Relapse or eradication of cancer is predicted by peptide-MHC affinity

Summary

Cancers often relapse after adoptive therapy, even though specific T cells kill cells from the same cancer efficiently in vitro. We found that tumor eradication by T cells required high affinities of the targeted peptides for MHC class I. Affinities of at least 10 nM were required for relapse-free regression. Only high-affinity peptide-MHC interactions led to efficient cross-presentation of antigen, thereby stimulating cognate T cells to secrete cytokines. These findings highlight the importance of targeting peptides with high affinity for MHC class I when designing T cell-based immunotherapy.

Introduction

Relapse of cancers is very common even following combinatorial therapy of surgery, chemotherapy, radiation, and/or immunotherapy. For maximal efficacy, drugs depend on reaching the necessary concentration in the tumor microenvironment (Skipper, 1986). This critical concentration concept also applies to cellular effectors such as neutrophils and T cells (Budhu et al., 2010; Li et al., 2002; Li et al., 2004). While cellular effectors or drugs at optimal concentrations can eradicate all sensitive cancer cells, relapse may still occur because of the outgrowth of variants. Cancer cells show extremely high genetic instability and cancers always contain variants that are resistant to destruction by a particular drug or T cell (Anders et al., 2011; Hanson et al., 2000), very similar to what is found for viruses (Hensley et al., 2009).

For complete eradication, it is important to eliminate every residual cancer cell including heritable variants (Singh et al., 1992; Spiotto et al., 2004; Zhang et al., 2007). However, factors responsible for T cell elimination of variants have not been determined. In experiments designed to explore the reason for failed T cell treatment, we took a reductionist approach, ultimately directing our focus to the target peptides, and in particular to their affinities for MHC class I. We selected several peptides that, when targeted, caused tumor eradication, and others that caused relapse. To reduce the influence of differences between cancers, we used two cancer cell lines that were both transduced to express the different peptides. To reduce differences due to expression levels, we used the same design of triple peptides fused to fluorescent proteins. Proteasomal cleavage of proteins may not generate (Chapiro et al., 2006; Popovic et al., 2011) or destroy immunogenic peptides (Schultz et al., 2002). To minimize differences in proteasomal cleavage of the fusion proteins, we designed peptide triplets separated by “Ala-Ala-Tyr” cleavage sites. We targeted antigens with no known oncogenic activity to reduce the possibility that the nature of a particular targeted antigen prevented the cancer from escaping. To exclude the influence of other T cells helping or regulating the relevant CD8⁺ T cells, TCR-transgenic T cells with a single specificity were adoptively transferred into hosts, which were TCR-transgenic for an irrelevant target. Finally, a single adoptive T cell transfer regimen was used, without providing any additional stimulation such as vaccination or administration of cytokine.

Results

Cancer cells expressing different peptides are killed by T cells with similar efficacy in vitro

EGFP was fused to minigenes encoding the peptides OVA₂₅₇, SIY, mouse Tyrosinase_369–377 (Tyr₃₆₉), mouse or human gp100_25–33 (mgp100₂₅ and hgp100₂₅, respectively) and EGP; EGP differs from mgp100₂₅ only in the third amino acid (EGPRNQDWL versus EGSRNQDWL), while it shares the proline at position 3 with hgp100₂₅ (KVPRNQDWL). A Cerulean fusion gene was generated only for SIY (Figure 1A and C, top). The fibrosarcoma line MC57 of C57BL/6 origin was used to generate lines that expressed the fusion genes at high levels (Figure 1B). Furthermore, the HHD MHC class I molecule was co-transduced with the Tyr₃₆₉-EGFP fusion protein to generate MC57-TyrHHD (Figure 1C, bottom).

Figure 1. Six transduced cancer cell lines that express antigens at high levels were effectively killed in vitro.

(A) Diagram of fusion proteins constructed to express antigen in MC57 cancer cells. Triple repeats of peptide and AAY proteasomal cleavage sites were fused to fluorescent proteins: OVA₂₅₇, SIY, mouse tyrosinase_369–377 (Tyr₃₆₉), mouse or human gp100_25–33 (mgp100₂₅ and hgp100₂₅, respectively) and EGP; a Cerulean fusion gene was generated only for SIY. (B) Flow cytometric analysis of peptide-EGFP fusion proteins expressed by the transduced MC57 fibrosarcoma lines. (C) MC57-mgp100/SIY expressed mgp100₂₅ and SIY antigens as EGFP and Cerulean fusion proteins, respectively. The HLA-A2/D^b chimeric protein HHD was co-transduced with the Tyr₃₆₉-EGFP fusion protein to generate MC57-TyrHHD. Parental MC57 (gray) was analyzed for comparison. (D) Cytolysis of MC57 target cells overexpressing SIY, mgp100₂₅, hgp100₂₅, EGP, or Tyr₃₆₉ and HHD (Tyr) by 2C, pmel, AFH (Tyr-negative) or FH (Tyr-positive) T cells in a 4.5 h ⁵¹Cr-release assay. Cancer cells expressing non-cognate peptide were used as negative controls. These data are compiled of three experiments and are representative for seven independent experiments. See Figure S1 for induction of vitiligo by FH and pmel T cells.

