Peptide-HLA-based immunotherapeutics platforms for direct modulation of antigen-specific T cells

Targeted pharmacologic activation of antigen-specific (AgS) T cells may bypass limitations inherent in current T cell-based cancer therapies. We describe two immunotherapeutics platforms for selective delivery of costimulatory ligands and peptide-HLA (pHLA) to AgS T cells. We engineered and deployed on these platforms an affinity-attenuated variant of interleukin-2, which selectively expands oligoclonal and polyfunctional AgS T cells in vitro and synergizes with CD80 signals for superior proliferation versus peptide stimulation.


Scientific Reports
| (2021) 11:19220 | https://doi.org/10.1038/s41598-021-98716-z www.nature.com/scientificreports/ on the natural signals governing T cell function: peptide-HLA and costimulatory ligands, embodied in the Immuno-STAT (Selective Targeting and Alteration of T cells) and Neo-STAT immunotherapeutics platforms. Immuno-STAT and Neo-STAT utilize compact, Fc-based architectures amenable to clinical manufacturing, and are designed to focus optimized signals for potent costimulatory axes such as IL-2 directly to AgS T cells in vivo, thereby enhancing anti-tumor T cell responses while avoiding indiscriminate immune activation.

Results
The Immuno-STAT framework comprises a covalent fusion of peptide epitope, MHC class I allele, co-modulator, and Fc, which imparts avidity and symmetrical multivalency, sufficient for cognate T cell activation (Fig. 1a) 10 . Potential co-modulators, costimulatory and co-inhibitory ligands, may be fused to the N-or C-terminus of the MHC-Fc heavy chain or the C-terminus of β2m, allowing organizational, compositional and stoichiometric flexibility. Our initial exploration of the Immuno-STAT platform utilized IL-2 as the co-modulator to activate and expand AgS cytotoxic effector T cells (Teff).
To identify an optimized IL-2-based Immuno-STAT framework, we evaluated a panel of constructs for relative potency, AgS selectivity, and manufacturability. Constructs comprised LCMV gp 33-41 /H-2D b , recognized by the murine TCR P14, fused at its N-terminus to variants of human IL-2 and C-terminally to an effectorattenuated murine IgG2a Fc. IL-2-attenuating mutations were included to limit IL-2Rα-dependent toxicity and Treg engagement as well as to reduce IL-2 affinity in favor of pMHC selectivity (see Supplementary Fig. 1) 11,12 . IL-2 stoichiometries were limited to two or four based on manufacturability which showed a significant drop in protein titer beyond four copies of IL-2 (see Supplementary Fig. 2). Human IL-2 exhibits potent activity on both human and mouse cells including phosphorylation of STAT5, an IL-2 receptor (IL-2R) proximal signaling molecule and phosphorylated STAT5 (pSTAT5) serves as an index of IL-2R engagement which correlates well with downstream consequences of IL-2R agonism such as proliferation and phenotypic marker expression 12 . We compared pSTAT5 induction for purified CD8 splenocytes from AgS P14 TCR transgenic mice to non-AgS C57BL/6 mice and ranked constructs based on logEC50 P14 (potency index), P14 minus C57BL/6 signal at EC50 P14 (selectivity index), and protein expression titer (see Supplementary Fig. 3 and Supplementary Table 1). Construct LCMV-IST-IL2.FH 4 ranked highest followed by LCMV-IST-IL2.F 4 , bearing four copies of IL-2 F42A, H16A or IL-2 F42A , respectively. Dose responses for top candidates LCMV-IST-IL2.FH 4 and LCMV-IST-IL2.F 4 from this initial screen were reevaluated over a broader concentration range and with greater resolution and compared with reference constructs bearing two or four copies of wildtype IL-2, LCMV-IST-IL2 2 or LCMV-IST-IL2 4 , respectively ( Fig. 1b and see Supplementary Fig. 4). AgS potency, as indicated by the logarithm of the pSTAT5 EC50 for responding P14 splenocytes (logEC50 P14 ), was not significantly different between LCMV-IST-IL2. FH 4 , LCMV-IST-IL2.F 4 , LCMV-IST-IL2 2 and LCMV-IST-IL2 4 (see Supplementary Table 2 and Supplementary  Fig. 5). Likewise, non-AgS potency (logEC50 B6 ) was also similar for these constructs (see Supplementary Table 2 and Supplementary Fig. 5). AgS selectivity was measured first by the difference in normalized pSTAT5 signal at EC50 P14 for AgS (P14) splenocytes (i.e. 50%) relative to non-AgS (C57BL/6) splenocytes as well as by the difference in the logEC50 for AgS versus non-AgS splenocytes (see Supplementary Fig. 6). Both measures of AgS selectivity were not significantly different across constructs LCMV-IST-IL2.FH 4 , LCMV-IST-IL2.F 4 , LCMV-IST-IL2 2 and LCMV-IST-IL2 4 . Thus, significant F42A-and H16A-mediated decreases in IL-2R signaling potency or selectivity were not observed relative to reference constructs bearing two or four copies of wildtype IL-2 (LCMV-IST-IL2 2 or LCMV-IST-IL2 4 ), presumably masked in part by the increased IL-2 stoichiometry.

