pos MC38-derived tumor-infiltrating murine CD8+ T cells display an altered mitochondria morphology when compared with PD1 neg counterparts. Indeed, they are characterized by a reduced Drp1 activation and a more fused morphology of mitochondria. Of note, these data are shared also in a corresponding human context of colon tumor, in which tumor-infiltrating lymphocytic elements almost never co-express PD-1 and active Drp1. Mechanistically, we provided evidence that PD-1 signaling downregulates Drp1 activating phosphorylation on Ser616 (and consequently mitochondria fragmentation) via the inhibition of ERK1/2 kinases, which directly phosphorylate Drp1 on this residue (15,16,25). Also, our data indicate that concomitant PD-1-dependent inhibition of mTOR pathway contributes to such a reduced Drp1 activity. This occurs presumably through mTOR ability to support the ERK pathway, as suggested before (27), even though we cannot exclude that mTOR may modulate Drp1 independently from its role on ERK.
Also, we explored the functional consequences of such PD-1-dependent downregulation of Drp1 activity in a tumor context. Of the highest importance, altogether, our data suggest that PD-1 signaling may exploit the downregulation of Drp1 activity to dampen some of the processes required for an optimal T cell functionality. Specifically, Drp1 downregulation seems to be involved in the inhibition of motility and proliferation of PD1 pos CD8+ TILs. Indeed, anti-PD-1 treatment in mice fails to restore motility and proliferation of PD1 pos CD8+ TILs (as it should), when Drp1 is not present in these cells (and therefore cannot be re-activated). Of note, we observed that the Drp1-dependent mitochondria relocation at the cell rear edge during cell migration (a phenomenon reported to occur in healthy effector CD8+ T cells) is completely lost in exhausted CD8+ T cells, which are unable to activate Drp1 upon chemokine stimulation. Our data are thus consistent with previous observations made in a persistent infection mouse model, in which the recovered motility of PD1 pos T cells upon anti-PD-1 treatment were associated with an increased ERK activation (21). Indeed, we previously reported that, during T cell migration, ERK1/2 mediates Drp1 phosphorylation, which is then required to sustain T cell motility (15), and we here report that exhausted T cells fails to activate Drp1 and fragment mitochondria upon chemokine stimulation. Therefore, downregulation of ERK/Drp1 axis may be exploited by PD-1 signaling to dampen T cell motility. Regarding the role of Drp1 in T cell proliferation, we previously reported that, in the absence of Drp1, T cells show an abnormal length of mitosis, due to the acquisition of aberrant centrosome morphologies (15), as observed also in cancer cells (36).
In sum, our data indicate that downregulation of Drp1 mediated by PD-1 signaling may be required to attain an efficient inhibition of T cell response. Therefore, we dare to propose Drp1 as a therapeutic target to ameliorate exhausted T cell functionality during anti-cancer approaches, although drugs able to activate this protein need to be developed yet (26). Of note, CAR T cellbased approaches are currently being exploited for the treatment of solid cancers (37), but they frequently fail to confer long term tumor regression due to a poor ability of CAR T cells to survive and infiltrate within a solid tumor mass. This may be partially explained by the tendency of CAR T cells to undergo functional exhaustion, similar to endogenous T cells (38); however, whether or not a PD-1-dependent downregulation of Drp1 activity is present also in exhausted CAR T cells is still not known. Should such modulation be observed, targeting Drp1 activity in CAR T cells (either pharmacologically or genetically) could represent a new strategy to ameliorate CAR T cell survival (by increasing proliferation) or infiltration (by increasing motility).
In addition, although we here provided evidence that Drp1 downregulation may be exploited by PD-1 signaling to reduce tumor-infiltrating T cell motility and proliferation, other processes may be touched as well. First, we recently reported that Drp1 is required to allow cytochrome-c release and activation of caspases during the Activation-Induced Cell Death (AICD) progression in T cell following TCR engagement (16). Interestingly, Tex cells require continuous engagement with antigen to persist long term (39), and this occurs in the absence of apoptosis induction, which normally takes place in short-lived Teff cells via AICD. We are thus tempted to speculate that PD-1dependent Drp1 downregulation in Tex cells may also contribute to their long-term survival by reducing their tendency to undergo AICD. However, PD-1 signaling is also an apoptosis-inducer in T cell (40), this contributing to its inhibitory effect on T cell activation. Therefore, how exactly PD-1 and Drp1 correlate each other during T cell apoptosis is still not clear. Also, these data suggest that PD-1 signaling may regulate apoptosis in a Drp1-and mitochondria-independent way. Further studies will be required to shed light on this aspect.
