Engineered red blood cells as an off-the-shelf allogeneic anti-tumor therapeutic

Checkpoint inhibitors and T-cell therapies have highlighted the critical role of T cells in anti-cancer immunity. However, limitations associated with these treatments drive the need for alternative approaches. Here, we engineer red blood cells into artificial antigen-presenting cells (aAPCs) presenting a peptide bound to the major histocompatibility complex I, the costimulatory ligand 4-1BBL, and interleukin (IL)-12. This leads to robust, antigen-specific T-cell expansion, memory formation, additional immune activation, tumor control, and antigen spreading in tumor models in vivo. The presence of 4-1BBL and IL-12 induces minimal toxicities due to restriction to the vasculature and spleen. The allogeneic aAPC, RTX-321, comprised of human leukocyte antigen-A*02:01 presenting the human papilloma virus (HPV) peptide HPV16 E711-19, 4-1BBL, and IL-12 on the surface, activates HPV-specific T cells and promotes effector function in vitro. Thus, RTX-321 is a potential ‘off-the-shelf’ in vivo cellular immunotherapy for treating HPV + cancers, including cervical and head/neck cancers.

T cells play a critical role in anticancer immunity through the specific recognition of cancer antigens by T-cell receptors (TCRs) 1 . Approaches designed to manipulate or mimic the T-cell response (such as checkpoint inhibitors and chimeric antigen receptor T [CAR-T] cells, respectively) have been an important focus of anticancer therapeutics in recent years. However, the limitations associated with these treatments (including fatal adverse effects [2][3][4] and the development of resistance in some cases 5,6 ), as well as the complex and costly personalized production processes required for autologous CAR-T cells [7][8][9] , have driven continued investigation into alternative approaches to stimulate T-cell-mediated antitumor responses Effective T-cell activation requires three distinct signals: engagement of the TCR by a peptide bound to a major histocompatibility complex (MHC) molecule (signal 1); a costimulatory signal to promote T-cell activation, function, and survival (signal 2); and a cytokine signal to facilitate further expansion and differentiation of T cells, enhanced effector function and the formation of immunological memory (signal 3) 10,11 . Signals 1 and 2 are typically delivered by antigen-presenting cells (APCs) 12 . Dendritic cells or artificial APC (aAPC) platforms utilizing fibroblasts, microbeads, biodegradable polymers, and K562 cells have been investigated as anticancer therapeutics to stimulate T cells either in vivo or ex vivo [13][14][15][16] .
Red blood cells (RBCs) have unique properties that make them attractive for allogeneic cell therapy, including inherent biocompatibility when O-negative donor blood is utilized 17 . This has led to their long-term use and familiar history in transfusion medicine with very few side effects 18 . Moreover, RBCs are associated with multiple mechanisms of immune privilege that protect them from inhibitory or adverse host responses 18 . From the manufacturing standpoint, cell culture and differentiation of hematopoietic progenitor cells allows for many-fold expansion of erythroid cells. In addition, via genetic modification 19 , it is possible to generate engineered human RBCs that express biotherapeutic proteins, herein termed Red Cell Therapeutics ™® or RCTs ™ . As erythroid precursor cells enucleate, the expressed protein is maintained, but the genetic modifications performed on progenitor cells are not carried into the final product intended for patient administration 19 . Lastly, since RBCs are confined to the vasculature in most parenchymal organs and the spleen, they have the potential to avoid some of the on-target toxicities seen with circulating immunomodulatory proteins 20,21 .
