Unbiased chemokine receptor screening reveals similar efficacy of lymph node- and tumor-targeted T cell immunotherapy

Localization is a crucial prerequisite for immune cell function and solid tumors evade immune control by modulating immune cell infiltration into the tumor stroma. Immunosuppressive cells like regulatory T cells are attracted, while cytotoxic CD8+ T cells are excluded. Engineering CD8+ T cells with chemokine receptors is a potent strategy to turn this mechanism of directed immune cell recruitment against the tumor. Here, we utilized fluorescent tagging to track the migratory behavior of tumor-specific T cells engineered with a library of all murine chemokine receptors in vivo. We then asked whether chemokine receptor-mediated redirection of antigen-specific T cells into tumors or tumor-draining lymph nodes showed superior anti-tumoral activity. We found that both targeting approaches showed higher therapeutic efficacy than control T cells. However, multiple receptors conveying the same homing pattern did not augment infiltration. Instead, in the MC38 colon carcinoma model, anti-tumoral efficacy as well as lymph node vs. tumor-homing patterns were mostly driven by CCR4 and CCR6, respectively. Overall, our data, based on fluorescent receptor tagging, identify the tumor-draining lymph node and the tumor itself as viable targets for chemokine receptor-mediated enhancement of adoptive T cell therapy. Supplementary Information The online version contains supplementary material available at 10.1007/s00262-023-03472-w.


Introduction
Traditionally, CD8 + T cells are considered to require three signals for optimal functionality: engagement of the antigen receptor, co-stimulation and stimulation by cytokines. However, none of these signals induce protective immunity without correct localization of the T cell. 'Signal 0' for a T cell is therefore its active migration toward a priming location or target organ and its interacting cell partners such as antigen-presenting or target cells.
Solid tumors capitalize on the location dependency of anti-tumoral immune cell function by controlling the composition of the intra-tumoral milieu. Through immunoediting, tumors can evolve to secrete a chemokine profile that attracts immunosuppressive cells such as Treg [1][2][3] and immunosuppressive myeloid cells [3][4][5]. Conversely, chemokines that facilitate infiltration of anti-tumoral immune cells such as CD8 + T cells and NK cells are suppressed. A crucial chemokine receptor that theoretically can recruit anti-tumor immune cells into tumors is CXCR3 [6,7]. If solid tumors fail to suppress secretion of ligands for CXCR3 such as CXCL9 and CXCL10, this renders such tumors susceptible to immune checkpoint blockade [8]. If tumors successfully shut down expression of such anti-tumoral chemokines, immune cells are deprived of a path into the tumor.
Genetic engineering of T cells used for adoptive T cell therapy can overcome the crucial bottleneck of impaired tumor infiltration and even turn the tumor's weapons against itself. Expression of chemokine receptors that 1 3 match ligands secreted by the tumor such as CCL2, CCL22, CXCL16, CX3CL1, CXCL1 and CXCL8 can enable CD8 + T cells to enter the tumor [9][10][11][12][13]. Importantly, these chemokines are secreted by the tumor for a reason, namely because they are pro-tumorigenic [14]. Chemokines such as CCL2 and CXCL16 can have direct effects on tumor cells, promoting cell motility and invasion [15,16]. CCL2 and CXCL8 can promote angiogenesis by acting on intra-tumoral endothelial cells [17,18]. Furthermore, immunosuppressive cells are recruited into the tumor. CCL2 for example facilitates infiltration by immunosuppressive myeloid cells [3], whereas chemokines such as CCL22 promote recruitment of Treg cells [19]. Expression of the corresponding chemokine receptors in CD8 + T cells is particularly attractive because the tumor can only shut down expression of these chemokines at the cost of losing their tumorigenic function [13].
Despite the potential of approaches to genetically engineer T cells to home into the tumor, so far none has been approved for therapy in humans. A possible explanation for this is that one chemokine receptor alone is often not sufficient to induce homing into the tumor site. Another explanation could be that strategies that attempt to drive T cells directly into the tumor are sub-optimal. Since stem-like CD8 + T cells have recently been found to reside mostly in tumor-draining lymph nodes [20], it could be important to engineer T cells to promote homing into draining lymph nodes instead.
To answer such questions in an unbiased manner, we generated a library of all known murine chemokine receptors outfitted with one of four distinct fluorescent tags. After transduction of this fluorescently-tagged receptor library into tumor antigen-specific T cells, we assessed the homing pattern conveyed by each chemokine receptor. Next, we defined batches of tumor-homing receptors and receptors that directed T cells into tumor-draining lymph nodes and investigated if combinatorial transduction [21][22][23] with multiple chemokine receptors improved homing to the target organ. To our surprise, we found that while both tumor-and lymph node-homing batches enhanced tumor regression in the MC38 colon carcinoma model, transduction with multiple receptors did not enhance target organ infiltration. Instead, organ infiltration and therapeutic efficacy were largely dependent on transduction with CCR4 in the lymph node-homing batch and CCR6 in the tumor-homing batch. Interestingly, combination of these two receptors appeared to reduce treatment efficacy. Taken together, CCR4 and CCR6 provided T cells with a unique but likely not additive quality that rendered them superior over other chemokine receptors with similar homing characteristics.