Second, engagement of PD-1 signaling downregulates both glycolysis and OXPHOS in stimulated T cells (41). However, glycolysis is downregulated to a higher extent than OXPHOS, this forcing PD1-engaged T cells to rely more on an OXPHOS-based metabolism (22,41). Although this mechanism may allow tumor-infiltrating T cells to survive in a tumor microenvironment characterized by low glucose availability (22), the underlying mechanism favoring such metabolic adaptation is not known yet. Ogando et al. recently reported that PD-1 signaling mediates disassembly of mitochondrial cristae (41). Since the cristae are required for an optimal OXPHOS-based metabolism (42), this observation may provide a putative mechanism by which PD-1 modulate the OXPHOS rate in T cells. This said, no clues are available about the mechanism by which PD-1 regulates glycolysis. Mitochondrial fragmentation is frequently associated with the engagement of a glycolytic metabolism, especially in cancer cells (10). This occurs in T cells, too. Indeed, upon T cell activation Drp1 sustains cMyc activation via a calcium/AMPK/mTOR axis (15), this in turn promoting upregulation of glycolytic enzymes and full activation of glycolysis (18). Also, Drp1-dependent mitochondria fragmentation reduces the functionality of the ETC complexes (43), further favoring an OXPHOS-to-glycolysis shift in T cell metabolism upon activation (44). Interestingly, we recently observed that Drp1-KO T cells show a glycolysis-to-OXPHOS metabolic shift upon activation, this explained by an increased FAO engagement and a reduced upregulation of cMyc-dependent glycolytic genes (15). Therefore, our observation that PD1 pos CD8+ T cell downregulates Drp1 activation upon stimulation, besides contributing to their reduced proliferation and motility, may provide also a mechanistic basis for the downregulation of the glycolysis in these cells. Although further studies are required to firmly establish this point, we suggest the possibility that PD1 signaling could mediate the reduction of the glycolytic potential of CD8+ TILs at least partially via Drp1 downregulation.
Third, CD8+ TILs infiltrating human clear cell Renal Cell Carcinoma (ccRCC) were reported to show increased mitochondrial mass and ROS production (35). Drp1 may facilitate dismissal of small mitochondria via mitophagy (45). Interestingly, downregulation of Drp1 activity in PD1 pos CD8+ TILs may provide a mechanism to reduce the mitophagy rate in these cells, preventing the generation of small mitochondria that can be targeted to degradation. This is consistent with the observation of a higher mitochondrial mass in CD8+ TILs (35).
Of note, Ogando et al. recently analysed the effect of PD-1 signaling on the assembly of mitochondrial cristae (41). Although they reported no variations in the rate of mitochondrial fragmentation in T cells stimulated in the presence or absence of PD-1 engagement, they also reported no differences in the rate of mitochondrial fragmentation between unstimulated and stimulated T cells (41). However, fragmentation of the mitochondrial network upon T cell stimulation (independently of PD-1 signaling) has been reported extensively by us and others (15,17,23). We suppose that such discrepancy may rely partially on their use of the circularity as a parameter to estimate the rate of mitochondrial fragmentation upon T cell activation. Indeed, we believe that such a parameter may not be suitable to compare mitochondria morphology in stimulated and unstimulated naïve T cells, since the former cells grow and enlarge significantly compared to the latter ones upon in vitro culture. In addition, Ogando et al. only analysed total levels of Drp1, without focusing on its specific phosphorylated residues (41), which are more reliable indicators of Drp1 activation compared to the total protein amount.
To conclude, the modulation of Drp1 in tumor-derived exhausted T cells may represent a valuable target to ameliorate anti-cancer immune response in a number of instances and the manipulation of CAR T systems to this aim may represent a valid future strategy.