Here, we describe the development, characterization, and preclinical testing of our allogeneic, Red Cell Therapeutics aAPC platform (RCT-aAPC) comprised of a tumor-specific peptide bound to MHC class I, a costimulatory ligand (4-1BBL) and a cytokine signal, interleukin 12 (IL-12). We demonstrate the ability of such an RCT-aAPC to drive antigen-specific T-cell expansion and acquisition of effector function both in vitro and in vivo, and to control tumors in preclinical mouse models. Importantly, tumor control is associated with the development of long-term memory and epitope spreading, leading to efficacy against tumors that do not express the original target antigen but are otherwise identical. These functions led to the creation of a clinical candidate, RTX ™ -321, which expresses human leukocyte antigen (HLA)-A*02:01 with human papillomavirus (HPV) 16 E7 peptide [11][12][13][14][15][16][17][18][19] (HLA-A2-HPV), 4-1BBL, and IL-12. We show that RTX-321 induces the activation of HPV antigen-specific primary human T cells, and that all three signals are sufficient for robust effector function and differentiation of effector memory cells. Given the poor survival rates associated with some recurrent HPV + cancers 22 , our study shows that RTX-321 represents a promising strategy for clinical investigation in multiple tumor types.
We found that treatment with mRBC-OVA-4-1BBL-IL-12 in the EG7.OVA plus OT-1 transfer model was superior to vaccination with long OVA peptide plus incomplete Freund's adjuvant (IFA; Fig. 1i). Median overall survival (OS) following vaccination with long OVA peptide plus IFA was 17.5 days, which was comparable to the median OS observed with mRBC-CTRL (20 days). In contrast, the median OS following treatment with mRBC-OVA-4-1BBL-IL-12 was 49 days (P = 0.0002 vs OVA vaccine; P = 0.0226 vs mRBC-CTRL). Signal 1 was required for optimal efficacy as treatment with mRBC-4-1BBL-IL-12 led to a median OS of 28.5 days, which was not statistically different from mRBC-CTRL. Only mRBC-OVA-4-1BBL-IL-12 treatment led to expansion of OT-1 cells in the blood (Fig. 1j), which correlated with efficacy.
Biodistribution and tolerability of mRBC-OVA-4-1BBL-IL-12. As toxicities have been previously observed with 4-1BB agonists 20,25 and recombinant IL-12 in the clinic 21,26 , we performed safety measurements after repeat doses of mRBC-OVA-4-1BBL-IL-12 in mice with or without OT-1 transfer (Fig. 2a), which showed that there were no significant changes in body weight compared to control treated mice (Fig. 2b). Plasma IFNγ levels ( Fig. 2c) and serum alanine aminotransferase (ALT) levels ( Fig. 2d) increased during the dosing phase, with the highest dose of 1 × 10 9 cells, but returned to baseline after a 2-week recovery period. There were minimal observed toxicities at the lower dose of 3 × 10 8 cells, which nonetheless was sufficient for antitumor efficacy (Fig. 1f, g). Additional safety measurements, including spleen and liver weights, liver enzyme levels, liver inflammation and macrophage infiltration, hematology changes, and plasma cytokine levels, indicated similar minimal and reversible toxicity with or without OT-1 transfer ( Supplementary Fig. 5). We hypothesized that the lack of significant toxicity seen with mRBC-OVA-4-1BBL-IL-12 may be due to the restricted biodistribution of RBCs compared with antibodies and cytokines. Minimal mRBC-OVA-4-1BBL-IL-12 was found in all tissues examined, with the notable exception of the spleen (Fig. 2e).
The anatomy of the splenic vasculature allows RBCs to directly interact with lymphoid cells in the parenchyma of the red pulp 27 , which led us to hypothesize that mRBC-OVA-4-1BBL-IL-12 may take part in cognate interactions with T cells in this location. Indeed, analyses of the spleen within 1 day indicated that OT-1 antigen-specific T cells were more frequently in contact with mRBC-OVA-4-1BBL-IL-12 than with mRBC-CTRL, as assessed by both colocalization (Supplementary Fig. 6a, b) and quantification of doublets by flow cytometry (Supplementary Fig. 6c). This also correlated with CD44 expression, confirming the activation of OT-1 cells (Supplementary Fig. 6d). We also demonstrated direct interactions between OT-1 cells and mRBC-OVA-4-1BBL-IL-12 by in vitro live cell imaging (Fig. 2f, Supplementary movie 1, and Supplementary movie 2). Together, these results indicate the ability of mRBC-OVA-4-1BBL-IL-12 to engage in cognate interactions with antigen-specific T cells and suggest that the spleen could be one of the primary locations for this in vivo.