Mice
Female C57BL/JOlaHSd of 6-14 weeks were purchased from Envigo. 8-20 weeks old SIINFEKL peptide-specific TCR-transgenic OT-1 mice (C57BL/6-Tg(TcraTcrb)1100Mjb/J) were originally obtained from The Jackson Laboratory and bred under specific pathogen-free conditions at our mouse facility at the Technical University of Munich. All animal experiments were approved by local authorities and performed in accordance with national guidelines.

RNA isolation and murine chemokine receptor cDNA amplification
Single cell suspensions from C57BL/JOlaHSd tissue samples (spleen, lymph node, Peyer's plaque) were resuspended in 1 mL TRI Reagent (Sigma Aldrich, Cat# 93289), briefly incubated and 200 µL Chloroform (Carl Roth, Cat# 3313) was added. Phase separation was performed at 12,000 g for 15 min at 4 °C in a tabletop centrifuge. The RNA precipitation was achieved by addition of 500 µL Isopropanol (Carl Roth, Cat# 9781) to the upper layer after phase separation. The RNA pellet was washed with 1 mL 70% ice cold Ethanol. Reverse transcription was performed with 5 µg of the isolated total RNA using the AffinityScript cDNA Synthesis Kit (Agilent, Cat# 200436).

Generation of retroviral fluorescently labeled murine chemokine receptor vectors
All fusion constructs of murine chemokine receptors and fluorescent proteins (FPs) were generated by overlap extension PCR and cloned into the retroviral pMP71 vector (a 1 3 kind gift from Wolfgang Uckert) for recombinant expression in target cells. For retrovirus production, Platinum-E packaging cells were transfected via calcium phosphate precipitation with retroviral vectors encoding for all known signaling murine chemokine receptors tagged by an individual fluorescent protein (GFP, YFP, T-Sapphire or BFP) via a translational skip motif (P2A). Virus-containing supernatants were collected 48 h after transfection and purified from remaining cells by centrifugation (1500 rpm, 4 °C, 7 min). Supernatants were stored at 4 °C and used within 4 weeks after harvest.

Subcutaneous inoculation of syngeneic tumors and growth monitoring
MC38-OVA and Panc-OVA cells were grown to a confluence of 90%. Medium was removed and cells were rinsed once with PBS prior to enzymatic detachment with trypsin. Digestion was stopped by addition of fresh cDMEM and cells were pelleted. Tumor cells were resuspended in an adequate volume of PBS and cell numbers were determined by trypan-blue staining and counting. Cell numbers were adjusted to 0.5-1 × 10 6 cells/100 µL. 100 µL of cell suspension was injected in the right flank of isofluorane anaesthetized C57BL/6JOlaHsd mice which had been subjected to total body irradiation (TBI) at 4.5 Gy 24 h prior to tumor implantation. Tumor sizes, defined as tumor area (length × width), were measured every 2-4 days using a digital caliper.

