High-throughput cloning reveals diverse properties of T-cell receptors targeting minor histocompatibility antigens

Key Points

• •A novel high-throughput single cell–based platform enabled the rapid cloning of TCR recognizing human miHA.
• •Multiple unique TCR recognizing HA-1 and HA-2 were cloned; these had a broad range of avidities despite recurrent TCRβ chain family usage.

Visual Abstract

Abstract

Hematopoietically restricted minor histocompatibility antigens (miHAs) presented on HLA-class I molecules are ideal targets for adoptive T-cell immunotherapy in the context of an allogeneic stem cell transplant (alloSCT). This is because CD8 cells that recognize these miHAs can mediate potent antileukemia effects with a low risk for graft-versus-host disease (GVHD). A barrier to translating this concept broadly to the clinic has been the difficulty and expense of cloning potentially high-value T-cell receptors (TCRs). Here, we describe the isolation of anti-miHA TCRs using a novel high-throughput platform (TCXpress) that enables the rapid cloning and characterization of TCRs from single cells without nucleic acid sequencing or gene synthesis. We cloned 7 unique TCRs recognizing the miHA HA-1^H from 50 mL of blood from an HLA-A∗02:01 HA-1^R/R woman who was immunized against HA-1^H during pregnancy. After in vitro expansion, 13 additional unique anti-HA-1^H TCRs were cloned from 740 HA-1^H-dextramer⁺ cells, with the screen completed in 35 days. We also cloned 12 unique anti-HA-2^V TCRs from 771 HA-2^V-tetramer⁺ cells from a patient with myelodysplastic syndrome (MDS) that had relapsed after alloSCT and who received a donor leukocyte infusion. Ten anti-HA-2^V TCRs were isolated from an unmanipulated sample collected at GVHD-onset, and 8 were isolated when GVHD and MDS were in remission. Anti-HA-1^H and -HA-2^V TCRs had a wide range of avidities measured by activation of cell lines expressing them. Taken together, our results validate a new method for TCR isolation and characterization with potentially broad applications, and provide insights into the nature of anti-miHA responses.

Introduction

In HLA-matched allogeneic stem cell transplant (alloSCT), alloreactive T cells recognize minor histocompatibility antigens (miHAs), which are the peptide products derived from coding polymorphisms that distinguish recipients from donors.^1,2 miHAs vary in tissue distribution patterns, with some being expressed in all tissues and others having a more limited pattern of expression. HLA-class I-restricted miHA, relatively restricted to hematopoietic cells, have been considered promising targets for graft-versus-leukemia (GVL) mediated by alloreactive CD8⁺ T cells.³ This is because CD8⁺ T cells must make cognate T-cell receptor (TCR) contact with nonhematopoietic tissues to cause significant graft-versus-host disease (GVHD).⁴ Therefore, it might be possible to engineer a high-magnitude CD8⁺ T-cell GVL response against a hematopoietically restricted miHA without causing serious GVHD. This general approach has been demonstrated in mouse GVL models targeting the hematopoietically restricted miHA, H60.^5,6

The miHA HA-1, encoded by ARHGAP45, is considered an ideal target for adoptive T-cell immunotherapy⁷ for several reasons: (1) its expression is relatively restricted to malignant and nonmalignant hematopoietic cells; (2) it is presented by the common HLA-A∗02:01 major histocompatibility complex I protein, which is carried by 47% of the US population; and (3) the allele frequencies of the immunogenic HA-1 allele (“HA-1^H”; VLHDDLLEA) and nonimmunogenic allele (“HA-1^R”; VLRDDLLEA) are relatively balanced, making it readily possible to find HLA-matched HA-1^R/R donors for recipients that are HLA-A∗02:01⁺HA-1^H+.

The HLA-A∗02:01-restricted miHA HA-2^V encoded by MYO1G is also an attractive target. Like HA-1^H, its expression is mostly limited to hematopoietic cells.^7,8 Unlike HA-1^H, however, the allele frequency of the immunogenic form (YIGEVLVSV; HA-2^V) is 79.3% globally.⁹ This makes it less likely that patients will have a related HLA-matched donor who is HLA-A∗02:01⁺ and homozygous for the nonimmunogenic (YIGEVLVSM) HA-2^M allele; however, because of how common HLA-A∗02:01 is in the population, HLA-matched HA-2^M/M donors are well represented in transplant registries. Also, HA-2^V can be targeted in haploidentical transplant using donors who lack HLA-A∗02:01-containing haplotypes. Clinical trials are underway targeting both HA-1^H and HA-2^V.10, 11, 12

A barrier to more thoroughly testing the selective targeting of miHAs to improve GVL is that the cloning of anti-miHA TCRs has been challenging, time-consuming, and costly. Historically, this has required the creation of T-cell lines specific for miHAs from which TCRs can be cloned or, alternatively, sequenced and then synthesized. This can be laborious because of the low frequency of miHA-reactive T cells among naïve T cells, and typically only a few TCRs are isolated.¹³ The frequency of alloreactive T cells may be increased in recipients after an alloSCT, but creating lines from such T cells is confounded by post-alloSCT lymphopenia and the use of calcineurin inhibitors and other immunosuppressants that reduce T-cell numbers. Efficient cloning of TCRs targeting antigens that may make for useful antipathogen or anticancer therapeutics suffers from similar barriers.

To overcome these hurdles, we developed a high-throughput platform, TCXpress, which enables the rapid cloning and characterization of TCRs from single cells without requiring the prior generation of T-cell lines or clones, TCR sequencing, or TCR gene synthesis. Using this approach, we isolated 20 unique TCRs against HA-1^H from a parous woman who was likely immunized against HA-1^H during pregnancy, as well as a single anti-HA-1^H TCR from a male donor who had never received a blood transfusion and thus likely was naïve to the HA-1^H epitope. We also cloned 12 TCRs against HA-2^V14 from an HA-2^V recipient of an HA-2^M/M allograft who developed GVHD following a donor leukocyte infusion (DLI) to treat relapsed myelodysplastic syndrome (MDS).