Assays in vitro demonstrated similar killing of the cancer lines by cognate peptide-activated T cells (Figure 1D). 2C T cells, whose TCR binds SIY, killed the MC57-SIY line, and pmel T cells killed MC57 cells expressing mouse gp100₂₅, human gp100₂₅ or EGP. Interestingly, Tyr₃₆₉-specific T cells derived from the FH TCR-transgenic, tyrosinase (Tyr)-deficient albino mouse (AFH) or Tyr-positive black mouse (FH) killed MC57-TyrHHD target cells similarly well. Together, the results imply there is sufficient direct presentation of all processed peptides and sufficient avidity of the T cells for efficient killing in vitro.

T cells targeting SIY, OVA₂₅₇ or Tyr₃₆₉ eradicate large tumors

SIY-expressing MC57-SIY cells were injected in TCR-transgenic mice of irrelevant specificity (OT-I). OVA₂₅₇-transfected cancer cells were injected in 2C TCR-transgenic mice; MC57-TyrHHD cancer cells were grown in OT-I TCR- and AAD-transgenic mice, which did (OTA) or did not express tyrosinase (albino, AOTA). In all cases, cancer cells produced progressively growing tumors within one week (Figure 2A). At least 2 weeks after cancer cell injection, when tumors reached about 500 mm³, mice were treated with T cells. As published by our laboratory, tumors expressing SIY and treated with 2C T cells were eradicated (Figure 2B upper left, Table 1 and Table S1) (Spiotto et al., 2004). Here we also show that OVA₂₅₇-expressing tumors treated with OT-I T cells were rejected (Figure 2B middle left) and FH T cells eradicated Tyr-positive tumors (Figure 2B lower left). In this last experiment, FH T cells derived from a Tyr-positive donor were transferred into a Tyr-positive host and eradicated a Tyr-expressing tumor. Together, this and other experiments using FH TCR-transgenic T cells from Tyr-negative donors (AFH) and/or Tyr-negative hosts (AOTA) showed that tumors could be rejected (i) whether the targeted peptide was self or non-self for the tumor-bearing host and (ii) whether the targeted peptide was self or non-self for the donor T cells (Figure S2A). This may be unique to our model, since a different model showed that low ubiquitous expression of a transgene prevented the rejection of antigen-expressing tumors through the induction of tolerance (Buschow et al., 2010). Levels of antigen expression in the host and/or tumor, type of cells that express the self-antigen and the source of T cells may likely influence the outcome. Taken together, targeting any of the three peptides SIY, OVA₂₅₇ or Tyr₃₆₉ caused eradication of established, large solid tumors.

Figure 2. Targeting SIY, OVA₂₅₇ or Tyr₃₆₉ eradicated large tumors while targeting mgp100₂₅, hgp100₂₅ or EGP caused initial tumor regression but was followed by relapse.

(A) Cancer cell lines formed progressively growing tumors within one week. TCR-transgenic mice with irrelevant specificity were challenged s.c. with 2×10⁶ cancer cells. MC57-SIY, -mgp100, -mgp100/SIY, -hgp100 and -EGP were injected into OT-I mice; MC57-OVA grew in 2C mice and MC57-TyrHHD grew in AOTA (Tyr-deficient, non-self) and OTA (Tyr-positive, self). Graphs represent single mice in 11 experiments, listed as non-treated controls in Tables 1, S1 and S3. (B) At least 2 weeks after cancer cell injection, when tumors reached about 500 mm³, each mouse was treated once with cognate T cells (treatment between days 13 and 26, depending on tumor size, as indicated by the horizontal bars (H)). Average size of tumors at day of treatment: 2C: 720 mm³, ranging from 448 to 995 mm³; OT-I: 608 mm³, ranging from 440 to 715 mm³; FH: 517 mm³, ranging from 250 to 848 mm³; pmel, targeting mgp100₂₅: 601 mm³, ranging from 325 to 980 mm³; targeting hgp100₂₅: 470 mm³, ranging from 264 to 936 mm³; targeting EGP: 337 mm³, ranging from 180 to 600 mm³. MC57-TyrHHD was grown in OTA (self) and treated with FH T cells (self). The number of rejected tumors per number of tumors treated is indicated. Data are derived from 15 independent experiments, compiled in Tables 1 and S1. See also Figure S2.

* Tumors were isolated and analyzed for antigen expression (see Figure 3).

Table 1.

Abbreviations, conditions and summary of results for key experiments.^A

Target peptide on cancer cells				Hosts		T cells		Tumor rejection
Desig- nation	Sequence	MHC	Affinity of peptide for MHC (IC50 [nM])B	Desig- nation	Relationship of antigen to recipient	Desig- nation	Relationship of antigen to donor
SIY	SIYRYYGL	Kb	1.1	OT-I	non-self	2C	non-self	5/5C,D,E,F
						none		0/6C
OVA257–264	SIINFEKL		0.9	2C	non-self	OT-I	non-self	4/4G
						none		0/4G
Tyr369–377	FMDGTMSQV	A2	4.2H	OTA	self	FH	self	6/7I
						none		0/5I
hgp10025–33	KVPRNQDWLJ	Db	186	OT-I	non-self	pmel	non-self	1/8D
						none		0/2
EGP	EGPRNQDWL		454	OT-I	non-self	pmel	non-self	1/6E
						none		0/5
mgp10025–33	EGSRNQDWL		22,975	OT-I	self	pmel	self	1/12F
						none		0/6

^ASee Table S1 for details.

^BIC₅₀ values represent the geometric mean of 5 or more experiments.