Cell sorting and adoptive transfer of chemokine receptor-engineered OT-1 T cells
48 h after transduction, cells were stained with respective antibodies for 30 min at 4 °C in the dark, propidium iodide was used for live/dead discrimination. For survival experiments, equal numbers of CD8 + FP + (FPs corresponding to chemokine receptor expression) PI − T cells were sorted. Sorting of chemokine receptor-engineered T cells was performed on a BD FACSAria III (Becton Dickinson) or MoFlo Astrios cell sorter (Beckman Coulter). T cells were sorted into tubes pre-loaded with FCS and cell numbers were adjusted. The resulting infusion product was injected i.p. into tumor-bearing C57BL/6JOlaHsd mice. For infiltration and phenotypic analyses, transduced T cells were not subjected to cell sorting but analyzed by flow cytometry to determine initial ratios of single or combinatorial chemokine receptor-engineered T cells.
For assessment of homing preferences, the number of T cells transduced with a fluorescently-tagged chemokine receptor was normalized to the number of untransduced T cells and then compared between the organ infiltrate (at day 4 after injection) and the infusion product (at day 0 before injection). If, e.g., 50 out of 500 cells (50/500 = 0.1), sampled from the infusion product, were CCR6 positive (day 0 before injection) and 300 out of 1000 transferred T cells (300/1000 = 0.3) were later found to be CCR6 positive in the tumor (day 4 after injection), the ratio day 4/day 0 would be 0.3/0.1 = 3.
Different groups received the same number of OT-1 T cells within each experiment. For experiments shown in Fig. 3A and B we transferred on average 400.000 cells, for

Flow cytometry
Spleens and lymph nodes were harvested and mashed through a 40 µm cell strainer to generate a single-cell suspension. Tumors were minced by scissors and digested in 1 mL RPMI 1640 supplemented with 1 mg/mL collagenase IV and 100 µg/mL DNase I (Sigma-Aldrich, respectively) for 1 h at 37 °C in a table top incubator. The tumor suspension was passed through a 100 µM mesh and rinsed with 10 mL PBS before pelleting and resuspending in 3 mL of 40% Percoll (Sigma-Aldrich) solution. Separation of the lymphocyte fraction was performed by centrifugation at 2600 rpm at room temperature for 20 min (acceleration 5 of 9, deceleration 0 of 9). After red blood cell lysis, cells from each tissue were incubated with mouse anti-CD16/32 for 15 min at 4 °C. Cells were washed with FACS buffer (PBS with 0.5% BSA and 2 mM EDTA) and stained with respective fluorophoreconjugated antibodies for 30 min at 4 °C in the dark. After washing, cells were resuspended in FACS buffer containing 1% paraformaldehyde and analyzed by flow cytometry (Cytoflex S or LX, Beckman Coulter). FlowJo software (FlowJo LLC) was used for analysis. Normalized mean fluorescent intensities (MFIs) were calculated by dividing the MFI of transduced cells by the MFI of co-transferred, untransduced cells in the respective organ.

Statistical analyses
Significances were calculated after testing for normal distribution using unpaired t-test. Significance of survival differences was calculated using Log-rank test.

Generation of a library of chemokine receptors
To comprehensively assess the homing profile conveyed to antigen-specific T cells by chemokine receptor engineering, we generated a library of all murine chemokine receptors. Since ligand binding can lead to rapid chemokine receptor internalization and to discern engineered from endogenous chemokine receptor expression, each of these receptors was linked via a P2A sequence to one of four fluorescent proteins: GFP, YFP, T-Sapphire or BFP. In absence of ligand binding, we exemplarily verified that fluorescent protein expression tightly correlated with expression of the respective chemokine receptor (Fig. 1A). Next, we transduced TCR-transgenic OT-1 T cells specific to the SIINFEKL peptide of chicken Ovalbumin (ova) with one chemokine receptor per population. To make assessment of homing patterns more efficient, we used distinct fluorescent tags to assemble up to four populations expressing distinct chemokine receptors and one untransduced population that served as an internal control. This mixed cohort was then transferred into mice bearing subcutaneous MC38-ova (colon carcinoma) or Panc-ova (pancreatic adenocarcinoma) tumors (Fig. 1B). On day 4 post injection, the relative abundance of fluorescentlylabeled versus untransduced T cells was measured in tumor, blood, spleen, contralateral and draining lymph nodes and then compared to the relative abundance of fluorescentlylabeled versus untransduced T cells in the infusion product (ratio day 4 / day 0; see Methods). Calculating this ratio of day 4 infiltration versus day 0 infusion, revealed that defined chemokine receptors were associated with unique homing patterns. For example, in mice inoculated with either MC38ova or Panc-ova, CCR7 transduced T cells were overrepresented in lymph nodes, while CXCR4 transduced T cells were overrepresented in spleen ( Fig. 1C and Supp. Figure 1).