Methods

Blood donors

Specimens were obtained under institutional review board–approved protocol 18-183 with informed consent.

Anti-HA-1^H donors

We aimed to clone TCRs against miHAs from parous women who may have been immunized against miHA during pregnancy.¹⁵ Using exome sequencing, we identified an HLA-A∗02:01^+/+ HA-1^R/R woman who had 3 children with an HLA-A∗02:01^+/− HA-1^H/R father; 50 mL of her peripheral blood was collected for HA-1^H-specific CD8⁺ T-cell isolation. A leukapheresis product (Precision for Medicine) was used to isolate TCRs against HA-1^H from an HLA-A∗02:01 HA-1^R/R male who had not received a blood transfusion.

The anti-HA-2^V patient

Using exome sequencing, we identified an HLA-A∗02:01⁺ HA-2^V male aged 66 years with relapsed MDS after an HLA-matched alloSCT from an HLA-A∗02:01⁺ HA-2^M/M donor who developed GVHD following a DLI.

Cloning TCRs using TCXpress

Major histocompaibility antigen complex class I (MHCI) multimer⁺ CD8 cells were single-cell sorted into 384-well plates. Within each well, the cell was lysed, and complementary DNA (cDNA) was synthesized using random hexamer primers. Half of the cDNA product from each well was transferred to a second 384-well plate, followed by first-round polymerase chain reaction (PCR) amplifications of TCRα and TCRβ variable regions (TRAVs and TRBVs) using the primer cocktail (supplemental Tables 3 and 4). After this first-round amplification, TRAVs and TRBVs were separately amplified using nested PCR; 40 and 30 forward primers capable of binding the 45 and 48 known TRAVs and TRBVs (supplemental Tables 5 and 6), each with overhanging anchor sequences to enable Gibson assembly, were combined with single reverse TCRα and TCRβ primers. These reverse primers bind just 3ʹ of the TRAV-J-C and TRBV-J-D-C junctions, which enables the amplification of sufficient constant regions for Gibson assembly. TCRα and TCRβ PCR products were purified with Sera-Mag Speed Beads (Cytiva, Marlborough, MA). The purified TCRα TCRβ products were mixed with a TCRα constant region fragment containing Gibson cloning ends and a lentivirus vector containing the TCRβ constant region with a 5ʹ Gibson cloning site followed by Gibson assembly.¹⁶ Each TCR was cloned into a lentivirus backbone (pLVX-EF1a-IRES-puro for HA-1^H and pLV230-PGK-IRES-Puro-WPRE for HA-2^V). Escherichia coli cells were transformed with the Gibson product. Plasmids purified from the bulk broth were used to create lentivirus.

Molecular deconvolution of TCR-encoding plasmids

After the Gibson reaction is transformed into bacteria, single colonies are not picked. Rather, plasmid DNA is generated from the “bulk” bacterial culture. In addition to including a correctly assembled construct with the TCRα and TCRβ chains that drove reactivity or major histocompatibility complex I-multimer binding, this DNA could also include plasmid molecules with: (1) an alternate TCRα chain as T cells frequently express 2, (2) errors from the reverse transcription or PCR amplification steps, or (3) errors induced by the Gibson assembly. In the initial screen, the pools of Jurkat reporter cells used for screening contain cells that express a mix of these molecules. We use “molecular deconvolution” to isolate single lentivirus backbones encoding single TCRαβ pairs. Briefly, the plasmid generated from the original Gibson assembly product is retransformed into E. coli which are cultured on agarose plates. Plasmid DNA derived from 3 to 5 colonies are sequenced to identify the TCRα identities and overall TCR integrity, and used to create individual lentiviruses. Separate Jurkat cells are transformed with these individual lentiviruses, selected in puromycin, and screened for functional assays. The detailed methods for TCR transduction of Jurkat cells, TCR affinity estimations, and cytotoxicity assays are described in the supplemental Materials.

Statistical analyses

Unpaired t tests were used to compare killing at different effector-to-target ratios. Simple linear regression was performed to test the correlations between 50% effective concentration (EC₅₀) and tetramer fluorescent intensity in the original sort. Prism (GraphPad) was used to calculate statistics.

Results

Cloning of HA-1^H-specific TCRs from a parous woman

We identified a multiparous HLA-A∗02:01⁺ HA-1^R/R woman who had 3 children with an HA-1^H/R HLA-A∗02:01^+/− father. Forty-eight CD8⁺dex^HA-1H+ cells were single-cell sorted into a 384-well plate from a total of 3.9 × 10⁷ peripheral blood mononuclear cells (PBMC) containing 3.1 × 10⁶ CD8 cells (Figure 1A). The TCXpress process was used to clone TCRs from these cells (Figure 1B). Briefly, cDNA was generated within each well using random hexamer primers. TRAVs and TRBVs were then amplified within the same well in the first-round PCR step. The PCR products were then split into 2 wells, followed by a nested second-round PCR amplification with primers that enabled Gibson cloning. The amplified TRAV and TRBV fragments were purified, then combined with a TCRα constant region fragment flanked by Gibson ends and a lentiviral backbone vector encoding the TCRβ constant region fragment (also with a Gibson cloning sequence) for Gibson assembly. Plasmid was then generated and used to express TCRs in 293-CD3-CD8 cells by transient transfection or to create lentivirus for cell transduction.

Figure 1.