^Cp = 0.002;

^Dp < 0.005;

^Ep = 0.015;

^Fp < 0.001;

^Gp < 0.029;

^HA higher IC₅₀ value of 65 nM was published for this peptide earlier (Colella et al., 2000). The differences in affinity measurements likely arose as a result of small differences in reagents, methodology, and procedures.

^Ip = 0.015

^Jonly the underlined amino acids differ between the three gp100 peptide variants

Large mgp100₂₅, hgp100₂₅, or EGP-expressing tumors relapse after initial regression caused by transferred T cells

In contrast, T cells targeting the self-peptide mgp100₂₅ and the non-self heteroclitic peptides hgp100₂₅ or EGP, did not result in tumor eradication. MC57 lines overexpressing mgp100₂₅, hgp100₂₅ or EGP were injected into OT-I TCR-transgenic mice and produced progressively growing tumors within one week (Figure 2A). At least 2 weeks after cancer cell injection, when the tumors reached about 500 mm³, the mice were treated with pmel T cells. The tumors regressed initially, but eventually almost all tumors relapsed (Figure 2B right panels, Table 1 and Table S1).

To exclude any non-antigenic differences in the cancer lines (caused by transduction and sorting), we used a cell line that expressed both SIY and mgp100₂₅ antigens (MC57-mgp100/SIY, Figure 1C). When mice bearing these tumors were treated with 2C or pmel T cells, the outcome was the same as when tumors from single antigen lines were treated (Figure 2B upper panels). In conclusion, neither human nor mouse gp100₂₅ expressed by the cancer cells supported rejection by pmel T cells.

These findings were not limited to the MCA-induced cancer line MC57 but were confirmed using the UV-induced cancer line 8101 (Figure S2A and D). The line was transduced to overexpress SIY, human or mouse gp100₂₅. Again, we observed eradication of established tumors by adoptive T cell transfer only when SIY was targeted. Interestingly, in this model, targeting hgp100₂₅ was more effective than targeting mgp100₂₅; tumors expressing hgp100₂₅ regressed after pmel transfer, while tumors expressing mgp100₂₅ continued to grow uninhibitedly.

Treatment of tumors expressing human gp100₂₅ but not murine gp100₂₅ or EGP results in outgrowth of antigen-loss variants

We isolated cancer cells from tumors expressing mgp100₂₅, hgp100₂₅ or EGP that had relapsed following treatment with pmel T cells (Figure 2B) and analyzed these for antigen-loss variants (ALV). All MC57-hgp100 tumors had lost EGFP expression, which indicated loss of hgp100₂₅, as both were expressed as a single fusion protein (one representative tumor shown in Figure 3). Importantly, the tumor isolated from a non-treated mouse retained EGFP expression. MC57-mgp100 and MC57-EGP tumors treated with pmel had also not lost EGFP expression. All lines expressed mgp100-EGFP or EGP-EGFP at levels similar to the isolate from a non-treated mouse (Figure 3). These data suggest that pmel T cells were capable of killing all hgp100₂₅-expressing MC57 cancer cells but were not capable of killing all mgp100₂₅- or EGP-expressing cancer cells in the respective tumors. These findings seem to be influenced also by the targeted cancer cell, as relapsed tumors formed by 8101-hgp100 cancer cells all retained expression of the antigen (data not shown).

Figure 3. Outgrowth of antigen-loss variants after pmel T cell treatment of cancer cells expressing hgp100₂₅ but not of cancers expressing mgp100₂₅ or EGP.

Cancer cells of relapsed tumors expressing mgp100₂₅, EGP (both gray) or hgp100₂₅ (black) were isolated after pmel T cell treatment, adapted to culture, and analyzed for peptide-EGFP fusion gene expression (left panels). MC57 cells (white histogram) cultured in vitro and MC57-mgp100, MC57-EGP (both gray) or MC57-hgp100 (black) cells isolated from non-treated mice (right panels) were analyzed as controls. Isolated lines from mgp100- or hgp100-expressing tumors are representative for four lines each, the isolate from the EGP-expressing tumor is representative for two lines; all lines were isolated after relapse (respective tumors were marked with * in Figure 2B). The repeatability in independent experiments strongly suggests that loss of antigen expression from the hgp100₂₅ cancer cells was not an artifact caused by adaptation or post-isolation culturing. See also Figure S3.

While we did not observe significant differences when targeting either human or mouse gp100₂₅ in treatments of established tumors, we did see differences in protection against cancer cell inoculations. Pmel T cells prevented the outgrowth of MC57-hgp100 but not of MC57-mgp100 tumors (Figure S3A and C). MC57-mgp100 cells formed tumors in which a large fraction of cells still expressed the antigen (Figure S3B). Taken together, pmel T cells showed a stronger effect when targeting hgp100₂₅ compared to mgp100₂₅ and EGP.

Tumor eradication correlates with high affinity of targeted peptides for MHC

In an effort to understand why targeting some peptides led to eradication while targeting others resulted in relapse, we first analyzed the activation status of the T cells transferred to treat the different tumors (Figure 4). Upon transfer, after peptide stimulation in vitro, all T cells showed the same CD44^hi and CD62L^hi phenotype of activated T cells (Figure 4) and demonstrated very similar killing capabilities in vitro (Figure 1). It is worth mentioning that splenocytes from self-reactive TCR-transgenic mice (pmel and AFH) showed an antigen-experienced phenotype (CD44^hi), while T cells from the non-self reactive TCR-transgenic 2C mice showed a truly naïve phenotype. However, this difference was overcome after peptide stimulation in vitro. Interestingly, when isolated from tumors four days after adoptive transfer, the T cells that led to eradication of tumors (2C and AFH) showed a more effector-like phenotype compared (CD62L^lo, CD44^hi) to the more central memory-like phenotype (CD62L^hi, CD44^hi) found for pmel T cells. Together, these data suggest that the differences in tumor rejection were not due to differing activation statuses of the T cells at the time of transfer.