Long-term memory formation and protection against tumor rechallenge. To investigate the effects of mRBC-aAPCs on immune memory, mice with EG7.OVA tumors that were previously cured following treatment with mRBC-OVA-4-1BBL-IL-12 were rechallenged with EG7.OVA 66 days after initial tumor injection. Age-matched naïve mice were transferred with OT-1 cells 1 day before EG7.OVA challenge to represent residual background levels of OT-1 cells (Fig. 3a). Four out of five naïve mice succumbed to EG7.OVA challenge (Fig. 3b), and OT-1 cells did not expand in these mice (Fig. 3d). In contrast, all seven mice cured of EG7.OVA tumors by prior mRBC-OVA-4-1BBL-IL-12 treatment rejected EG7.OVA rechallenge without additional treatment (Fig. 3b). This was associated with OT-1 and endogenous OVA-specific T-cell expansion 10 days after tumor rechallenge ( Fig. 3c-e) and demonstrates OVA-specific immunological memory formation following initial treatment with mRBC-OVA-4-1BBL-IL-12.
IL-12 has been shown to play a role in promoting epitope spreading 28 . To determine if protection against other tumor antigens in addition to OVA was established by mRBC-OVA-4-1BBL-IL-12 treatment, cured mice and age-matched naïve controls were challenged with the parental OVA-negative tumor cell line, EL4, 61 days after the second EG7.OVA challenge. Strikingly, previously cured mice showed either resistance to tumor growth (3/7) or delayed tumor growth (3/7) compared with control mice (Fig. 3f). This demonstrates that in the process of inducing an immune response to EG7.OVA, treatment with mRBC-OVA-4-1BBL-IL-12 led to epitope spreading to other tumor antigens.
Co-culture with RTX-321 for 2 or 24 h significantly increased Nur77 and CD69 expression, respectively, on E7-TCR cells, indicating productive ligation of the TCR (Fig. 6c, d). TCR signaling was independent of 4-1BBL and IL-12 expression as engineered RBCs expressing HLA-A2-HPV alone were able to induce these activation markers (Fig. 6c, d), as well as 4-1BB (Fig. 6e). While HLA-A2-HPV alone (signal 1) was sufficient for T-cell activation, all three signals were required for T-cell expansion (Fig. 6f), which was accompanied by induction of granzyme B and IFNγ secretion (Fig. 6g). Co-culture with RTX-321, but not RCT-CMV-4-1BBL-IL-12, led to the expansion of E7-TCR cells, which demonstrates that expansion is driven by antigen specificity. The CD8 + T cells that expanded in response to RTX-321 consisted of both T CM and T EM cells (Fig. 6h). The increase in T CM cells was predominantly dependent on signals 1 and 2, while all three signals contributed to the increase in T EM cells. These data demonstrate that RTX-321 promotes the expansion of primary CD8 + antigen-specific T cells, induces the production of effector molecules, and increases T-cell activation and the generation of T EM and T CM phenotypes in an antigen-specific manner.

Discussion
Here, we show that RBCs can be engineered into aAPCs that induce a tumor-specific immune response by expanding T cells against a target antigen in vivo. Of note, RTX-321 enabled the expansion of HPV-specific primary human T cells in vitro, highlighting the potential therapeutic application of this platform. This approach is designed to mimic human immunobiology through the cellular presentation of thousands of copies of biotherapeutic proteins on the cell surface of human enucleated RBCs, thus differing from other approaches that utilize synthetic receptors (as used in CAR-T cells), non-allogeneic cell lines for ex vivo T-cell expansion, or synthetic platforms (e.g. liposomes or biodegradable polymer particles) [30][31][32] .