The cloning HA-1^H-reactive TCRs from a parous woman. (A) Schema describing the TCXpress TCR cloning process and TCR characterization: (1) CD8⁺HA-1^H-multimer⁺ cells were single-cell sorted into plates, followed by cDNA generation; (2) TRAVs and TRBVs were amplified in a fashion that generates ends compatible for Gibson cloning; (3) amplified TRAV and TRBV fragments combined with a TCRα constant region fragment (TRAC) and a retroviral vector that includes the TCRβ constant region fragment (TRBC) underwent Gibson assembly; (4) single cell–derived TCRα and TCRβ chains were expressed stably in Jurkat cells via lentivirus infection or transiently in 293-CD3-CD8 cells; and (5) functional analyses of TCR clones by MHCI-multimer binding and via coculture of Jurkat reporter cell lines expressing TCRs with cells that do or do not present the putative target antigen. (B) Flow plot showing the 48 sorted dex^HA-1H+ cells (in red), gated on viable CD14⁻CD19^˗CD3⁺ cells. FMO panel shows acquisition of events with the same gate except without dex^HA-1H staining. (C) Dex^HA-1H and TCR expression of 293-CD3-CD8 cells transiently transfected with lentiviral constructs encoding 7 unique anti-HA-1^H TCR clones. (D) Interleukin-2 (IL-2) enzyme-linked immunospot data of TCR-transduced Jurkat cell lines stimulated with HA-1^H-pulsed T2 cells from 7 anti-HA-1^H TCRs determined to be unique by TRAV and TRBV CDR3 sequencing. The number in the left upper corner of each well is the count of IL-2 spots based on automated counter analysis. (E) CD69 expression of Jurkat cells expressing anti-HA-1^H clone 7, after coculture with T2 cells pulsed with different concentrations of HA-1^Hpeptide or HA-1^R peptide (500 nM). (F) Estimation of EC₅₀s of these 7 anti-HA-1^H TCR clones. Jurkat lines expressing these TCRs were cocultured with T2 cells pulsed with different HA-1^H peptide concentrations, and the percentages of CD69⁺ cells among CD3⁺ cells were determined. (G) Relationship of EC₅₀ and dex^HA-1H fluorescent intensity (FI) at the original index sorting. No correlation was observed. P = .66 by simple linear regression analysis. TCR peptide pulse:response assays were repeated at least twice for most TCRs with similar results. FMO, fluorescence minus one.

TCRα and TCRβ chains were amplified and cloned from 47 of 48 sorted dex^HA-1H+ cells. As an initial screen for TCR expression and dex^HA-1H binding, these plasmids were individually transiently transfected into 293-CD3-CD8 cells, followed by staining with dex^HA-1H. In these cells, expression of a TCRαβ pair mobilizes CD3 to the cell surface. We chose 29 TCRs for further characterization based on the fraction of CD3⁺ cells that bound dex^HA-1H (Figure 1C).

In parallel to the transient transfection approach, lentivirus expressing each of 29 TCRs chosen based on dex^HA-1 binding in 293-CD3-CD8 cells was used to stably transduce Jurkat reporter cells, which were puromycin-selected. To test whether these TCRs could mediate HA-1^H-specific signals, TCR-transduced Jurkat lines were cocultured with HLA-A∗02:01⁺ T2 cells pulsed with HA-1^H peptide. From these 33 TCRs, we identified 7 unique TCRs (determined by TCR sequencing) that supported interleukin-2 production in response to HA-1^H peptide as measured by enzyme-linked immunospot (Figure 1D). These 7 unique clones used TRBV7-9, as has been reported for other HA-1^H-reactive TCRs^13,17,18 (Table 1).

Table 1.

Summary of anti-HA-1^H TCR clones isolated from blood collected from a single parous woman

Clone ID	TCRα	TCRα CDR3	TCRβ	TCRβ CDR3	EC50 (nM)	Fresh PBMC	Expanded cells	Total
1	TRAV13-1:TRAJ56	AASNPTGANSKLT	TRBV7-9:TRBJ1-4	ASSLVRNEKLF	198	1	2	3
2	TRAV21:TRAJ40	AVRGSTSGTYKYI	TRBV7-9:TRBJ2-5	ASSLVAGETQY	34.6	1	0	1
3	TRAV8-1:TRAJ28	AVRDSGAGSYQLT	TRBV7-9:TRBJ2-1	ASLTGTVYNEQF	61.1	1	0	1
4	TRAV17:TRAJ33	ATEMNSNYQLI	TRBV7-9:TRBJ1-2	ASSPLKGSNYGYT	347	1	0	1
5	TRAV13-1:TRAJ49	AAIVGNQFY	TRBV7-9:TRBJ2-1	ASSETTLSEQF	38.4	1	0	1
6	TRAV17:TRAJ29	ATPSGNTPLV	TRBV7-9:TRBJ2-1	ASSQLAGVIEQF	804	1	0	1
7	TRAV3:TRAJ40	AVRAPTSGTYKYI	TRBV7-9:TRBJ2-1	ASSLVSGNEQF	5.8	1	1	2
8	TRAV3:TRAJ40	AVRGGSYKYI	TRBV7-9:TRBJ2-1	ASTPGTVYNEQF	23.2	0	11	11
9	TRAV13-1:TRAJ43	AMRNNNDMR	TRBV7-9:TRBJ2-3	ASSLTTLDTQY	72.2	0	1	1
10	TRAV21:TRAJ40	AVRPRTSGTYKYI	TRBV7-9:TRBJ1-3	ASTELSGNTIY	8.7	0	2	2
11	TRAV3:TRAJ40	AVRGPTSGTYKYI	TRBV7-9:TRBJ2-1	ASSLVSGNEQF	9.1	0	1	1
12	TRAV21:TRAJ36	AVTRTGANNLF	TRBV7-9:TRBJ1-2	ASSLGVLGIGYT	24.5	0	1	1
13	TRAV13-1:TRAJ10	AAHLTGGGNKLT	TRBV7-9:TRBJ2-5	ASSSRAGGETQY	26.1	0	1	1
14	TRAV26-2:TRAJ40	ILRAGSGTYKYI	TRBV7-9:TRBJ2-1	ASSPGTVYNEQF	112	0	1	1
15	TRAV17:TRAJ36	AGTGANNLF	TRBV7-9:TRBJ2-5	ASSLIRGETQY	40.4	0	1	1
16	TRAV17:TRAJ40	ATAPGSGTYKYI	TRBV7-9:TRBJ1-6	AAPPDTYNSPLH	47.6	0	1	1
17	TRAV13-1:TRAJ45	AAPLAGGGADGLT	TRBV7-9:TRBJ1-2	ASSTTLITGYT	59.2	0	2	2
18	TRAV12-3:TRAJ40	AMRATSGTYKYI	TRBV7-9:TRBJ2-1	AVPGGSSYNEQF	75.8	0	1	1
19	TRAV21:TRAJ40	AVRGTTSGTYKYI	TRBV7-9:TRBJ2-5	ASSFLAGETQY	54	0	1	1
20	TRAV26-1:TRAJ40	IVRASTSGTYKYI	TRBV7-9:TRBJ2-1	ASSPGTVYNEQF	360	0	2	2

