Graft γδ TCR Sequencing Identifies Public Clonotypes Associated with Hematopoietic Stem Cell Transplantation Efficacy in Acute Myeloid Leukemia Patients and Unravels Cytomegalovirus Impact on Repertoire Distribution

Hematopoietic stem cell transplantation (HSCT) has been used for decades to treat a wide range of hematological malignancies, such as acute myeloid leukemia (AML) (1). Although its efficacy has evolved in the last years, the procedure is still associated with substantial transplant-related morbidity, mainly underlying disease relapse, graft-versus-host disease (GvHD), and infections (2). To improve transplantation outcomes, the impact of donor graft T cell subsets on the clinical outcome after HSCT has been studied. Although it has been clearly shown that αβ T cells present in the grafts are the main component behind GvHD development (3–6), little is known about γδ T cells’ influence on HSCT clinical outcomes.

Although comprising around 5% of total peripheral blood T cells, γδ T cells are present in substantially greater numbers in peripheral blood stem cell (PBSC) grafts and can impact clinical outcomes after transplant (7). In matched, unrelated (8, 9), partially mismatched T cell–depleted (10–12) and haploidentical αβ-depleted (13, 14) HSCT settings, increased numbers of γδ T cells in peripheral blood after transplantation are associated with favorable outcomes. This could partly be due to the persistence of donor graft-derived γδ T cells in the peripheral blood of the patients that quickly reconstitute after transplantation (14, 15) and exert a potentially strong antileukemic effect (16–18). This indicates that the number of γδ T cells within the graft and their distribution may help prevent disease recurrence and infections (7, 9). However, the role of these cells remains unclear, and whether HSCT outcomes are influenced by the TCR repertoire of γδ T cells has not been adequately addressed. This emphasizes the need to elucidate which donor graft γδ T cell clones have been infused to the patient during transplantation and how this impacts clinical outcome.

The role of γδ T cells in GvHD development remains controversial. Some studies in mice show that γδ T cells are not primary initiators of GvHD (19, 20) whereas others describe a role for these cells in GvHD pathogenesis (21, 22). In humans, most of the literature shows that the transient increase of γδ T cells after HSCT is not associated with acute GvHD (aGvHD) (10, 11, 23–25), although there are contradictory results (26). In an earlier study, grafts with higher overall γδ T cell content were linked to increased cumulative incidence of aGvHD posttransplant (8). Given the functional plasticity and rich specificity of γδ T cells, in-depth analysis is required to fully unravel the role of these cells in GvHD, especially if specific clonotypes, repertoire characteristics, or usage of gene segments is associated with aGvHD development or protection.

The impact of CMV infection on T cell reconstitution after HSCT has been intensively investigated. CMV reactivation leads to posttransplant immune dysfunction (27) and γδ T cell expansion in the peripheral blood (16, 28, 29). γδ T cells exert anti-infectious activity (16), and CMV infection could be related to TCR γ-chain (TRG) distribution changes. It was recently shown that the peripheral blood TRG repertoire is dominated by the V-J segment V9-JP, presenting quick reconstitution after HSCT and adaptive clonal expansion in response to CMV infection (15). However, little is known about its distribution on PBSC grafts and how CMV can impact the TRG repertoire.

In this study, we used a next-generation sequencing (NGS) platform to analyze the TRG repertoire of γδ T cells within PBSC grafts given to 20 AML patients, evaluating its potential link to clinical outcome as well as to donor CMV status. We hypothesized that the γδ T cell composition of the grafts could be associated with clinical relapse or aGvHD development. Our findings show that the grafts from CMV⁺ donors presented a reshaped TRG repertoire, the intragraft TRG composition is not associated with aGvHD incidence, and several private and public sequences were associated with nonrelapse after HSCT.

Twenty AML patients (7 males and 13 females) who underwent HSCT from 2014 to 2016 at the Centre for Allogeneic Stem Cell Transplantation (Karolinska University Hospital, Huddinge, Sweden) and the respective donors (11 males and 9 females) were enrolled in this study. Patients’ and donors’ characteristics are shown in Table I. The median (range) recipient and donor age at time of enrollment were 59.5 (20–71) and 30 (19–71) y, respectively. All patients received unmanipulated PBSC grafts, comprising 17 grafts from matched, unrelated donors and 3 from matched siblings. All unrelated donors were fully matched for HLA-A, -B, -C, and -DRB1 loci. Patients were followed for a median period of 32.7 mo, and disease relapse was evaluated according to previously described criteria (30). Grading of aGvHD was performed according to the Glucksberg criteria over a 3 mo period post-HSCT. Donor and recipient CMV status was assessed following standard guidelines (31). The study was approved by the Regional Ethical Review Board in Stockholm, Sweden (2008/206-31, 2010/760-31/1, 2013/2215-32, 2017/469-32), and performed according to the Declaration of Helsinki.