In addition to a tumor antigen-loaded MHC I (signal 1), we used 4-1BBL as the T-cell costimulatory ligand (signal 2). Engagement of 4-1BB on the T cell by 4-1BBL promotes T-cell survival and proliferation, enhances effector function, and is critical for the formation of immunological memory 33 . In clinical trials of CAR-T cells, incorporation of a 4-1BB costimulatory domain favors development of central memory T cells and longer T-cell persistence compared with the use of a CD28 domain 33 .
We determined that the inclusion of IL-12 as the cytokine (signal 3) led to optimal effects by promoting antigen-specific Tcell expansion, memory and effector function, and tumor control. This is consistent with the known function of IL-12 as a potent proinflammatory cytokine that induces IFNγ release from NK cells as well as CD4 + and CD8 + T cells. Moreover, IL-12 signaling via STAT4 is essential for Th1 differentiation and the cytotoxic activity of CD8 + T cells 34 . We demonstrated T-cell expansion and activation that were associated with tumor regressions and cures in two murine tumor models. Our data show that mRBC-aAPCs can drive an antitumor response not only to the foreign antigen OVA, but also to a naturally expressed tumor antigen, gp100. mRBC-aAPC treatment provided superior survival benefits compared to vaccination with long peptide plus IFA (equivalent to Montanide™ ISA 51 in clinical trials; Fig. 1i). Long peptide plus IFA is the vaccination platform furthest along in clinical development for HPV 16+ cancers [35][36][37] for a single T-cell epitope and was therefore used as a comparator as opposed to other platforms, such as DNA vaccines, protein-based vaccines, and viral/bacterial vector vaccines. Our results suggest that the mRBC-aAPC format provides additional therapeutic benefits that vaccines and adjuvants may fail to provide. In addition, treatment with mRBC-aAPCs drove improved tumor control in comparison to anti-PD-1 treatment, which has been shown previously to have only a minimal effect in the B16-F10 tumor model [38][39][40] . Interestingly, intravenously administered mRBC-aAPCs appeared to be confined to the circulation and open vasculature of the spleen where interactions with antigen-specific T cells occurred (Supplementary Fig. 6), and proved effective at inducing an antitumor response. Indeed, recent data suggest that the spleen is an important source of cells that respond to immunotherapy 41 .
Importantly, cured mice in the EG7.OVA model were not only resistant to rechallenge with EG7.OVA tumor cells but also to challenge with the parental EL4 cell line lacking OVA, indicative of both memory formation and epitope spreading 42,43 ; this has been shown to improve outcomes in patients undergoing cancer immunotherapy 42 . While epitope spreading has been demonstrated previously with adoptive transfer of preactivated 44,45 or genetically modified T cells 28,46 , here, we find that a RBC engineered to be an artificial APC can promote epitope spreading in vivo with naïve T-cell transfer. This is a particularly promising finding in our study given that the use of a single tumor-antigen could be considered a potential limitation of our approach. For example, recent data from studies of CD19directed immunotherapies suggest that a proportion of patients whose disease relapses can be characterized by loss of CD19 from the tumor cell surface 47 . However, our data demonstrate that the surrogate mRBC-aAPCs not only drive the activation and expansion of antigen-specific T cells but also promote changes in the tumor microenvironment, including increases in endogenous T cells, reductions in Treg numbers, increases in proliferative Th1 cells, and increases in M1 macrophages ( Supplementary  Fig. 4). These changes may contribute to antigen spreading and the associated protection observed in animals rechallenged with tumor cells lacking the target antigen (Fig. 3f). In addition, broad immune activation (Supplementary Fig. 4) may contribute to improved efficacy over long peptide vaccination. Patients with HPV+ cancers have varying frequencies of HPV antigen-specific T cells in their circulation [48][49][50] . We modeled the range of antigen-specific T-cell frequency in patients through the presence or absence of OT-1 adoptive transfer. With a high frequency of antigen-specific T cells (i.e., with OT-1 adoptive transfer), an mRBC-aAPC that provided all three signals was more efficacious than an mRBC-aAPC lacking signal 1 ( Fig. 1i and Supplementary Fig. 3d, e). However, with a lower frequency of antigen-specific T cells (i.e., without OT-1 transfer), we observed similar efficacy with mRBC-OVA-4-1BBL-IL-12 and mRBC-4-1BBL-IL-12 (Fig. 4b). This indicates that the main drivers of the anti-tumor effects in the model without OT-1 transfer are the 4-1BBL and IL-12 components of the mRBC-aAPC. Although the provision of signal 1 led to an earlier expansion of target antigen-specific T cells, this did not translate into an additional effect on tumor growth. Thus, our data indicate that the relative importance of these signals in the clinic may be determined by the number of pre-existing target antigen-specific T cells in patients.