To further confirm reactivity (especially for clone 6 that only had a twofold increase in interleukin-2 positive spots with peptide stimulation) and to more quantitatively rank the relative EC₅₀s of the isolated TCRs, each TCR-expressing puromycin-selected Jurkat line was cocultured for 16 hours with T2 cells that had been preincubated with a range of HA-1^H peptide concentrations. TCR signaling was quantified by measuring the percentage of CD3⁺ Jurkat cells expressing CD69, with the EC₅₀ defined as the peptide concentration that led to 50% of the maximum CD69 upregulation (Figure 1E, representative flow plot of clone 7). The isolated HA-1^H-specific TCRs had a broad range of EC₅₀s (5.8-804.8 nM; Figure 1F). The maximum response (maximum percentage of CD69⁺ cells) varied from ∼20% to 80%. Generally, the TCRs with the lowest EC₅₀s elicited the highest maximal responses. Clone 6, which had the lowest enzyme-linked immunospot response, only responded to high HA-1^H peptide concentrations.

Because we index-sorted the original dex^HA-1H+ cells, we could compare the dex^HA-1 fluorescent intensity of the original sorted cells to TCR function as measured by their EC₅₀s. We found no correlation (P = .66; Figure 1G), indicating that quality receptors can be obtained from dextramer-low cells.

Cloning anti-HA-1^H TCR from expanded CD8 cells

We next attempted to clone additional anti-HA-1^H TCRs by expanding relatively rare HA-1^H-reactive CD8 cells through in vitro culturing. We took 2 approaches using PBMC harvested from the same parous donor. In the first, we generated autologous B-cell antigen-presenting cells (B-APC) that express high levels of HLA and costimulatory molecules (supplemental Figure 1). These were HA-1^H peptide-pulsed and cocultured for 10 days with unfractionated PBMC as a source of CD8 cells. In the second approach, PBMC were HA-1^H peptide-pulsed and cultured for 10 days without B-APCs. Pre-culture dex^HA-1H+ cells comprised 0.049% of CD8 cells (Figure 2A). After culture, 1.28% and 7.23% of CD8 cells were dex^HA-1H-Fluorescein Isothiocyanate (FITC)⁺ dex^HA-1H-APC⁺ in the peptide only and B-APC cultures, respectively (Figure 2A), indicative of meaningful expansion of anti-HA-1^H CD8 cells. Seven hundred sixty-eight dex^HA-1H-FITC⁺ dex^HA-1H-APC⁺ double-positive CD8⁺ T cells were single-cell sorted from these cultures using more (gate 1) and less (gate 2) stringent criteria for dex^HA-1H-binding (Figure 2B), and subjected to TCXpress. TCRα and TCRβ chains were amplified from 704 of 768 sorted cells, and each TCRα/TCRβ pair was cloned into a lentivirus backbone. These lentivirus plasmids were individually transfected into 293-CD3-CD8 cells. Of 704 TCRs, 440 resulted in dex^HA-1H binding of >10% of CD3⁺ cells (Figure 2C-D), with dex^HA-1H+ TCRs cloned from both culture conditions and from cells sorted with greater or lesser dex^HA-1H binding stringency. The time from single-cell sorting to screening in 293-CD3-CD8 cells was 35 days.

Figure 2.