Distinct homing patterns of chemokine receptors in vivo
To identify the best targets for increased tumor homing, we compared the abundance of each chemokine receptor-engineered T cell population in the tumor at day 4 post injection to its abundance in the infusion product ( Fig. 2A and Supp. Figure 2A). The top targets for homing into MC38ova tumors were CCR6, CXCR1, CXCR2 and CX3CR1. Of these, CCR6 and CXCR1 also showed increased infiltration in Panc-ova tumors (Supp. Figure 2A). Since CXCR1 and CXCR2 bind the same targets, we picked only one, namely CXCR1 for further experiments. To answer whether homing into tumor or draining lymph node was more important for anti-tumor function of engineered T cells, we assessed all chemokine receptors regarding their potential to increase immediate homing into lymph nodes ( Fig. 2B and Supp. Figure 2B). To particularly select for chemokine receptors that promote homing into the draining lymph nodes and not lymph nodes in general, we furthermore calculated normalized ratios between enrichment into draining lymph nodes and contralateral lymph nodes or tumor, respectively (Fig. 2C). Here, we selected CCR4 and CCR7, which showed the highest ratio of draining to contralateral lymph node enrichment and a particularly high specific enrichment into draining lymph node over the tumor.

Targeting tumor-specific T cells into tumor-draining lymph nodes or the tumor itself are both successful therapeutic strategies
Next, we used combinatorial transduction to transduce the most promising chemokine receptors in two batches: One batch comprised CCR6, CXCR1 and CX3CR1-all chemokine receptors facilitating homing into MC38-ova tumors. The other batch contained CCR4 and CCR7, which conveyed T cell enrichment in the tumor-draining lymph nodes. In contrast to mice receiving T cells transduced with constructs encoding fluorescent proteins only (color-transduced), both chemokine-receptor-engineered T cell batches showed significant treatment efficacy in relation to untreated mice (Fig. 3A). Next, we analyzed distribution and phenotype of T cells engineered with the two batches at different time points post injection. As expected, the tumor-directed batch led to better T cell accumulation in the tumor at day 4 post injection, whereas the lymph-node homing batch led to relatively more T cell migration to the tumor-draining lymph node (Fig. 3B). However, regarding tumor infiltration, on day 8 post injection the lymph node batch caught up with the tumor batch. Comparing markers for activation and exhaustion, we found that the tumor-homing batch showed higher PD-1 expression early in the tumor, potentially as a sign for immediate antigen encounter (Fig. 3C). The lymph node batch in turn showed higher expression of stemness markers in the draining lymph nodes and Slamf6 in the tumor side (Fig. 3D). Together, it appeared that the tumor-targeting

Expression of two or more chemokine receptors is not superior over expression of one high-quality receptor
We hypothesized that by expressing more than one receptor that improves homing into the respective target site, migration could be improved through increased signal strength and specificity. To this end, we separated single and multiple transduced OT-1 T cells on days 4, 8 and 14 ( Fig. 4A and B). To our surprise, we could not observe a clear synergistic pattern in the tumor batch (Fig. 4A). Instead, CCR6 alone was as efficient or even more efficient in directing T cells into the tumor as was dual transduction with CCR6 and CXCR1 or CCR6 and CX3CR1. Tripletransduced T cells could not be detected in sufficient numbers throughout, although we initially hypothesized that such T cells might outcompete single-transduced T cells over time. For the lymph node batch, the two chemokine receptors individually appeared to have very different homing patterns over time. In the tumor, CCR4 was detectable in numbers that were as low as CCR7 or double transduced T cells on day 4 but then rapidly increased in numbers and persisted better toward day 14. In the draining lymph node, CCR7 transduced T cells were more abundant on day 4 but then slowly decreased over time, whereas CCR4 transduced T cells peaked at day 8, both