Despite existing treatment options, there is a high unmet need in the treatment of HPV + cancers, and survival of patients with recurrent disease remains poor 22 . This led us to develop an RCT-aAPC to target an HPV-specific antigen for potential therapeutic applications (RTX-321). The viral oncoprotein E7 of HPV is known to drive the development of numerous cancers through the disruption of pRB tumor suppressor activity 51 . The E7 [11][12][13][14][15][16][17][18][19] peptide from HPV16 was chosen for RTX-321 because it is from an invariant region of the oncoprotein 52 , demonstrated prior cytotoxic T-lymphocyte induction, and is presented by HLA-A*02:01, the most common HLA allele among the US population 53,54 . We showed that RTX-321 induced TCR signaling, led to the expansion of HPV antigen-specific T cells, promoted effector function, induced the expansion of an effector memory cell population, and established a repertoire of T CM cells, suggesting that RTX-321 treatment may lead to a sustained antitumor response 55 .
Our study presents a genetically engineered RBC-based aAPC platform for the treatment of cancer and shows the activity of this approach in primary human cells. RBCs have been used in transfusion medicine for decades and O-negative donor blood can be reliably transfused to most of the population. Allogeneic engineered RBCs could be dosed across multiple patients and produced using a scalable manufacturing process without the need for the complex, personalized production processes used in TCR-T-cell and CAR-T therapies [7][8][9] . The RCT-aAPC platform demonstrates broad antigen applicability that is designed to mimic the biology of T-cell-APC interactions for potential application as an immunotherapeutic in a range of cancers 19,56,57 . . Cells were washed and analyzed on a NovoCyte ® 3000 flow cytometer (Acea Biosciences ™ , Inc., CA, USA). Flow cytometry data were analyzed using FlowJo ™ software (BD Biosciences). All requests for materials will be promptly reviewed by Rubius Therapeutics to verify whether the request is subject to any intellectual property or confidentiality obligations. Following review, materials can be shared via a material transfer agreement.
EG7.OVA and EL4 subcutaneous tumor model. Pep Boy (CD45.1 B6) mice were injected subcutaneously in the flank with 2 × 10 6 EG7.OVA at a 1:1 ratio in RPMI 1640 media and Matrigel ® matrix (Corning). Tumor growth was measured every 2-3 days. Tumor volume was determined as length (mm) × width (mm 2 ) × 0.5. Tumor-bearing mice were randomized when the average tumor volume reached 150-230 mm 3 and naïve CD8 + OT-1 cells were administered to all groups no later than 1 day after randomization by tail vein injection. Several hours after T-cell injection, mice were treated intravenously with PBS (Thermo Fisher Scientific), mRBC-CTRL, or mRBC-aAPC in 200 μL PBS every 3-4 days. For comparison with therapeutic vaccines, a group of mice received 40 nM OVA 241-270 58 (AnaSpec ® , Inc., CA, USA) emulsified in IFA (Sigma-Aldrich). In studies without OT-1 transfer, mRBC-CTRL or mRBC-aAPC treatment started 1 day after subcutaneous injection of 2 × 10 6 EG7.OVA cells in C57BL/6 mice. For the tumor rechallenge experiment, cured mice were challenged in the alternate flank subcutaneously with 2 × 10 6 EG7.OVA cells at a 1:1 ratio in RPMI 1640 media and Matrigel matrix. Mice cured following rechallenge with EG7.OVA tumor were then challenged subcutaneously in the shoulder with 1 × 10 5 EL4 at a 1:1 ratio in RPMI 1640 media and Matrigel matrix. Mice were euthanized after a tumor volume greater than 2000 mm 3 was reached.