Cloning of HA-1^H-reactive TCRs from expanded CD8⁺ T cells from the same donor as inFigure 1. PBMC from the same multiparous female described in Figure 1 were stimulated with HA-1^H peptide or HA-1^H peptide-pulsed B-APC as in “Methods,” and live/CD14⁻CD19⁻CD8⁺dex^irrelevant-dex^HA-1H-FITC⁺ dex^HA-1H-APC⁺ double-positive (DP) cells were single-cell sorted with different stringencies of dex^HA-1H binding. (A) dex^HA-1H-FITC/dex^HA-1H-APC staining of samples taken before expansion and from postexpansion cultures 1 day prior to the cell sorting, gating on CD8⁺ cells. (B) The gates used to sort dex^HA-1H DP cells from HA-1^H peptide pulsed PBMC (left panel) or cultured with HA-1^H peptide-pulsed B-APCs (middle and left panels). (C-D) Lentiviral plasmids encoding TCRs from 704 sorted cells from panel A were individually transfected into 293-CD3-CD8 cells and stained with anti-CD3 and dex^HA-1H. Representative staining of 293-CD3-CD8 cells transfected with 3 unique clones is shown in panel C. (D) Plotted is percent dex^HA-1H+ of CD3⁺ cells. Symbol colors indicate from which sort (see panel B) the TCRs were derived. Of 704 tested TCRs, 440 supported dex^HA-1H binding of >10% of CD3⁺ cells. (E) Anti-HA-1^H TCR-expressing Jurkat lines were cultured with T2 cells pulsed with different concentrations of HA-1^H peptide, and the percentages of CD69⁺ cells were used to determine EC₅₀s. (F) TRBV family usage of TCRs categorized as binding dex^HA-1H (F, top panel) or not binding dex^HA-1 (F, bottom panel). TCR peptide pulse:response assays were repeated at least twice for most TCRs with similar results.

Based on TCRβ sequencing, among the clones that resulted in dex^HA-1H-binding in >10% of CD3⁺ cells, 13 additional unique TCRs, all using TRBV7-9, were identified. In addition, 2 of the 7 original unique clones were reisolated, including clone 7 (Table 1), which had the lowest EC₅₀. Jurkat cells transduced with 12 of the 13 additional TCRs specifically recognized HA-1^H-pulsed T2 cells; clone 20 was at best only weakly reactive against HA-1^H (Figure 2E).

In total, 20 unique and functionally diverse HA-1^H-reactive TCRs were cloned from a single parous woman (Table 1). Of note, 30 of 115 (26%) TCRs scored as not binding dex^HA-1 in the screen in Figure 2D also used TRB7-9 (Figure 2F), which is greater than the TRBV7-9 usage in unselected CD8 cells,¹⁹ suggesting that there may have been additional HA-1^H-specific TCRs that were not identified due to technical reasons.

Cloning anti-HA-1 TCR from a nonimmunized male donor

Given the success in isolating anti-HA-1^H TCRs from a healthy female who may have been immunized through pregnancies, we next attempted to clone anti-HA-1^H TCRs from a pheresis product from an HLA-A∗02:01⁺ HA-1^R/R male who had never received a blood transfusion. Purified CD8⁺ T cells were cocultured with autologous monocyte–derived dendritic cells pulsed with HA-1^H peptide. Ten days later, 107 dex^HA-1-Phycoerythrin (PE)⁺/dex^HA-1-APC⁺ CD8 cells were sorted from 674,000 CD3⁺CD8⁺ cells (Figure 3A), and 93 TCRα/TCRβ pairs were amplified. In contrast to what was observed in cells from the multiparous donor, TRBV sequencing revealed a monoclonal expansion of a single TCR, also using TRBV7-9. Two independent isolates of this TCR (TCR 4 and 7) bound dex^HA-1H when transiently transfected into 293-CD3-CD8 cells (Figure 3B). Jurkat cells stably transduced with these TCRs had similar EC₅₀s (9.5 and 10.8 nM) with superimposable response curves (Figure 3C), consistent with their being identical.

Figure 3.

Cloning an HA-1^H-specific TCR from a healthy male donor who had not received a blood transfusion. Purified CD8 cells from an apheresis product were cocultured with autologous monocyte–derived dendritic cells as in “Methods.” (A) Dex^HA-1H staining of CD3⁺CD8⁺CD14⁻CD19⁻ live cells. (B) Dex^HA-1H binding of Jurkat cells transduced with an empty vector or 2 independent isolates of the same TCR clone (TCR 4 and 7) that bound HA-1^H-MHCI-multimers when expressed in 293-CD3-CD8 cells. (C) Peptide dose:response curves for these 2 identical TCRs.

Cloning anti-HA-2^V TCRs from an alloSCT recipient after a DLI

We identified an HLA-A∗02:01⁺ HA-2^V+ male, aged 66 years, with MDS that had relapsed after an HLA-matched unrelated alloSCT from an HA-2^M/M donor.²⁰ He received a DLI containing 10⁷ CD3⁺ cells per kg. His PBMC were collected at 3 time points: the onset of acute GVHD prior to immunosuppressive therapy (time point 1; 73 days after DLI); while on tacrolimus and corticosteroids (time point 2; 87 days after DLI); and after he achieved a complete MDS remission and was only on low-dose tacrolimus (time point 3; 140 days after DLI). To improve sorting specificity, cells were stained with an HLA-A∗02:01 tetramer loaded with an irrelevant peptide (ALIAPVHAV) labeled in FITC (tet^irr), and with HA-2^v tetramers labeled in PE and APC. Tet^HA-2V-PE⁺tet^HA-2V-APC⁺tet^irr- CD8 cells were single-cell sorted into 384-well plates for TCR cloning. PBMC from time points 1 and 2 were also HA-2^V-peptide pulsed and cultured for 7 days to expand HA-2^V-reactive CD8 cells (Figure 4A), which were fluorescence-activated cell sorter-isolated using the same strategy.

Figure 4.