Figure 4. T cells transferred to treat the tumors expressing the different peptides showed the same phenotype of activated T cells.

T cells of 2C AFH and pmel TCR-transgenic mice were tested for their activation status in “naïve”, untreated mice (splenocytes), at day 3 after peptide activation in vitro, and on day 4 after adoptive transfer (tumor infiltrating cells). Cells were analyzed by flow cytometry for expression of CD44 and CD62L, and gated on CD8⁺ T cells expressing the cognate Vβ-chain: Vβ8 for 2C, Vβ11 for AFH and Vβ13 for pmel. Data are representative for two or three independent experiments for the data in vivo and in vitro data, respectively.

As another variable that could influence the efficacy of tumor rejection, we analyzed the affinities of the peptides for the presenting MHC molecules. In a cell-free competition binding assay, the concentration of inhibitor peptides needed to displace half of the probe peptide (IC₅₀ in nanomolar (nM)) was determined. IC₅₀ values are reasonable approximations of real K_D values (see Methods). A wide range in binding affinities was measured (Table 1). There was a strong correlation between affinity of the peptide for MHC and tumor eradication. The three target peptides supporting tumor eradication, OVA₂₅₇, SIY, and Tyr₃₆₉, displayed strong binding to their cognate MHC (0.9, 1.1 and 4.2 nM, respectively). These high affinities stood in stark contrast to the affinities we measured for mgp100₂₅, EGP and hgp100₂₅ (22,975, 454 and 186 nM, respectively). These three peptides bound the MHC poorly and when targeted resulted in relapse rather than tumor eradication.

Stromal cells isolated from tumors formed by cancer cells expressing peptides with high affinity to MHC stimulate cognate T cells effectively

To analyze why high peptide-MHC affinities were required for tumor eradication, we performed assays to evaluate the level of cross-presentation in growing tumors. MC57 and 8101 lines expressing SIY, mgp100₂₅ or hgp100₂₅ were grown in OT-I TCR-transgenic or Rag1^−/− mice, respectively; MC57-TyrHHD was grown in AOTA (non-self) mice. SIY was used as a representative peptide for the two highest binding peptides OVA₂₅₇ and SIY, and only the relatively best and worst binding gp100 peptides (human and mouse gp100) were analyzed in comparison. Enriched populations of CD11b⁺ stromal cells were obtained from at least 2-week-old untreated tumors and were compared in their ability to stimulate T cells in vitro to analyze the level of cross-presentation of the different peptides expressed by the tumors. For comparison, we used the transduced MC57 and 8101 cancer lines grown in vitro. As seen for the similar killing in vitro of MC57 cells presenting the different peptides in Figure 1D, direct presentation also led to comparable amounts of IFN-γ and TNF-α secretion by cognate T cells (Figure 5). However, in the 8101 model more IFN-γ was found when targeting SIY versus hgp100₂₅ and mgp100₂₅ (Figure S2C). Even bigger differences occurred in both cancer models, when T cells were stimulated with stromal cells. CD11b⁺ stromal cells cross-presenting SIY and Tyr₃₆₉ stimulated cognate T cells even more strongly than directly presenting cancer cells (Figure 5). In contrast, both gp100₂₅ peptides were very poorly cross-presented. While stromal cells from hgp100₂₅-expressing tumors stimulated T cells to secrete low levels of both cytokines, stromal cells from mgp100₂₅ tumors did not stimulate T cells at all (Figure 5 and Figure S2C). The heteroclitic peptide EGP behaved similarly to hgp100₂₅ when cross-presented; it stimulated pmel T cells to secrete low amounts of IFN-γ (Figure S4). Thus, the peptides with high affinities for MHC (SIY and Tyr₃₆₉) were well cross-presented, while peptides with low affinities (all three gp100₂₅ peptides) were so poorly cross-presented that the respective stromal cells could not efficiently stimulate cognate T cells ex vivo.

Figure 5. Only SIY and Tyr₃₆₉ are cross-presented, as detected by cytokine secretion by T cells stimulated by stromal cells isolated from untreated tumors.

CD11b⁺ stromal cells were isolated from established untreated tumors and were co-cultured with peptide-activated T cells. Enriched stromal cells from tumors grown from MC57-SIY, -hgp100 and -mgp100 cells (all grown in OT-I mice) and -TyrHHD (grown in AOTA mice (non-self)) were co-cultured with 2C pmel AFH (non-self) or FH (self) TCR-transgenic T cells. Stromal cells from the various tumors were compared to cultured cancer cells expressing the same antigen. Supernatants were harvested after 24 h of co-culture and amounts of IFN-γ and TNF-α measured by ELISA. Data are shown as percent of maximal cytokine secretion (anti-CD3 and -CD28 antibody stimulation, defined as 100 %). For TNF-α, cytokine secretion by stromal cells without T cells was subtracted to obtain specific TNF-α secretion by T cells (443 to 975 pg/ml by 1×10⁵ cells cultured in 200 µl for 24 h, depending on experiment). Unstimulated T cells served as negative control (below 0.7 % for all responders, not shown). Data shown are combined from two experiments and are representative for four independent experiments. See also Figure S4.