B16-F10 lung metastasis model. C57BL/6J mice were injected intravenously with 1 × 10 5 B16-F10 in 200 μL RPMI 1640 media on day 0. The next day, all groups were intravenously administered 2 × 10 6 pmel-1 cells, followed by intravenous administration of mRBC-CTRL or mRBC-gp100-4-1BBL-IL-12 on days 1, 4, and 8. One group of mice received 100 μg anti-PD-1 (InVivoMab™ anti-mouse PD-1 clone RMP1-14, Bio X Cell, Inc., NH, USA) intraperitoneally on days 1, 4, and 8 in addition to mRBC-CTRL treatment. On day 14, the animals were euthanized. The lungs were perfused with ice-cold PBS through the right atrium and removed, and the lobes were teased apart with small forceps in a Petri dish. The left lobe was processed immediately for immune cell phenotyping. The remaining four lobes were fixed in 10% buffered formalin (VWR International, LLC, PA, USA) for 24 h. Metastases were visualized and quantified under the microscope by an operator blinded to the treatments. All mice requiring euthanasia or found dead for reasons other than tumor burden were excluded from analysis.
Immune profiling in mice and flow cytometry. Single-cell suspensions were prepared from blood, spleens, lymph nodes, lungs, and tumors. Blood was collected by submandibular bleed or cardiac puncture in BD Microtainer ® blood collection tubes with K 2 EDTA (BD Biosciences) and the tubes were kept on ice until analysis. For flow cytometry, 50 μL of blood was mixed with 550 μL of Alfa Aesar ™ RBC Lysis Buffer for mouse RBCs (Thermo Fisher Scientific) in a 96-well deep well plate (VWR International, LLC) and incubated at room temperature for 15 min to lyse RBCs. All cells were transferred to a round bottom 96-well plate (Corning) after pelleting for staining. Tumors or lungs were digested in gentleMACS ™ C-tubes (Miltenyi Biotec, Inc.) using the murine Tumor Dissociation Kit or the murine Lung Dissociation Kit (Miltenyi Biotec, Inc.), respectively. Cell suspensions were generated using a gentleMACS Octo Dissociator with Heaters instrument (Miltenyi Biotec, Inc.) according to the manufacturer's instructions. Digested tissue was then filtered through a 70 µm cell strainer and washed and resuspended with PBS containing 0.1% bovine serum albumin (PBSA; Cytiva) for staining.
For flow cytometry, cells were resuspended in a staining mixture containing Fc block (1:100 from 1 mg/mL working stock, anti-mouse CD16/32, clone 2.4G2, Bio X Cell, Inc.), directly conjugated antibodies, and LIVE/DEAD Fixable Aqua Dead Cell Stain or Zombie NIR ™ dye (BioLegend) in PBSA and stained for 30 min on ice. For detecting endogenous OVA-specific T cells by tetramer staining, cells were incubated with H-2K b -SIINFEKL tetramer (1:20, MBL International Corporation) with Fc block at room temperature for 10 min before surface staining. Samples were washed twice with PBSA and resuspended in PBSA for flow cytometry. To measure intracellular expression, eBioscience™ Cell Stimulation Cocktail (Thermo Fisher Scientific), containing phorbol 12-myristate 13-acetate (PMA), ionomycin, brefeldin A, and monensin, was incubated with single-cell suspensions from tumors for 3-4 h at 37°C. After incubation, cells were washed, surface stained, fixed, and permeabilized using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific), before being stained for intracellular or intranuclear marker expression according to the manufacturer's instructions.