Human samples
Peripheral blood samples were purified from buffy coats of healthy volunteer blood donors (

Mice
WT and Drp1 fl/fl Lck:cre+ c57BL/6 mice were bred and maintained under conventional conditions at the Plaisant Srl (Castel Romano) Animal Facility. Drp1 fl/fl Lck:cre+ mouse strain has been previously described (15). Mice were kept in cages of no more than 5-6 mice each, divided by sex, under 12h/12h light/dark cycles, with standard temperature, humidity and pressure conditions according to FELASA guidelines. Small red squared mice house and paper were used for cage enrichment. Mice health was monitored daily by veterinary staff and health analysis for pathogens were performed every three months according to FELASA guidelines. All mice were sacrificed by neck dislocation at 2-3 months of age. All efforts were made to minimize animal suffering and to reduce the number of mice used, in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC). The mice protocol has been approved by the Allevamenti Plaisant Srl Ethical Committee as well as by the Italian Ministry of Health (Authorization #186/2020-PR). It has been written following the ARRIVE Guidelines, and the numeric details have been chosen following the criteria described in The National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs) (http://www.nc3rs.org.uk/). Sample size for the experiments performed has been established using power analysis method. Experiments involving growth of tumor cells in mice where performed using male mice.
To induce isolated murine naïve CD8+ T cells into an exhaustion-like state in vitro, cells have been stimulated up to 4 times with 5mg/ml anti-CD3 (plate-coated) (eBioscience 14-0031-86) and 1mg/ml anti-CD28 (Invitrogen 14-0281-86) for 24h in 96well plate. Between each stimulation, cells have been expanded using 20ng/ml mouse IL-2 (R&D System 402-ML). Cells were considered into exhaustion-like state after 4 stimulations (Tex) and were compared with effector-like (Teff) cells isolated from sibling mice and stimulated only once. For in vitro migration experiments, Tex and Teff cells were finally expanded in vitro for additional 6 days in IL2-containing medium and then used for the assays.

Western Blot
Western blot were performed as previously described (15 . All primary antibody incubations were followed by incubation with appropriated secondary HRP-conjugated antibodies (GE Healthcare or Cell Signaling) in 5% milk plus 0.1% Tween20 (Sigma P2287). Detection of protein signals was performed using Clarity Western ECL substrate (Biorad 170-5061) and Amersham Imager 600. Stripping of the membranes for re-probing has been performed using buffer containing 1% Tween-20 (Sigma P2287), 0.1% SDS (Sigma 71729), and 0.2M glycine (VWR M103) at pH 2.2 (two washes for 10min). Original scan of western blot images shown throughout the manuscript are reported in Supplemental Figure 4.

Immunofluorescence
Immunofluorescence staining has been performed as previously described (16). Anti-TOM20 (Santa Cruz sc-11415) primary antibody was used to identify the mitochondrial network. Nunc Lab-Tek Chamber Slides (Thermo Fisher 154534) have been used to culture in vitro T cells directly on slides before fixation and were coated with 10ng/ml fibronectin (Millipore FC010) for 1h at RT before adding the cells. Images were acquired using a Perkin Elmer Ultraview VoX microscope. The mitochondrial network has been always evaluated upon 0.4mm slices z-stack reconstructions.

Immunohistochemistry
Formalin-fixed and paraffin embedded (FFPE) tissue samples of human colon cancer moderately differentiated (G2) cases were selected for in situ immunophenotypical analyses. Fourmicrometers-thick sections were deparaffinized, rehydrated and unmasked using Novocastra Epitope Retrieval Solutions pH 9 in PT Link Dako at 98°C for 30 minutes. Subsequently, the sections were brought to room temperature and washed in PBS. After neutralization of the endogenous peroxidase with 3% H2O2 and Fc blocking by a specific protein block (Novocastra), the samples were incubated with anti-pSer616-Drp1 (clone D9A1 Cell Signaling, 1:100) and anti-PD-1 (clone NAT105 Abcam, 1:50) antibodies. IHC staining was revealed using "MACH 2 Double Stain 1" kit and DAB (3,3'-Diaminobenzidine, Novocastra) and Vulcan Fast Red or Ferangi Blue as substrate-chromogens. Slides were analyzed under a Zeiss Axioscope A1 and microphotographs were collected using a Zeiss Axiocam 503 Color with the Zen 2.0 Software (Zeiss).

Flow Cytometry
The following antibodies have been used to stain extracellular proteins: anti-CD8-Alexa488 To evaluate IFNγ production in tumor-derived T cells, cells have been restimulated for 4h in presence of 50ng/ml PMA (Sigma 79346), 1μg/ml ionomycin (Sigma I9657) and 2μM monensin (added for the last 2h, Sigma M5273) and then fixed and processed using Foxp3 Transcription Factor Staining Buffer Set (00-5523-00, eBioscience) and anti-IFNg antibody (eBioscience 12-7311-82). Acquisitions have been performed using BD Accuri C6 and BD FACSCelesta cytometers. Cell sorting have been performed by staining cells with the aforementioned extracellular antibodies and using BD FACS Aria III flow cytometer.
For the evaluation of the mitochondrial membrane potential, 1mM TMRE (Thermo Fisher T669) has been added for 20 min and then the cells were washed and analysed. As a positive control for mitochondria depolarization, cells have been pre-treated with 50μM FCCP (Sigma Aldrich C2920).