Cloning anti-HA-2^V TCRs from a post-alloSCT DLI recipient who developed GVHD. PBMC were analyzed at 3 time points after the DLI: day +73 at GVHD onset; day +87 when on corticosteroids and tacrolimus; and day +140 when the patient’s MDS was in a complete remission. Tet^HA-2V+ CD8 cells were sorted directly from unmanipulated samples or from cells that had been cultured with HA-2^V peptide. (A) Flow plots showing tet^HA-2V+ staining from each sort (gated on CD8⁺ cells). Two hundred ten and 234 tet^HA-2V+ CD8⁺ T cells were sorted from unexpanded and expanded cultures, respectively, at time point 1 (left panels). At time point 2, 21 and 96 tet^HA-2v+ cells were sorted from the uncultured and cultured samples, respectively (middle plots). At time point 3, 210 tet^HA-2V+ CD8 cells were sorted from unmanipulated PBMC (right plot). (B) Seven hundred forty-four cloned TCRs were individually transiently expressed in 293-CD3-CD8 cells and tested for tet^HA-2V binding. Two hundred eighty out of 744 TCRs supported tet^HA-2V binding of ≥10% of CD3⁺ cells. (C-F) Selected TCRs that bound tet^HA-2V were expressed in Jurkat reporter cells and cocultured with T2 cells pulsed with different concentrations of HA-2^V peptide or HA-2^M peptide (500 nM), and the percentages of CD69 expression were measured by flow cytometry. (C) Tet^HA-2V binding of representative TCRs expressed in Jurkat cells. (D) CD69 expression of Jurkat clones expressing TCRs 5 and 7 after coculture with peptide-pulsed T2 cells. Both TCR were reactive against HA-2^V in a dose-dependent manner. Clone 7 was partially reactive against HA-2^M peptide at a high concentration (500 nM). (E) HA-2^V-peptide dose:response relationships for 12 unique anti-HA-2^V TCRs. (F) Most clones were specific for HA-2^V; clone 7 and clone 10 were partially reactive against HA-2^M. (G) Anti-HA-2^V TCR EC₅₀ values were plotted against tet^HA-2V FI from the original index sort. No significant correlation was found by simple linear regression analysis (P = .58). TCR peptide pulse:response assays were repeated at least twice for most TCRs with similar results.

From uncultured PBMC, at time point 1, 210 tet^irr-tet^HA-2-PE⁺tet^HA-2-APC⁺ cells were sorted from 36 000 CD8 cells; at time point 2, 21 tet^irr-tet^HA-2-PE⁺tet^HA-2-APC⁺ cells were sorted from 22 954 CD8 cells (0.09%); and at time point 3, 210 tet^irr-tet^HA-2-PE⁺tet^HA-2-APC⁺ cells were isolated from 133 206 CD8 cells (0.16%). From PBMC collected at time points 1 and 2 that were cultured with HA-2^V peptide, 234 and 96 CD8⁺ tet^irr-tet^HA-2-PE⁺tet^HA-2-APC⁺ cells were sorted, respectively (Figure 4A).

From these cells, using TCXpress, 771 TCRs were cloned into a lentivirus backbone, and lentivirus plasmid was transfected (individually) into 293-CD3-CD8 cells. Two hundred eighty TCRs resulted in ≥10% of surface CD3⁺ 293-CD3-CD8 cells binding tet^HA-2V (Figure 4B). TCR sequencing revealed that these contained ≥12 unique TCRs.

Lentiviruses encoding each of the 12 unique TCRs were used to create TCR-expressing Jurkat reporter lines (Figure 4C, representative flow cytometry). These were cocultured with T2 cells pulsed with different concentrations of HA-2^V or HA-2^M peptide to estimate their relative EC₅₀s based on CD69 induction. All 12 TCRs were confirmed to be reactive against HA-2^V with a range of EC₅₀s (Figure 4D-E). Interestingly, high-affinity clones 7 and 10 had some reactivity against the nonimmunogenic HA-2^M peptide at a high 500 nM concentration. This was not seen with the other TCRs (Figure 4F). The observed EC₅₀s did not correlate with the tet^HA-1V binding of the sorted source CD8 cells (P = .58; Figure 4G).

The identities, numbers, and sources of cloned anti-HA-2^V TCR are summarized in Table 2. From the time point 1 directly sorted sample, we cloned 10 unique TCRs, 7 of which were isolated more than once (range, 2-11), indicative of in vivo clonal expansion. Five unique TCRs were recovered from the time point 1 expanded sample, all of which were also cloned from unexpanded cells. No functional anti-HA-2^V TCRs were recovered from the time point 2 uncultured sample. However, the cultured sample yielded 4 anti-HA-2^V TCRs, of which 2 were new. At time point 3, 8 unique TCRs were recovered, all of which were isolated at a previous time point. The most frequently isolated clone from uncultured cells was clone 1 (22-times), followed by clones 4 and 8 (12-times). Seven unique anti-HA-2^V TCRs cloned from the time point 1 unmanipulated day +73 sample were also recovered at the remission day +140 sample.

Table 2.