The following fluorochrome-conjugated antibodies were purchased from BioLegend or BD Biosciences TCR variable beta chain sequencing. Genomic DNA (gDNA) was extracted from 50-60 μL of anticoagulated blood 5-10 min after OT-1 transfer (pre-treatment or pre-rechallenge) or post treatment/tumor rechallenge using DNeasy ® Blood and Tissue Kit (Qiagen ® Inc., MD, USA) according to the manufacturer's instructions. Single-cell suspensions were extracted from whole tumors, similar to the tissue processing for flow cytometry described above, and gDNA was extracted from one quarter of the tumor single-cell suspensions using the DNeasy Blood and Tissue Kit (Qiagen Inc.) according to the manufacturer's instructions. Immunosequencing of the CDR3 regions of mouse TCRβ chains was performed using the immunoSEQ ® Assay (Adaptive Biotechnologies Corporation, Seattle, WA). In brief, extracted DNA was amplified in a bias-controlled multiplex polymerase chain reaction (PCR), followed by high-throughput sequencing. Sequences were then collapsed and filtered in order to identify and quantitate the absolute abundance of each unique TCRβ CDR3 region for further analysis with an Adaptive ImmunoSEQ analyzer v3.0 59-61 .
Biodistribution analyses. mRBC-CTRL or mRBC-OVA-4-1BBL-IL-12 cells were incubated with 10 μM CellTrace Far Red dye at room temperature for 6 min followed by quenching with FBS and extensive washing before dosing. Pep Boy (CD45.1 B6) mice were dosed with 1 × 10 9 CellTrace Far-Red dye-labeled mRBC-CTRL or mRBC-OVA-4-1BBL-IL-12. At 1 h or 17 h post mRBC-aAPC dosing, mice were euthanized and perfused with PBS. One quarter of the spleen, the whole mandibular lymph node (LN), one of the mesenteric LNs, the bottom half of the left lung lobe, a 5-mm 3 portion of liver, the bottom half of the heart, a 5-mm 3 portion of brain, a 5-mm 3 portion of large intestine, a whole kidney, an ovary, and the testes were frozen in Tissue-Tek ® Cryomold ® plastic molds (VWR International, LLC) containing Tissue-Tek O.C.T. Compound (VWR International, LLC) in a liquid nitrogen bath. Frozen tissues in O.C.T. Compound were stored at −80°C before preparation of 7 μM sections that were placed onto slides. Slides were fixed with 4% formalin (Sigma-Aldrich), washed, and incubated with Hoechst dye (Thermo Fisher Scientific). After washing, coverslips were mounted on slides with ProLong ™ Gold Antifade Mountant (Thermo Fisher Scientific) and scanned on a 3DHISTECH PANNORAMIC™ scanner v2.1.1.10094 RTM (Thermo Fisher Scientific). Images were analyzed for mRBC density per tissue area using HALO image analysis software V2.2.1870.44 with High Plex FL module V2.0. Samples that were not sectioned or stained properly were excluded from analysis and noted in the source data.