PBSC samples were obtained during preparation of the graft for infusion and processed as previously described (3). Mononuclear cells were separated by density gradient centrifugation (800 g, 20 min, Lymphoprep; Fresenius Kabi, Oslo, Norway) and cryopreserved in liquid nitrogen in 10% DMSO in RPMI 1640 medium (HyClone; Thermo Fisher, Waltham, MA) enriched with 10% human AB-serum (Karolinska University Hospital), 100 IU/ml penicillin G (Life Technologies, Paisley, U.K.), and 100 mg/ml streptomycin (Life Technologies).

PBSC mononuclear cell staining was performed as described previously (3). Anti-CD3 (clone UCHT1; BD Biosciences, San Jose, CA) and anti-TCR pan γδ (clone REA591; Miltenyi Biotec, Teterow, Germany) Abs were used to assess the percentage of total T cells and total γδ T cells. Anti-TCR Vγ9 (clone B3; BioLegend), anti-TCR Vδ1 (clone TS8.2; Thermo Scientific), and anti-TCR Vδ2 (clone B6; BioLegend) Abs were used for further graft γδ T cell characterization. Samples were acquired by a FACSCanto instrument (BD Biosciences) using FACSDiva software (BD Biosciences). Gating strategy is provided in Supplemental Fig. 1.

γδ T cells were isolated from the grafts with the Anti-TCRγ/δ MicroBead Kit (Miltenyi Biotec) following the manufacturer’s instructions. Purity was checked by flow cytometry and was higher than 90% for all samples. Immediately after selection, genomic DNA was extracted using the EZ1 DNA Blood Kit and EZ1 instruments (QIAGEN, Hilden, Germany) and stored at −20°C. The extracted DNA (1 μg) was used for survey-level deep sequencing of the γ-chain using the ImmunoSEQ platform by Adaptive Biotechnologies (Seattle, WA). Briefly, bias-controlled V and J gene primers were used to amplify rearranged V-J segments for high-throughput sequencing (32, 33). After correcting sequencing errors via a clustering algorithm, CDR3 segments were annotated according to the International ImMunoGeneTics collaboration, identifying which V and J genes contributed to each rearrangement (34). NGS data are available at the Adaptive Biotechnology ImmunoSEQ site (https://clients.adaptivebiotech.com/immuneaccess, https://doi.org/10.21417/LA122018).

Sequencing data were initially analyzed using the ImmunoSEQ Analyzer (https://clients.adaptivebiotech.com/login). Postanalysis of TRG repertoire diversity, clonal space homeostasis, segment usage, spectratyping, and repertoire overlap were performed using the tcR package (35), VDJTools (36), and VDJviz browser (37).

Patient characteristics and the distributions of the continuous variables are described as median ± interquartile range. The Mann–Whitney nonparametric two-tailed test was used to assess differences between any two group comparisons. The p values <0.05 were considered statistically significant. Statistical analysis was performed using GraphPad Prism version 7 (San Diego, CA) or R Project Package in VDJTools and tcR environments.

HSCT patients were grouped into relapse (n = 8) or nonrelapse (n = 12) groups based on clinical follow-up. All grades of aGvHD (I–III) were grouped (aGvHD Yes, n = 12) and compared with patients who did not develop aGvHD (aGvHD No, n = 8). Donor grafts were divided into CMV⁺ (n = 7) and CMV⁻ (n = 13). Individual patient characteristics are shown in Table I.

The median total T cell frequency (percentage) was comparable between the relapse versus nonrelapse groups (30.5 versus 32.23%) and aGvHD Yes versus aGvHD No (33.55 versus 28.04%). CMV⁺ grafts presented a higher T cell percentage than CMV⁻ ones (43.68 versus 26.7%, p = 0.006). The same profile was observed for the percentage of γδ T cells, with no differences between relapsing or aGvHD incidence groups (9.03 versus 10.07% and 11.59 versus 7.89%, respectively) but a higher percentage in CMV⁺ donor grafts (12.3 versus 7.0%, p = 0.023). Regarding the major γδ T cell subfamilies Vγ9, Vδ1, and Vδ2, no differences were observed between the groups (Supplemental Fig. 1).