Tumor induction
5*10 5 MCA38 cells were injected subcutaneously into the right flank of two months-old male WT or Drp1 fl/fl and Drp1 fl/fl Lck:cre+ mice. Mice were kept for up to 17/18 days in animal facility, and tumor growth was monitored twice or three times per week and recorded as [longest diameter]*[shortest diameter] 2 in cubic millimetres. At days 7, 9, 11, 14 and 16 from tumor inoculation, mice were inoculated i.p. with 150µg of InVivoMab anti-mouse PD-1 (CD279), clone RMP1-14 (Bioxcell, BE0146) or 150µg InVivoMab rat IgG2a isotype control, clone 2A3 (Bioxcell, BE0089) antibodies (in 150µl of saline). Mice were randomly subdivided into each experimental group (a-IgG or a-PD1) before inoculation of antibodies (no specific randomization method was used). At day 17, mice were sacrificed and tumors were collected. Tumor tissues were mechanically dissociated over 70 mm-cell strainers, and mononuclear cells were enriched from tumor-derived cell suspensions by 40%/80% Percoll (GE Healthcare GE17-0891-01) density gradient, by collecting cells at the interface between 40% and 80% Percoll solution.
Isolated TILs have been used for subsequent proliferation and migration in vitro analyses or stained for flow cytometric measurements.

Seahorse analysis
Basal OCR has been measured in T cells during acute phase (unstimulated or stimulated for 12h) or after 48h of stimulation and 4 days of in vitro expansion with IL2, as previously described (15).

In vitro proliferation and migration assays
To evaluate the in vitro proliferation potential of tumor-infiltrated CD8+ T cells isolated from tumor-bearing Drp1 fl/fl and Drp1 fl/fl Lck:cre+ mice, 2x10 5 isolated TILs have been cultured in vitro in 96well plate in presence of 20ng/ml mouse IL-2 (R&D System 402-ML), 20ng/ml mouse IL-7 (R&D System 407-ML) and 20ng/ml mouse IL-15 (R&D System 447-ML). The total number of cells were estimated each 2/3 days using BD Accuri C6 flow cytometer and the percentage of CD8+ T cells in each plate evaluated by flow cytometry staining with anti-CD8-A488 (Biolegend 100723) antibody.

Statistical Analysis
In the Figure legends, "n" indicates either the number of independent experiments (in vitro primary cells) or the number of mice used. Data are expressed as mean ± SEM from at least three independent experiments unless specified otherwise (Microsoft Office Excel and SigmaPlot v12.5 have been used for analysis). The number of mice used has been estimated using the power analysis method. All the acquisitions of the experiments have been performed blinded without knowing the specific conditions of each sample. Comparisons between groups were done using two-tailed Student's T-test (two groups) or One-way and Two-way ANOVA (multiple groups and repeated measurements, adjustments for pairwise comparisons were performed using Holm-Sidak method). Mann-Whitney Rank Sum Test or ANOVA on ranks have been used if samples did not meet assumptions of normality and/or equal variance. Chi-square test has been used to evaluate data in Figure 1B Fig. 5A. Then, TILs were starved from serum for 2h and allowed to migrate in response to 10% Fetal Bovine Serum for 2h using transwell migration assay. The graph indicates the relative (anti-PD1 / anti-IgG) migration index calculated for PD1 neg and PD1 pos CD8+ TILs isolated from tumor-bearing control Drp1 fl/fl or conditional-KO Drp1 fl/fl Lck:cre+ mice inoculated with anti-IgG or anti-PD1 antibodies as indicated in Fig. 5A (n=7). (B-D) Murine exhausted T cells (Tex) have been generated in vitro through 4 cycles of a-CD3/28-mediated stimulation (24h) and IL2-mediated expansion (6 days) and compared to effector T cells (Teff) generated through a single cycle of stimulation and expansion. After the last 6 days in IL2-containing medium, cells have been starved from serum for 2h and then the following assays were performed. In (B) the migration index in response to CCL19/CCL21 gradient for 2h has been calculated using transwell migration assay (n=6). In (C) cells have been left to adhere for 30min to fibronectin-coated slides. Then, cells have been stimulated with 50nM CCL19 and 50nM CCL21 chemokines for 15min and then fixed and processed for immunostaining. Representative images showing the mitochondrial network (anti-TOM20 staining) in effector T (Teff) and exhausted T (Tex) cells are shown on the left. Quantification of the percentage of cells showing fragmented mitochondria in each condition is reported in the graph on the right (n=3). In (D) the cells have been stimulated with 50nM CCL19 and 50nM CCL21 chemokines for 15min in an eppendorf tube and then proteins were extracted. The expression level of the indicated (phospho)-proteins has been evaluated by western blot (n=3). Data are shown as mean ± SEM. Scale bar: 5µm in C. Significance is indicated as follows: *=p<0.05; **=p<0.01; ***=p<0.001.