CCR4 and CCR6 enhance adoptive T cell therapy in a non-synergistic manner
Due to their distinct phenotypes, we chose to investigate the therapeutic potential of CCR4 and CCR7 individually (Fig. 4C). While CCR4-transduced T cells showed a therapeutic efficacy, surprisingly, CCR7 did not. For CCR4, we could previously show that exogenous expression facilitates interaction of CD8 + T cells with dendritic cells [13]. Since cDC1 express ligands for CCR4 [27], but not CCR7, we hypothesized that the selective treatment advantage of CCR4 vs. CCR7 despite highly similar homing patterns could be attributed to this mechanism. PD-1 is best known as an exhaustion marker, but is also induced by antigen receptor signaling [28]. Therefore, early during immune responses, PD-1 is an activation marker that is upregulated upon effective antigen recognition. We wondered, if T cells transduced with different chemokine receptors differ in their expression of PD-1 early during their anti-tumoral response. Indeed, CCR4 transduced T cells expressed much higher PD-1 levels in the draining lymph nodes, than did T cells transduced with CCR7 on day 4 post injection (Fig. 4D). This suggested that indeed CCR4 enhanced, whereas CCR7 potentially even impaired antigen recognition in draining lymph nodes, which might explain why the treatment efficacy of the lymph node batch appeared to be largely conferred by CCR4. Lastly, because CCR4 and CCR6 both conferred increased 1 3 therapeutic efficacies, we tested whether co-transduction of these two chemokine receptors improves anti-tumor efficacy (Fig. 4E). However, we could not detect a synergistic effect.
Single-chemokine receptor transduced T cells with CCR4 or CCR6 both led to longer median survival than CCR4/CCR6 double-transduced T cells. Since CCR6 transduced T cells showed increased PD-1 expression in the tumor and CCR4 transduced T cells in draining lymph nodes (Fig. 4D), one possible reason why this combination approach did not work was that CCR4/CCR6 double-transduced T cells become exhausted by additive PD-1 upregulation via CCR4 and CCR6. Indeed, when analyzing PD-1 expression in draining lymph nodes late after infusion, we found CCR4/CCR6 double-transduced T cells to express significantly higher levels of PD-1 than CCR6 transduced T cells (Supp. Figure 3B). Combining data from all conducted experiments, CCR6transduced T cells significantly and CCR4-transduced T cells almost significantly (p = 0.0505) prolonged survival compared to control-transduced T cells (median survival of 47 and 44.5 vs. 33 days, respectively). Thus, it appears that both tumor-and tumor-draining lymph node targeting of T cells can improve therapeutic efficacy of adoptive T cell therapy. However, at least for lymph node targeting, it appears that additional factors beyond simple organ enrichment must be considered in order to identify a therapeutically active chemokine receptor.