Discussion
We have described two modular immunotherapeutics platforms, Immuno-STAT and Neo-STAT, for the copresentation of defined peptide-HLA (pHLA) and co-modulatory ligands to AgS T cells, and engineered an affinity-attenuated variant of IL-2, IL2.FH, for use thereon. IL2.FH-bearing constructs, including an HPV-specific clinical candidate, CUE-101 (NCT03978689), are associated with 110-fold and 3.1-fold decreases in binding to IL2Rα and IL2Rβ, respectively, with minimal Treg and non-AgS T cell responses in vitro and in vivo 14 . While greater AgS selectivity was anticipated for LCMV-IST-IL2.FH 4 over LCMV-IST-IL2 2 due to attenuation of IL-2:IL-2R versus pHLA:TCR binding, this trend was not statistically significant. Potentially more selective variants (e.g. F42A D20K) affecting critical contacts with IL-2Rβ were associated with low peak activities, suggesting trade-offs between selectivity and Rmax 15,16 . Integration of Immuno-STAT signals through the TCR and IL-2R and/or more IL-2R distal events such as cytokine release may reflect greater selectivity than pSTAT5 induction. Proliferation showed greater selectivity versus pSTAT5 induction for AgS CD8 T cells in response to IST-IL2. FH 4 , extending similar results obtained with CUE-101. In addition, through modestly attenuated IL-2Rβ binding, H16A may enable strong, but submaximal IL-2 signals, thereby limiting burnout/terminal differentiation while driving proliferation, effector function and memory T cell generation 16,17 .
AgS cytotoxic CD8 T cell responses have previously been achieved via heterologous expression of membrane pMHC complexes, for example, by DNA vaccination 18 . Alternatively, nanoparticle display of pMHC can stimulate or inhibit AgS T cell responses depending on the presence and nature of costimulatory and co-inhibitory ligands 19,20 . In contrast, soluble pMHC complexes, including streptavidin-based tetramers, largely suppress T cell responses 21,22 . To our knowledge, Immuno-STAT and Neo-STAT molecules are the first embodiments of soluble pMHC-comodulator signaling complexes directed to AgS T cells in cis.
Immuno-STAT and Neo-STAT access diverse HLA alleles and epitopes in compact architectures which may benefit clinical manufacturability and biodistribution 23 . HLA-A*0201, HLA-A*1101 and HLA-A*2402, surveyed here in the context of Immuno-STAT, are collectively expressed by over half the global population and Neo-STAT extends potential pHLAs to include post-translational modifications and low affinity peptides 24 . In preclinical murine studies, CUE-101 and mCUE-101 elicit AgS T cell responses from both naïve and antigen-experienced precursors, with responses to mCUE-101 detectable in blood, spleen, and tumors, suggesting the potential to both invigorate T cell responses within tumors and initiate responses outside of immunosuppressive tumor microenvironments, which are relatively resistant to exhaustion 14,25 . Compelling applications of these platforms include enhancing T cell responses against essential antigens in cancer or infectious disease, especially where relevant AgS T cells are limiting or dysfunctional, and more efficiently activating or maintaining ACT cells ex/in vivo. In addition, conjugating clinical grade peptides and Neo-STAT precursors enables a rapid and parallelizable manufacturing paradigm for immunotherapeutics capable of addressing neoantigens, multi-epitope combinations to address tumor heterogeneity and mitigate tumor escape and rapid responses to emerging pandemics.
As shown here, distinct Immuno-STAT co-modulators can synergize for superior AgS T cell activation in vitro. Moreover, unlike current vaccines, Immuno-STATs and Neo-STATs may bypass APC in vivo to drive therapeutic T cell responses. The same core architectures bearing appropriate comodulators may also be used to engage other costimulatory pathways, such as CD137 and CD70 26,27 . Likewise, Immuno-STATs and Neo-STATs may be used to engage or block inhibitory pathways such as PD-1 and CTLA-4 for AgS activation or downregulation, respectively, while limiting toxicity 7 . Whereas the present work has focused on classical HLA class I restricted T cells, alternative versions of these scaffolds may, in principle, be used to engage any AgS T cell, including T cells specific for non-classical HLA class I ligands as well as HLA class II-restricted ligands.
Immuno-STAT and Neo-STAT proteins were expressed by transient transfection in Expi-CHO cells (Ther-moFisher). Proteins were purified from the conditioned media using a two-step method of ProteinA capture with MabSelect SuRe (GE) followed by size exclusion chromatography. For SDS-PAGE analysis, proteins were boiled in SDS sample buffer with or without reducing agent for 5 min before loading 2 µg per gel lane.
Empty Neo-STAT precursor proteins were linked to maleimide-conjugaged peptides using standard maleimide chemistry. Briefly, empty Neo-STAT precursors were exchanged into and partially reduced with a TCEPbased reducing buffer before two rounds of conjugation with a 20 fold molar excess of peptide-maleimide in the absence of TCEP. Conjugated Neo-STAT proteins were washed at low pH to remove excess unconjugated peptides in solution before purification by size exclusion and mass confirmation by electrospray ionization time of flight mass spectrometry (ESI-TOF MS).
Animals studies. Spleens were collected from C57BL/6J and P14 T cell receptor (specific for LCMV gp 33-41 /H-2D b ) transgenic mice (Jackson Labs) at least six weeks of age following euthanasia by CO 2 inhalation and confirmation of euthanasia by cervical dislocation. All studies requiring animal tissues were approved by the  (clone HIB19, BioLegend), CD56 (clone HCD56, BioLegend), CD4 (clone SK3, BioLegend) and CD8 (clone SK1, BD Biosciences) for 30 min. Tetramer flow cytometric data was acquired using the Attune NxT cytometer (ThermoFisher) and analyzed using FlowJo software (Tree Star). Peak fold expansion per donor per trial was calculated as the maximum tetramer-positive frequency observed for PBMC expanded with specific IST-IL2. FH 4 over tetramer-positive frequency for media-incubated PBMC. Peak fold expansion data was derived from between one and three expansion trials per donor, with mean peak fold expansion values used for donors with multiple expansion trials.