Targeting dendritic cells with a PD-L1 based bispecific antibody rejuvenates specific anti- tumor T cells

Bispecific T-cell engagers (BiTEs) that preferentially target tumor-associated antigens (TAA) to 13 reengage CD3 signaling have been approved to treat acute B-cell lymphoblastic leukemia. 14 However, their applications in solid tumors have been hampered due to short half-life, weak anti- 15 tumor activity, and severe toxicity at therapeutic doses. To explore new targets, we designed a 16 bispecific antibody (BsAb) which simultaneously targets CD3 and immune checkpoint PD-L1. 17 Compared with conventional TAA based targeting, PDL1xCD3 generates far superior anti-tumor 18 immune responses in vivo . Mechanistically, blockade of PD-L1 on dendritic cells instead of tumor 19 cells can potently rejuvenate preexisting tumor reactive CD8 T cells in a B7-1/2 dependent manner 20 for a durable anti-tumor responses. This study argues that targeting DC-T cell instead of current 21 tumor-T cell can achieve much better T cell rejuvenation in BsAb therapy. on 725 cytometry GFP were detected by K (G) CBA. two independent as mean ± SEM (n=5). Statistical analysis was performed by two-tailed unpaired t test (C-G), two-way ANOVA and Log-rank test (H-I)

Introduction the therapeutic effect of BiTE treatment can be improved in combination with checkpoint blockade 86 37 . 87 To overcome the limitations of current tumor cell targeting BiTE therapy, we designed a 88 novel bispecific antibody that targets immune checkpoint PD-L1 to redirect T cells to APCs. We 89 unexpectedly observed that PDL1xCD3 generates much better anti-tumor effect than conventional 90 TAA targeting bispecific antibody (EGFRxCD3) in vivo. We also reveal a new target on APCs to  Figure 2C). More importantly, PDL1xCD3 treatment also induced OTI-specific IFNγ producing 157 cells in the spleen of MC38OVA bearing mice, further confirmed the efficient generation of 158 antigen specific T cell response ( Figure 2D and S1A). Thus, in contrast to convention BiTE, we 159 hypothesized that PDL1xCD3 might provide a distinct signal to T cells, which triggers a specific 160 immune response against tumor without causing severe side effect. 161 In order to test this hypothesis, we first explored whether PDL1xCD3 could generate  Figure S2F). In addition, 168 tumor-free mice from PDL1xCD3 treated groups also obtained memory immunity to reject re-169 challenged tumor cells ( Figure 2F and 2H). To exclude the dose effect which may cause ErbxCD3 170 to be ineffective, we treated mice with ErbxCD3 intratumorally. Even though i.t. injection of 171 ErbxCD3 had an improved anti-tumor effect compared to i.p. injection, the overall anti-tumor 172 effect is still weaker than PDL1xCD3 and no ErbxCD3 treated mice become tumor-free after 173 treatment ( Figure S2D). Therefore, dose effects did not contribute to the resistance of ErbxCD3 in 174 vivo. Taken together, using syngeneic mouse models of multiple cancer types, we demonstrated 175 PDL1xCD3 generate superior anti-tumor effect than ErbxCD3. These data also raises the 176 possibility that PDL1xCD3 creates a unique microenvironment by engaging different signal 177 pathways or inducing different cell-cell interactions.

178
Pre-existing CD8 T cells are required for PDL1xCD3 treatment. 179 Next, we investigated the mechanisms underlying the therapeutic effects of PDL1xCD3.

180
PDL1xCD3 has no effect on MC38 bearing Rag1 -/mice, which confirms that adaptive immunity 181 is essential for the therapeutic effect of PDL1xCD3 ( Figure 3A). Moreover, CD8 T cells but not 182 CD4 T cells contribute to this effect ( Figure 3B). To further determine whether PDL1xCD3 183 treatment depends on pre-existing T cells in the tumor microenvironment (TME) or recruitment of T cells from peripheral tissue, we used FTY720, a S1P receptor agonist, to block T cell trafficking to tumor tissue during PDL1xCD3 treatment. As shown in Figure 3C, additional FTY720 blocking 186 did not affect the therapeutic effect of PDL1xCD3, indicating the critical role of pre-existing CD8 187 T cells in the TME for this treatment. To further clarify whether PDL1xCD3 is targeting tumor 188 tissue to induce the anti-tumor effect, we intratumorally treated mice with the fusion protein. As   PDL1xCD3 reshapes a distinct immunophenotypic signature in tumor-bearing mice 229 Since PDL1xCD3 targets PD-L1 on DCs to facilitate a superior immune response to ErbxCD3, we 230 want to further investigate how the TME is reshaped to promote CD8 T cell responses.   As previous studies have shown, antigen presenting cells maintain the stem-like CD8 T 243 cell niche in the TME 31,43 . We have also observed that our fusion protein targets PD-L1 on DCs to reactivate T cells. Thus, it is possible that stem-like CD8 T cells but not terminally exhausted CD8 T cells were preferentially activated by PDL1xCD3, due to their physiological co-localization with 246 DCs. More importantly, the percentage of antigen specific CD8 T cell in the tumor also increased 247 after PDL1xCD3 treatment ( Figure 5F). Thus, PDL1xCD3 may preferentially activate a 'DC-248 interacting' population of CD8 T cells to establish specific anti-tumor immunity. In addition, we 249 also examined the dynamics of myeloid cells in the TME. The percentage of both macrophages 250 and MDSCs dramatically decreased after treatment since they are considered as 'PD-L1 + targets' 251 ( Figure 5G and 5H). Meanwhile, the percentage of DCs also significantly decreased even though 252 they are required for the initiation of the anti-tumor effect ( Figure 5I). Notably, the percentage of 253 Tregs was also decreased despite the increase in total CD4 percentage ( Figure S5F). Even though 254 the mechanism is still unknown, the expression of PD-L1 on Tregs has been reported, which may  Co-stimulatory signaling is required for PDL1xCD3 mediated anti-tumor effects 261 The generation of protective T cell immunity is one of the most desired goals in cancer 262 immunotherapy. However it remains the major hurdle of current tumor cell targeting BiTE therapy.

263
Our data has shown that PDL1xCD3 can target DCs to promote antigen specific T cell immunity.

264
Therefore, we want to further investigate the underlying mechanisms participating in this process.

265
Previous studies have shown that the therapeutic effect of anti-PD-L1 treatment is CD28 dependent 266 which highlights the importance of co-stimulatory signaling in generation of T cell immunity 45,46 . 267 Therefore, we hypothesized that the therapeutic effect of PDL1xCD3 may depend on T cell co-

300
The implementation of bispecific T-cell engagers to solid tumors has been hampered presumably 301 due to not only short half-life, poor anti-tumor activity and severe toxicity. It raises the possibility 302 that current targeting TAA might not be right strategies. In the presented study, we designed and   PDL1xCD3 brings T cells to this rare but potent APC for their re-activation. 391 As shown in our data, redirecting T cells to tumor cells for killing only induces a limited 392 immune response. T cells that are activated by CD3 engagement also undergo AICD due to lack 393 of CD28 co-stimulation. IFNγ will not only kill tumor cells but also induce T cell apoptosis 56 .