Discussion
Lack of migration of effector immune cells into solid tumors is a crucial bottleneck for adoptive T cell therapy. Engineering T cells with chemokine receptors can overcome this limitation. However, so far, no successful strategy using chemokine receptor transduced T cells has been approved for therapy in humans.
In this study, we characterized different homing patterns for each chemokine receptor. Over time the organ distribution of engineered T cells will be influenced by factors other than homing, such as proliferation. Therefore, we assessed homing patterns at day 4 post injection, an early time point at which we believe organ distribution will largely be influenced by homing. We identified chemokine receptors that improved homing into tumor or tumor-draining lymph nodes. So far, pre-clinical studies have largely focused on chemokine receptors that increase homing directly into the tumor. However, one could imagine that migration into tumor-draining lymph nodes, the side with most cells with a stem-cell like phenotype [20], could confer better anti-tumor efficacy. In our screen, we found two chemokine receptors to very specifically direct T cells into tumor-draining lymph nodes: CCR4 and CCR7.
At first sight, it is surprising that CCR7 expression leads to preferential recruitment to the draining lymph node, as the ligands CCL19 and CCL21 are thought to show homeostatic expression in all lymph nodes [29]. CD8 + T cells endogenously express chemokine receptors that lead to homing into activated lymph nodes such as CXCR3 [14,30]. Thus, we suspect that expression of CCR7 might increase the dwell time in lymph nodes in general, but additional, endogenously expressed receptors lead to preferential accumulation in tumor-draining lymph nodes.
Despite remarkably similar homing patterns, only CCR4transduced T cells were responsible for the observed treatment effect of the lymph node batch. We hypothesized that it might be crucial which cell type expresses the attracting chemokine in the lymph node. CCR4 ligands are predominantly expressed by dendritic cells [27], while CCR7 ligands are expressed by stromal cells [31]. Fitting this hypothesis, CCR4 and CCR7 transduced T cells showed clearly distinct expression of PD-1 on day 4 post injection. This early after injection, high PD-1 expression is unlikely to be a sign of exhaustion, but rather suggests recent antigen receptor signaling [32]. While CCR4 showed the most significant upregulation of PD-1 of all tested chemokine receptors in draining lymph nodes, CCR7 showed lower than average expression of PD-1. In our screen, all chemokine receptors showing increased PD-1 expression on day 4 post infection either match chemokines secreted by dendritic cells (CCR5, CXCR3) [30,33] or respond to chemokines that also attract dendritic cells (XCR1). The ligands for CCR7, however, are predominantly expressed by fibroblastic reticular cells in lymph nodes [31]. Thus, once in the lymph node, CCL19 and CCL21 might, in fact, lead T cells away from antigenpresenting cells, while CCR4 enhances interaction with dendritic cells in part via increasing the affinity of LFA-1 [13]. Significances in D were calculated as t-test against the fixed value 1. * = p < 0.05, ** = p < 0.01, *** = p < 0.001, 50.000 (4C), 100.000 (4E) and 65.000 (4F) OT-1 T cells were transferred on average ◂ Concerning the tumor-homing batch, we asked whether transduction with multiple chemokine receptors that direct T cells into the tumor might increase functionality. However, we could not observe such a synergistic effect. Instead, overexpression of CCR6 in the MC38-ova model was sufficient and even superior over T cells co-transduced with other tumor-homing receptors. This might be due to various reasons. Potentially, transduction with multiple receptors decreases fitness of the transduced T cells. In this case, generating strategies that rely on expression of receptors via alternative delivery methods, such as CRISPR/Cas9 editing might improve efficacy [34,35]. Alternatively, co-expression of two or more chemokine receptors could lead to immediate immigration at the expense of maintaining a pool of stemlike cells in the secondary lymphoid organs, which could lead to decreased persistence over time.
Last, we assessed whether a combination of the best tumor-and lymph node-homing chemokine receptors might show a synergistic effect. However, co-transduction of CCR4 and CCR6 was not superior over single transductions, again suggesting that engineering of T cells with one high-quality receptor provides better efficacy than co-transductions. An additional reason for decreased functionality of multiple-transduced cells may be a synergistic PD-1 induction dependent on different chemokine receptors: CCR6 showed increased PD-1 expression in the tumor and CCR4 in draining lymph nodes. We found CCR4/CCR6 doubletransduced T cells to express higher levels of PD-1 than CCR6-transduced T cells in endpoint analyses, suggesting that double-transduced T cells might have an increased risk of undergoing T cell exhaustion.
Our study is the first to assess the potential of all murine chemokine receptors to influence homing of CD8 + T cells in tumor-bearing mice. Our data indicate that equipment of T cells with one high-quality chemokine receptor is sufficient for enhancing the therapeutic potential of adoptively transferred T cells into solid tumors. Transduction with CCR6, a chemokine receptor known to recruit immunosuppressive myeloid cells and Treg into the tumor [36], showed the best efficacy. The efficacy of CCR6-transduced T cells for immunotherapy has only been studied in xenograft models, where the tumor xenograft is the only source of CCR6 ligands [37,38]. Furthermore, we could show that directing T cells into the tumor-draining lymph node via CCR4 can enhance therapeutic efficacy. However, the success of directing tumor-reactive T cells into draining lymph nodes likely depended on the cell type releasing the attracting chemokine, as CCR7-transduced T cells showed similar homing but no therapeutic effect.