* not done

Destruction of tumor stroma is stronger when targeted peptides have high affinity for MHC

We analyzed regressing MC57 tumors to help us understand how tumor eradication correlated to peptide-MHC affinities. Tumors were dissected on day five after adoptive T cell transfer and we analyzed the viability of CD11b⁺ stromal cells. Stroma from tumors expressing SIY or Tyr₃₆₉ showed a high percentage of dead cells (Figure 6), 6- and 5.8-fold increase over background, respectively. In accordance with the relapse of tumors from gp100₂₅ peptide-expressing cells, death of tumor stroma was low, with only 1.9- and 2.2-fold increases over background for hgp100₂₅ and mgp100₂₅, respectively. In conclusion, tumors that were eradicated by cognate T cell therapy showed a high rate of stromal death, while relapsing tumors contain less dead CD11b⁺ stromal cells.

Figure 6. Death of stromal cells in T cell-treated SIY and Tyr₃₆₉ expressing tumors.

MC57-SIY, -TyrHHD, -hgp100 and -mgp100 tumors were treated with 2C, AFH (non-self) or pmel T cells, respectively. Five days after adoptive transfer, single cell suspensions were generated from treated and untreated tumors and analyzed by flow cytometry. Histograms show CD11b⁺ T cells stained with propidium iodide (PI) to identify cell death. Numbers indicate the percentage of dead cells among CD11b⁺ cells. Data are representative for four independent experiments, with single mice per group.

Discussion

Our results have a direct impact on the design of adoptive immunotherapy. First, and most importantly, the affinity of the targeted peptide for the presenting MHC was highly predictable of success or failure of T cell therapy, indicating that this is a key variable. Only high-affinity peptides that were efficiently presented by cancer cells and/or stroma induced cytokine secretion by T cells, stromal death and relapse-free regression of tumors. Second, targeting self-antigens on tumors did not preclude eradication of large cancers even though the treated mice developed vitiligo (data not shown). Such autoimmunity was also observed in patients treated with anti-self T cells (Morgan et al., 2010; Palmer et al., 2008; Parkhurst et al., 2011; Yee et al., 2000). As might be expected, we found that the self-peptide with the higher affinity for the presenting MHC molecule was associated with stronger autoimmunity (mTyr₃₆₉ as opposed to mgp100₂₅). Vitiligo was also detected earlier in mice transgenic for the FH TCR compared to pmel (Figure S1 and (Gregg et al., 2010)).

We have analyzed the potential of the different peptides to be cross-presented by tumor stroma. This is an effective readout to evaluate different affinities of peptides for MHC and more sensitive than direct presentation by cancer cells. All peptides were overexpressed by the cancer lines, therefore, no differences in direct presentation were detected in killing and cytokine secretion assays, but cross-presentation reflected the results obtained from the cell-free affinity measurements. Death of stroma correlated with the amount of cross-presentation and tumor relapse. Though it seems to be required for tumor eradication, we do not know whether cross-presentation is essential for stromal death. For example, direct T cell stimulation provided by cancer cells expressing peptides with high affinity for MHC can lead to strong cytokine production, Fas ligand upregulation, and bystander killing (Wang et al., 1996), which could destroy stroma. Stromal cross-presentation may also not be needed when an essential oncogene on the cancer is targeted (Anders et al., 2011; Listopad et al., 2013).

Targeted peptides that led to tumor eradication fell into a category of high-affinity MHC binders (IC₅₀ < 10 nM), whereas affinities of peptides that led to relapse fell into a category of intermediate (IC₅₀ between 50 and 500 nM) or low binders (IC₅₀ > 500 nM). These data are consistent with the low nanomolar affinities needed to provide full protection against lethal Vaccinia virus infection (Moutaftsi et al., 2009). Low affinity peptides may allow perforin-mediated killing, which requires only two to three peptide/MHC complexes and brief T cell–target cell interactions (Purbhoo et al., 2004). However, the efficacy of adoptively transferred T cells to eradicate tumors does not depend on perforin (Garcia-Hernandez Mde et al., 2010; Listopad et al., 2013). The high affinity of peptides for MHC is probably needed for tumor eradication because this allows the formation of stable synapses between T cells and antigen-positive cancer cells and/or stromal cells cross-presenting the antigen. At least ten peptide/MHC complexes need to be engaged for the prolonged interactions required to stimulate T cells to secrete cytokines (Purbhoo et al., 2004), which are essential for tumor eradication (Garcia-Hernandez Mde et al., 2010; Listopad et al., 2013; Zhang et al., 2008). It appears that targeting peptides with affinities below a certain threshold will result in a level of stimulation of effector T cells that is insufficient to eradicate the cancer, resulting in relapse of antigen-positive or -negative cancer cells.

Several studies have tried to overcome relapse after adoptive T cell therapy. They show that the anti-tumor effects of adoptively transferred T cells can be enhanced by selecting for more effective T cell populations, multiple transfers of T cells, high-dose IL-2, vaccinations, and/or total body irradiation (Cheever et al., 1980; Cho et al., 2012; Dummer et al., 2002; Ho et al., 2003; Ly et al., 2010; Matsui et al., 2003; North, 1982; Overwijk et al., 2003). But even under these conditions, relapse was often observed when peptides with low affinities for MHC were targeted (Antony et al., 2005; Gattinoni et al., 2005; Gattinoni et al., 2009; Overwijk et al., 2003).