Spleen interaction analyses. CD45.1 Pep Boy mice were transferred with 2 × 10 6 naïve CellTrace Yellow dye-labeled OT-1 cells before dosing with 1 × 10 9 CellTrace Far Red dye-labeled mRBC-CTRL, or mRBC-OVA-4-1BBL-IL-12 cells. At 1 h or 17 h post mRBC dosing, mice were euthanized and perfused with PBS. Half of each spleen was processed into a single-cell suspension for flow cytometry analyses without RBC lysis to determine the amount of mRBC that formed doublets with OT-1 cells. One quarter of each spleen was frozen in Tissue-Tek O.C.T. Compound. Frozen tissues in Tissue-Tek O.C.T. Compound were stored at −80°C before preparation of 7 μM sections that were placed onto slides. Slides were fixed with 4% formalin, washed, blocked with Background Sniper blocking reagent (Biocare Medical, LLC, CA, USA), and stained with hamster anti-mouse-CD31 antibody (1:250, clone 2H8, MilliporeSigma, MA, USA). Slides were washed, stained with anti-hamster antibody conjugated with AF488 (1:125, Jackson ImmunoResearch Laboratories, Inc., PA, USA), and incubated with Hoechst dye (Thermo Fisher Scientific). After washing, coverslips were mounted on slides with ProLong™ Gold Antifade Mountant (Thermo Fisher Scientific) and scanned on an Aperio ™ ScanScope ® FL scanner with ScanScope Console v102.0.0.33 (Leica Microsystems, Inc., IL, USA). The percentage of OT-1 cells that colocalized with mRBCs was determined using the HALO image analysis software with Object Colocalization FL module V1.0. Confocal images of spleen sections were obtained using a Leica TCS SP8 STED 3X confocal microscope (Leica Microsystems, Inc.) and LAS X Life Science 3.5.5.19976 software (Leica Microsystems, Inc.).
TCR transduction of primary human T cells and co-culture assay with engineered RBCs. CD8 + T cells from human HLA-A*02:01-positive donors (Cellero ™ , MA, USA), collected under Advarra IRB approval and informed consent, were stimulated with Dynabeads human T-Activator CD3/CD28 beads (Thermo Fisher Scientific) at a 1:1 ratio in X-VIVO ™ 15 medium containing 100 IU/mL human IL-2 (PeproTech, NJ, USA). The next day, the beads were removed and activated CD8 + T cells were transduced with lentivirus encoding HPV E7 11-19 -specific TCR in X-VIVO 15 media containing 100 IU/mL human IL-2. Culture media was changed 1, 4, and 8 days after transduction. On day 10, 8 × 10 4 untransduced T cells or TCRengineered primary CD8 + T cells with~20% TCR expression were incubated with 3.2 × 10 5 RTX-321 or control RBCs in 200 μL X-VIVO 15 media without human IL-2. At 2 h, cells were harvested, surface stained, fixed, and permeabilized using Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific), and stained for Nur77 expression. On day 1, cells were treated with brefeldin A (Bio-Legend) for 4 h before harvest and were stained with a similar protocol for CD69 and granzyme B expression. On day 5, cells were harvested for flow cytometry analyses of HPV E7 [11][12][13][14][15][16][17][18][19]  Statistics. Graphs were made and statistical analyses were performed using Prism software v8.4.2 (GraphPad Software, Inc., CA, USA). Data were expressed as mean ± standard deviation. Analysis of two groups was performed by a two-tailed, unpaired Student's t test where applicable. For analyses of three or more groups, a one-way analysis of variance (ANOVA) test was performed with Dunnett's multiple comparison test at each time point compared to controls. Biodistribution significance was determined by one-way ANOVA with Tukey's multiple comparison test within each tissue type. T-cell expansion post tumor rechallenge was determined by one-way ANOVA with Dunnett's multiple comparison test compared to before challenge. Body weight and plasma cytokine levels over time were analyzed by repeated measure two-way ANOVA with Tukey's multiple comparisons test at each time point, and statistical significance was reported for each time point compared to similarly treated mRBC-CTRL. Tetramer+ CD8 + T cells in the blood over time were analyzed by repeated measure two-way ANOVA with Dunnett's multiple comparison test compared to mRBC-CTRL after log transformation of data points (data do not follow normal distribution as determined by Shapiro-Wilk test). Statistical differences in survival were determined by log rank (Mantel-Cox) analyses. P < 0.05 was considered statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The TCR sequencing data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE168826. Source data are available as a source data file. The authors declare that all data supporting the findings of this study are available within the paper and its supplementary information files or available from the authors upon request. All requests for materials will be promptly reviewed by Rubius Therapeutics to verify whether the request is subject to any intellectual property or confidentiality obligations. Following review, materials can be shared via a material transfer agreement by contacting the corresponding author. Source data are provided with this paper.