Summary of HA-2^V TCR clones isolated from post-DLI samples

Clone ID	TCRα	TCRα CDR3	TCRβ	TCRβ CDR3	EC50 (nM)	Time point specimens collected (days after DLI) and number of times clones were isolated					Total	Total uncultured
						73		87		140
						Fresh	Exp	Fresh	Exp	Fresh
1	TRAV21:TRAJ42	AVRPKNYGGSQGNLI	TRBV7-8:TRBJ2-7	ASSFPGLASYEQY	18.6	11	1	0	15	11	38	22
2	TRAV21:TRAJ42	AVRQGYGGSQGNLI	TRBV7-8:TRBJ1-6	ASSLIQGTNSPLH	16.9	4	0	0	10	1	15	5
3	TRAV20:TRAJ42	AVQAWDYGGSQGNLI	TRBV18:TRBJ2-3	ASSTREGTDTQY	13.1	3	0	0	0	0	3	3
4	TRAV21:TRAJ42	AVRRGYGGSQGNLI	TRBV7-8:TRBJ2-1	ASSLTAGAYNEQF	42.7	9	2	0	0	3	14	12
5	TRAV21:TRAJ42	AVRTKDYGGSQGNLI	TRBV7-8:TRBJ2-7	ASSLALPAADSYEQY	14.6	2	2	0	0	3	7	5
6	TRAV38-2DV8:TRAJ34	AYRRFYNTDKLI	TRBV19:TRBJ1-1	ASRSAGYAEAF	21.0	0	0	0	3	0	3	0
7	TRAV21:TRAJ42	AVRSGYGGSQGNL	TRBV7-8:TRBJ2-5	ASSLAGGSLQETQY	0.7	0	0	0	5	3	8	3
8	TRAV21:TRAJ42	AVRPGNYGGSQGNLI	TRBV7-8:TRBJ1-4	ASSPPGGLGEKLF	33.5	5	0	0	0	7	12	12
9	TRAV21:TRAJ42	AVRPVNYGGSQGNLI	TRBV7-8:TRBJ2-5	ASRSPRLETQY	82.2	4	2	0	0	2	8	6
10	TRAV21:TRAJ42	N/A	TRBV7-8:TRBJ2-7	N/A	4.9	1	0	0	0	0	1	1
11	TRAV21:TRAJ42	N/A	TRBV7-8:TRBJ2-5	N/A	4.5	1	0	0	0	0	1	1
12	TRAV25:TRAJ26	N/A	TRBV7-9:TRBJ2-5	N/A	17.3	1	2	0	0	3	6	4

Exp, expanded; N/A, not currently available.

As reported for other anti-HA-2^V TCRs, TRBV7-8 usage dominated,²¹ present in 9 of 12 unique clones. There was also an enrichment for TRAV21TRAJ42 usage, present in 9 unique TRBV7-8⁺ clones (Table 2).

Functionality of anti-miHA TCR expressed in primary CD8 cells

Next, we tested whether primary CD8 cells transduced with an anti-HA-1^H or anti-HA-2^V TCRs would recognize or kill hematopoietic targets that naturally express HA-1^H or HA-2^V. Anti-HA-1^H TCR clone 7 was subcloned into a lentiviral vector that also expresses an epitope from CD34. Purified CD8 cells from an HLA-A∗02:01⁻ donor were stimulated with anti-CD3/CD28 beads, followed by CRISPR/Cas9-editing of the endogenous TCRα and TCRβ chains, transduction with the anti-HA-1^H TCR lentivirus (modified to avoid recognition by the TCRα/TCRβ guide RNAs), and expanded in culture for 10 to 12 days. Control CD8 cells were similarly prepared except without TCR transduction. Clone 7-transduced CD8 cells were cocultured for 4 hours with a panel of lymphoblast cell lines (LCL) with diverse HLA types (supplemental Table 2) that were positive or negative for HLA-A∗02:01. We also specifically tested reactivity against HA-1^H+ and HA-1^H- HLA-A∗02:06 LCL as a previously described anti-HA-1^H TCR recognized HA-1^H presented by HLA-A∗02:06.²² CD69 was only induced in cells that coexpressed either HLA-A∗02:01 or HLA-A∗02:06 and HA-1^H (Figure 5A). Importantly, this TCR was not reactive against other HLA molecules, indicating that it is not broadly alloreactive.

Figure 5.

Primary CD8 cells expressing anti-HA-1^H or anti-HA-2^V TCRs kill HLA-A∗02:01 cells that naturally express the targeted miHA. (A) Primary CD8 cells expressing the anti-HA-1^H TCR clone 7 (with knockdown of the endogenous TCRα and TCRβ) were cocultured with a panel of LCLs with different HLA and HA-1 genotypes, and CD69 expression was quantitated. TCR-transduced CD8 cells only recognized HLA-A∗02:01⁺ or A∗02:06⁺ LCLs that were also HA-1^H/H or HA-1^H/R (P < .0001, comparing HLA-A∗02:01 LCL HA-1^H+ vs HA-1^R/R, or HLA-A∗02:06 LCL HA-1^H+ vs HA-1^R/R, unpaired t test). (B) Cytotoxicity assays of anti-HA-1^H TCR-transduced (clone 7) CD8 cells against immunogenic HLA-A∗02:01⁺ HA-1^H/H LCL222 cells or nonimmunogenic HLA-A∗02:01⁺ HA-1^R/R LCL224 cells. (C) These anti-HA-1^H TCR-transduced CD8 cells also killed HLA-A∗02:01⁺ HA-1^H THP1 cells, whereas nontransduced but TCR-edited cells did not. (D) Cytotoxicity assays of anti-HA-2^V TCR-transduced (clone 4) CD8 cells (with endogenous TCR knockdown) against immunogenic HLA-A∗02:01⁺ HA-2^V/V LCL174 cells or nonimmunogenic HLA-A∗02:01⁺ HA-2^M/M LCL177 cells. Data in panel A are representative of at least 2 repetitions for most target LCL lines. Killing assays in panels B-C are representative of at least 3 independent repetitions. Killing assays in panel D are representative of at least 2 repetitions. ∗P, <0.05; ∗∗P, <0.01; ∗∗∗P, <0.001; ∗∗∗∗P, <0.0001; HH, HA-1 H/H homozygous; HR, HA-1 H/R heterozygous; ns, not significant; MM, HA-2^M/M homozygous, RR, HA-1^R/R homogygous; VV, HA-2^V/V homozygous.

Anti-HA-1^H clone 7 TCR-transduced CD8 cells also specifically killed HLA-A∗02:01⁺HA-1^H/H LCL222 cells (Figure 5B) and HLA-A∗02:01⁺ HA-1^H/R THP-1 leukemia cells (Figure 5C), whereas HA-1^R/R HLA-A∗02:01⁺ LCL224 cells were not killed (Figure 5B). Similarly, primary CD8 cells transduced with anti-HA-2^V TCR (clones 1, 4, 7) specifically killed immunogenic HA-2^V/V HLA-A∗02:01 LCL174 cells relative to HA-2^M/M HLA-A∗02:01 LCL177 cells (Figure 5D). Among the 3 anti-HA-2^V TCRs tested in killing assays, clone 7, which had the lowest EC₅₀ in Jurkat reporter cells (Figure 4), had the best killing at the lowest effector-to-target ratio.