TCR affinity can undoubtedly be an important factor (Gottschalk et al., 2012). However, in the study presented here, the affinity of the peptides for MHC seemed to determine if T cells could eradicate tumors or not. As reasons for this, we propose that the affinities (K_D) of the majority of natural TCRs, measured by surface plasmon resonance, are 1 to 100 µM (Davis et al., 1998; Williams et al., 1999). This is a very narrow range considering the affinity range from under 1 to more than 20,000 nM measured for the different peptides binding MHC. In the same line of argument, a study by Bowerman and colleagues demonstrated in vitro that the magnitude of T cell activity against peptide/MHC was influenced more by peptide binding to MHC than by binding of TCR to peptide/MHC, especially for higher affinity TCRs (Bowerman et al., 2009). Finally, T cells expressing the 2C TCR even when targeting a peptide/MHC complex with a 30-fold higher affinity could not prevent relapse in the absence of cross-presentation. Using 2C T cells to treat MC57-SIY and MC57-L^d tumors, SIY-expressing tumors were rejected (Figure 2B and (Spiotto et al., 2004); affinity of 2C TCR for SIY-K^b (K_D = 30 µM)) while MC57-L^d cancer cells grew out as ALV ((Spiotto et al., 2004); affinity of 2C for QL9- and p2Ca-L^d (K_D ≈ 1 µM) (Corr et al., 1994; Garcia et al., 1997; Holler and Kranz, 2003)). In contrast to SIY, QL9- and p2Ca-L^d, recognized by 2C as alloantigens, cannot be cross-presented, as the entire peptide/MHC complex would need to be taken up by the stromal cells and then be re-expressed on their surface.

While we did not study the influence of several TCRs with different affinities to one (same) peptide MHC complex, the influence of one TCR (pmel) on tumors expressing three peptides with different affinities for MHC was studied here. The affinities of the pmel TCR for the three peptide/MHC complexes studied, mgp100₂₅, EGP and hgp100₂₅ binding D^b, are not known; however, alanine-scans of murine and human gp100₂₅ suggest similar affinities since the first three amino acids, which harbor the only differences between the three peptides, do not contribute to the binding of the peptide/MHC complex to the TCR (but are important for the binding of the peptides to MHC (Overwijk et al., 1998)). Also, structural studies of these peptides (van Stipdonk et al., 2009) in complexes with D^b indicate that the two positions that influenced binding to D^b (p2 and p3) are both pointing down into the MHC pocket. In fact, the authors did not see significant differences in any of the exposed regions of the peptides, which would be in contact with the TCR. Nevertheless, only cancer cells expressing the peptide with the highest affinity for D^b (hgp100₂₅) were effectively killed in vivo, tumors expressing the other peptides relapsed being antigen positive. As none of the three peptides supported complete tumor eradication, we further analyzed their affinity for MHC in detail, which has given us insight into the importance of the stability of peptide–MHC interactions. In the original description of EGP (van Stipdonk et al., 2009), two different RMA-S cell-based assays were employed to determine the relative affinity of EGP for D^b compared to the murine gp100 peptide (EGS) and human gp100 peptide (KVP). These assays were: first, a binding assay that measured cell surface D^b levels as a function of peptide concentration; in this assay EGP was almost 100-fold better than hgp100₂₅ (KVP) and 1000-fold better than mgp100₂₅ (EGS). Second, a stabilization assay that measured the cell surface lifetimes of the peptide/D^b complexes. In this stabilization assay EGP and hgp100₂₅ (KVP) showed similar lifetimes, whereas mgp100₂₅ (EGS) had a considerable shorter lifetime (i.e. the complex was less stable). Consistent with their results, we observed a 50-fold increased affinity of EGP over mgp100₂₅, and this affinity of EGP was similar to that of hgp100₂₅ (454 nM and 186 nM, respectively). We used a cell-free competition binding assay, which is influenced by on- and off-rates (giving an approximation of the dissociation constant (K_D)). Taken together, EGP demonstrated similar MHC stabilization compared to hgp100₂₅ (van Stipdonk et al., 2009). Indeed, like hgp100₂₅, the affinity of EGP for D^b was insufficient to allow for tumor eradication.

Heteroclitic peptides can induce strong T cell responses that include TCRs with high affinities (Gold et al., 2003; van Stipdonk et al., 2009). However, these T cells will not be able to eradicate tumors if the targeted tumor antigen has low affinity for its presenting MHC. An example is a recent clinical trial that showed no improvement of anti-melanoma effects by addition of vaccinations with heteroclitic gp100 peptides to the immune stimulating anti-CTLA-4 antibody (Hodi et al., 2010). The affinities of the corresponding natural peptides are gp100₂₀₉: 83 – 172 nM and gp100₂₈₀: 94 – 455 nM (Kawakami et al., 1995; Parkhurst et al., 1996; Tsai et al., 1997)).

Since our data show that high affinity of peptide for MHC results in tumor eradication along with strong stimulation of T cells to secrete cytokines, future studies should concentrate on targeting peptides that have high affinities for presenting MHC class I. There are several algorithms that are constantly being improved to give a fairly reliable prediction of peptide affinities for MHC (e.g. Immune Epitope Database Analysis Resource). Nevertheless, the predicted affinities of the peptides show 2 to 20-fold differences when compared to the measured affinities (as analyzed here for SIY and OVA₂₅₇, respectively). While the affinities of peptides for MHC can be accurately measured in standardized cell-free assays, natural processing and presentation of these putative peptides needs also to be confirmed before selecting a peptide as a therapeutic target (Popovic et al., 2011). Together, it should be possible to identify optimal targets for T cell therapies when analysis of peptide-MHC affinity is included.

Experimental Procedures

Cell lines

Phoenix-ampho (Fujita et al., 1992) cells were cultured in DMEM (Mediatech, Manassas, VA), 10% non-heat inactivated FCS (Sigma-Aldrich, St. Louis, MO) at 37°C in a 5% CO₂ humidified incubator. Cancer cells lines were cultured in DMEM, 5% FCS (Gemini Bio-Products, West Sacramento, CA) at 37°C in a 10% CO₂ dry incubator. 8101 originated in a UV-treated C57BL/6 and has been described (Dubey et al., 1997; Schreiber et al., 2001). P. Ohashi (University of Toronto, Toronto, Ontario, Canada), with permission of H. Hengartner (University Hospital Zurich, Zurich, Switzerland), provided the MC57G methylcholanthrene-induced, C57BL/6-derived fibrosarcoma (MC57). Its transfectant MC57-SIY-1 (MC57-SIY) has been described previously (Spiotto et al., 2002). The new cell lines 8101- and MC57-hgp100, 8101- and MC57-mgp100, 8101-SIY, MC57-EGP and MC57-OVA were generated by transductions of 8101 or MC57 with MFG retroviral vectors expressing peptide-EGFP fusion genes (see Supplemental Experimental Procedures for details on retroviral vectors and transductions). MC57-mgp100/SIY was derived from MC57 by subsequent transduction with MFG-(SIY)₃-Cerulean and MFG-mgp100-EGFP. MC57-TyrHHD was obtained by sequential transductions with MFG-Tyr-EGFP and MP71-HHD, encoding a fusion protein of a HLA-A2/D^d chimera and human β2m (Pascolo et al., 1997).

Mice

A list of the pairs of mice used as hosts of tumors and donors of T cell can be found in Table 1. OVA₂₅₇-K^b-specific TCR-transgenic OT-I mice were provided by M. Mescher (University of Minnesota, Twin Cities, MN), the SIY-K^b-specific TCR-transgenic 2C mice were provided by J. Chen (Massachusetts Institute of Technology, Cambridge, MA), and the human and murine gp100₂₅-D^b-specific pmel-1 (referred to as pmel) were provided by N. Restifo (National Cancer Institute, Bethesda, MD) (Overwijk et al., 2003). Other TCR-transgenic mice used in this study are the murine Tyr₃₆₉-A2-specific AFH mouse (Nichols et al., 2007), which is also AAD-transgenic (HLA-A2 and D^b chimera (Newberg et al., 1996)) and albino (Tyr-deficient) (Colella et al., 2000). It is important to note that the FH TCR used for targeting the self-peptide mTyr₃₆₉ has been derived in a non-self setting. The TCR was obtained from a Tyr-deficient albino mouse (Nichols et al., 2007), while the mouse from which pmel was obtained expressed mgp100₂₅ (Overwijk et al., 2003). This does not imply that the pmel TCR specific for mgp100₂₅ is of lower affinity, but rather that TCRs with a certain affinity for peptide/MHC can only be found naturally, if the target peptide is not expressed (FH) or of low affinity for MHC (pmel). The tyrosinase-positive, Tyr₃₆₉-A2-specific, AAD-transgenic FH mice were generated by crossing AFH to C57BL/6J (The Jackson Laboratory, Bar Harbor, ME) and selecting black (Tyr⁺) mice. The OT-I-, Thy1.1- and AAD-transgenic strains AOTA (albino, Tyr⁻) and OTA (Tyr⁺) were obtained by crossing ATA (Nichols et al., 2007) to OT-I/Thy1.1 (Thy1^a; provided by T. Gajewski, The University of Chicago, Chicago, IL) and selecting for mice with white or black fur color, respectively. All colonies, including Rag1^−/−(B6.129S7-Rag1^tm1Mom/J, The Jackson Laboratory), were maintained at the University of Chicago facilities. The Institutional Animal Care and Use Committee at the University of Chicago approved all animal experiments and all experiments were performed conform to the relevant regulatory standards.

Peptides

The peptides EGP (EGPRNQDWL), hgp100₂₅ (KVPRNQDWL), mgp100₂₅ (EGSRNQDWL), OVA₂₅₇ (SIINFEKL), SIY (SIYRYYGL), and Tyr₃₆₉ (FMDGTMSQV) were made by solid-phase peptide synthesis using standard FMOC chemistry (see Supplemental Experimental Procedures for details).

T cell cultures

NH₄Cl-treated splenocytes were cultured at 4 × 10⁶ cells/ml, 3 ml per well of a 6-well plate in RPMI, 10% FCS (Sigma-Aldrich), 2 mM glutamine, 50 µM β-mercaptoethanol, 1 mM Hepes, 1 mM sodium pyruvate, 1x non-essential amino acids, 100 U/ml penicillin, 100 µg/ml streptomycin, 50 µg/ml gentamicin (all Gibco/Invitrogen, Carlsbad, CA). T cells were activated with 1 µg/ml anti-CD3 (145-2C11) and anti-CD28 (37.51, both eBioscience, San Diego, CA) for killing assays in vitro or 1 µM cognate peptide: SIY for 2C, Tyr₃₆₉ for AFH and FH, OVA₂₅₇ for OT-I, and hgp100₂₅ for pmel. Activated T cells were used for adoptive transfer after 3 days and for assays in vitro after 4 days of culture.