Discussion

This study makes 3 important contributions. First, it describes a novel technology to rapidly clone, express, and functionally assess large numbers of TCRs in parallel. Second, this approach is validated by the cloning of TCRs specific for 2 miHAs derived from different biological contexts, including naturally induced memory cells, in vitro expanded cells, and from a patient undergoing GVL after a DLI to treat MDS relapse. The power of the technology is evidenced by the ability to clone and concurrently functionally express multiple TCRs in parallel at high speed and low cost, enabling the screening of many more TCRs than would have previously been possible. Finally, this technology allowed us to more deeply investigate the nature of the miHA-specific T-cell compartment, which is of both basic and applied relevance. Most notable was that both pregnancy and a graft-versus-host response generated T cells bearing TCRs with a broad range of functional avidities. Finally, the TCRs isolated provide the basis for clinical development of TCR that can potentially be used to promote GVL with minimal or no GVHD in the alloSCT context. An ongoing phase 1/2a trial (ClinicalTrials.gov identifier: NCT06704152) is testing the safety of a donor-derived CD8⁺ T-cell product expressing one of the anti-HA-1^H TCR clones described here in in high-risk HA-1^H leukemia patients undergoing an HLA-matched alloSCT with an HA-1^R/R donor.

The development of adoptive T-cell therapies faces challenges in isolating, characterizing, and qualifying appropriate TCRs. More recently, single-cell approaches have enabled the sequencing of TCRα and TCRβ chains of isolated T cells.²³ However, for this approach to test the function of the sequenced TCRs requires TCR gene synthesis and cloning prior to expression in a capable responding cell. In situations wherein reactive receptors are rare, this can require screening many nonreactive receptors, which is costly and time-consuming. The need for high-quality TCRs to redirect the specificities of T cells is not only of importance to alloSCT, but broadly for the development of T-cell therapies against cancers and to treat pathogens, especially in immunocompromised individuals.

An approach for cloning TCRs from single cells using a nested PCR approach and Gibson assembly was previously reported.^24,25 However, there are notable differences that make TCXpress simpler and with a higher throughput. First, the prior approach requires next-generation sequencing of the TCRα amplicons. This is because the first-round amplification does not capture the entire 5ʹ TRAV coding region, and therefore each specific amplified 3ʹ portion of the TRAV must be inserted into 1 of 13 different vectors encoding the TRAV family-specific 5ʹ portion that was not cloned. In contrast, our approach amplifies the entire TCRα and TCRβ variable regions which enables direct Gibson cloning of the second-round amplicons into a single accepting vector, without the need for sequencing.

The high-throughput nature of our system enabled a deeper understanding of in vivo responses against miHAs. In the case of HA-1^H, 20 unique anti-HA-1^H TCRs with a range of avidities were cloned from 100 mL of blood. In contrast, only a single TCR was recovered from an unimmunized donor. Taken together, these data indicate that exposure of the mother to miHA-expressing fetal cells without an adjuvant or other obvious stimulator of innate immunity can engage a diverse polyclonal response. One of the high avidity clones is now in the clinic; had we only been able to clone and characterize a few TCR, it is possible that a TCR with lower avidity, which might be less clinically active, would have been advanced. Indeed, it is possible that some of these TCRs, especially those with low functional avidities, would not direct potent killing in vitro or in vivo.

We also cloned 12 unique anti-HA-2^V TCR from a patient who had received a DLI from his HLA-matched HA-2^M/M donor to treat relapsed MDS. From a sample drawn at GVHD onset, 10 unique TCRs were isolated. The most dominant clone was independently isolated from unexpanded cells 22 times, with other clones isolated 12, 6, 5, and 3 times. These data indicate that coincident with GVHD and the initiation of a process that yielded a durable complete MDS remission, there was a substantial polyclonal anti-HA-2^V CD8 response. Notably, 7 of the day 1 clones were still present at day +140 after DLI, when GVHD and MDS were in remission, and that the ranking of clone frequencies paralleled the ranking at time point 1. Such long-lasting alloreactive clones may have contributed to the depth and durability of the remission.

Another interesting aspect of our results was the diverse range of avidities (among 31 TCRs assayed) observed in responses to single miHAs, even with recurrent TCR family usage. It remains to be determined whether engaging a mix of lower and higher avidity TCRs is advantageous (presumably for antipathogen responses), and thus whether such diversity of avidities has been selected for by evolution. Inflationary T-cell memory responses to cytomegalovirus eventually select for lower-affinity TCRs, and it has been postulated that this benefits the host.^26,27 On the other hand, the generation of TCRs with diverse properties could be a consequence of a repertoire that evolved to recognize a broad set of pathogen-derived epitopes. In our case, we know that a single miHA epitope is driving the expansion of the family of TCR that encompasses a broad range of avidities. In this connection, our TCR cloning and characterization approaches could be employed to efficiently analyze other focused T-cell responses, which would shed further light on this question.

Conflict-of-interest disclosure: S.I. received research support from BlueSphere Bio. M.J.S. and W.D.S. have equity interests in have received research support from and are paid consultants for BlueSphere Bio. S.S., J.R., E.B.C., E.M., C.P., A.B., A.T., R.J., and J.J. are or were employed by BlueSphere Bio. A.M.R. has equity interests in BlueSphere Bio. The remaining authors declare no competing financial interests.