Cytotoxicity assay

Cell-mediated lysis of target cells by activated T cells was determined by standard 4.5-h ⁵¹Cr-release assay. Briefly, target cells were labeled for 1 h with 100 µCi sodium chromate-51 (Perkin Elmer, Waltham, MA) and incubated with T cells using E:T ratios from 50:1 to 1.3:1, using 5 × 10³ target cells. The ⁵¹Cr-released was measured using a gamma counter (Titertek, Huntsville, AL). The percentage of specific lysis was calculated as: % specific lysis = [(experimental release-spontaneous release)/(maximum release - spontaneous release)] × 100.

Tumor challenge and treatment

Cultured cancer cells were trypsinized and washed with PBS. Cancer cells in suspension (MC57: 2 × 10⁶ / 200 µl, 8101: 5 to 10 × 10⁶ / 200 µl) were injected subcutaneously onto the shaved back of mice. Tumor volumes were measured along three orthogonal axes (a, b, and c) every 3 to 4 days and tumor volume calculated as abc/2. MC57 tumors were treated after at least 14 days, when tumors reached approximately 500 mm³; 8101 tumors were treated after at least 5 weeks, when tumors reached approximately 300 mm³. Mice were treated with 3-day activated T cells, one spleen per recipient. We injected 5.5 ± 1.3 × 10⁷ activated 2C T cells, 5.3 ± 2.4 × 10⁷ activated FH T cells and 6.9 ± 2.2 × 10⁷ activated pmel T cells (Numbers were derived from six independent experiments). 8101 tumors were treated with cells from half a spleen only, and with naïve T cells in some of the experiments (see Table S2). T cell suspensions of one spleen were injected into the recipient via the retro orbital plexus in two doses of 150 µl. For tumor protection, T cells were injected on the day of tumor challenge or 3 days later, as indicated.

Isolation of stromal and cancer cells from tumors

Two-week old, untreated tumors were used for functional analysis; tumors of mice treated with T cells 4 or 5 days prior were used for flow cytometric analysis of T cells and stromal death, respectively. Tumors were surgically excised and single cell suspensions generated by enzymatic digestion (see Supplemental Experimental Procedures). For stromal cross-presentation, CD11b⁺ cells were enriched using magnetic beads (Dynabeads FlowComp Flexi (Invitrogen Dynal, Oslo, Norway) and anti-CD11b antibody (M1/70, BD Bioscience, Franklin Lakes, NJ)).

To analyze antigen loss, relapsed tumors were surgically excised under sterile conditions and placed in DMEM on ice. Tumors were minced to 1 – 2 mm pieces and seeded in DMEM, 10% FCS, 100 U/ml penicillin, 100 µg/ml streptomycin, 50 µg/ml gentamicin and 50 µg/ml nystatin. Cells and fragments in the flask were not moved for the initial three days and then cultured normally.

Cytokine release assay

T cells activated in vitro for 4 days were incubated with cancer cells cultured in vitro or tumor stromal cells obtained ex vivo. 1 × 10⁵ responders were cultured with 1 × 10⁵ stimulators per well of a 96-well U-bottom plate for 24 h. Wells coated with 1 µg/ml of anti-CD3 (145-2C11) and anti-CD28 (37.51, eBioscience) served as positive controls and maximal stimulation. All supernatants were removed and tested for IFN-γ and TNF-α using ELISA Kits (‘Femto-HS’ High Sensitivity, eBioscience) according to the manufacturer’s protocol.

Flow cytometry

Cells were stained using directly labeled antibodies (see Supplemental Experimental Procedures). Flow cytometry data were acquired on FACSCalibur or FACSCanto machines (BD) and data were analyzed using FlowJo (Tree Star, Ashland, OR) software. Cell sorting was performed using FACSAria (BD) or MoFlo-HTS (Beckman Coulter, Brea, CA) at the Flow Cytometry Facility of The University of Chicago.

MHC peptide binding assays

MHC purification, and quantitative assays to measure the binding affinity of peptides to purified H2-K^b, H2-D^b, and HLA-A*0201 molecules were performed as previously described (Assarsson et al., 2007; Sidney et al., 2001) (see Supplemental Experimental procedures for details). Under the conditions used, where [label] < [MHC] and IC50 ≥ (MHC), the measured IC50 values are reasonable approximations of the true K_D values.

Statistical analysis

Results of treatment of small groups of mice were analyzed using the two-tailed probability calculated by the Fisher’s exact probability test (p ≤ 0.05 is considered significant, p ≤ 0.01 highly significant).

Highlights.

• Tumor relapse versus eradication is determined by affinity of peptide for MHC

• Outcome of adoptive T cell therapy is determined by affinity of peptide for MHC

• Stroma is only destroyed in tumors expressing peptides with high affinity for MHC

• Efficient cross-presentation is dependent on high peptide-MHC affinity

Significance.

Cancer relapse remains the greatest obstacle to virtually any cancer therapy. Our data show that high affinity of the targeted peptides for MHC is required for strong stimulation of T cells to secrete cytokines and cause relapse-free tumor eradication. Adoptive T cell transfer therapies should, therefore, target peptides that have high affinities for the presenting MHC class I.

Nonstandard abbreviations used

ALV

antigen-loss variant

nM

nanomolar

Tyr

tyrosinase