Deep sequencing showed similar median total productive templates (i.e., the sum of template counts for all productive rearrangements in the sample) in grafts given to patients experiencing a relapse or not (2.08 × 10⁴ versus 1.06 × 10⁴) and patients with or without aGvHD (1.07 × 10⁴ versus 1.96 × 10⁴, Fig. 1A). The same was observed for the number of unique reads (i.e., number of clonotypes) between the two sets of patient groups. However, grafts given to patients who later developed aGvHD trended toward fewer reads than those who did not develop aGvHD (respectively, median values 3.30 × 10³ versus 2.91 × 10³ and 2.33 × 10³ versus 4.86 × 10³, Fig. 1B).

All groups presented similarly low unique CDR3 TRG ratios (i.e., TCR diversity; 0.004 versus 0.002 and 0.003 versus 0.004, Fig. 1C). The unique CDR3 ratio ranges from 0 to 1, with 0 representing the lowest diversity and 1 representing the maximal diversity. The median ratio found in all samples was 0.003, indicating that the γδ TCR repertoire in the grafts contains oligoclonal TRG sequences and low diversity. A closer TCR inspection by quantile plots revealed that the grafts presented multiple clonal overrepresentations, contributing to a skewed clonotype size distribution (Fig. 1D). To summarize the quantile plots, we assessed the TCR richness. We did not find any differences between the patient groups regarding the Efron–Thisted estimator (0.65 × 10⁴ versus 0.66 × 10⁴ and 0.58 × 10⁴ versus 0.94 × 10⁴, Fig. 1E), Simpson D index (72.58 versus 60.91 and 52.78 versus 75.70, Fig. 1F), or sample productive clonality (0.30 versus 0.29 and 0.29 versus 0.31, Fig. 1G). Altogether, our data suggest that CDR3 TRG diversity in PBSC grafts is overall reduced but does not seem to affect clinical relapse or aGvHD incidence.

We then analyzed the TRG clonal proportion and space homeostasis in PBSC samples (35). The top 10 most abundant clonotypes represented a large part of the repertoire (Fig. 2A), and almost all grafts presented a significant amount of hyperexpanded clones (i.e., clones with frequency ≥1%, Fig. 2B), leading to the reduced diversity. The median proportion of top 10 clones was similar between both relapse and nonrelapse patients (23.01 versus 25.67%) and whether patients experienced aGvHD or not (25% in both) (Fig. 2C).

To address the impact of TRG clonal distribution on patient outcome, we classified clonotypes according to the TCR proportion taken up by clones measured as rare (0–0.001%), small (0.001–0.01%), medium (0.01–0.1%), large (0.1–1%), or hyperexpanded (1–100%). All patients showed few rare clonotypes, further explaining the reduced TRG diversity observed in all samples. Grafts given to patients who did not relapse after HSCT showed a contraction of clones classified as small (respective medians of 7.04 versus 18.86%, p = 0.01) and an expansion of clones classified as medium (32.61 versus 18.23%, p = 0.01) and large (29.38 versus 18.61%, p = 0.04) when compared with relapse patients (Fig. 2D). No differences were observed when looking at the occurrence or absence of aGvHD.

To investigate the profiles of CDR3 lengths in the TRG repertoire, the length of productive nucleotide CDR3 sequences was analyzed for every TRGV/TRGJ gene combination. The length of the CDR3 sequences ranged from 24 to 60 nt, with a median of 45 nt (Fig. 2E). Grafts received by the nonrelapse patient group had an increased proportion of CDR3 with long sequences (54 and 57 nt long) compared with relapse patients (median of 0.41 versus 0.27% [p = 0.02] and 0.11 versus 0.04% [p = 0.04], respectively). Patients who did not develop aGvHD received grafts with more enriched CDR3 sequences of 24 and 33 nt than the non-aGvHD patient group (0.80 versus 0.38% [p = 0.03] and 4.1 versus 3.3% [p = 0.04]). In general, these groups showed a bell-shaped pattern of the CDR3 region (Fig. 2F), characteristic of a Gaussian distribution associated with healthy TCR repertoire distribution. Our results indicate that γδ T cell clonal distribution and TRG CDR3 spectratyping could be related to disease relapse.

CMV infection is associated with γδ T cell expansion and TRG reshaping in the peripheral blood (28). We investigated whether TRG distribution in PBSC grafts was also affected by donor CMV status. The median total productive templates was similar between CMV⁺ and CMV⁻ graft donors (0.54 × 10⁴ versus 1.85 × 10⁴, Fig. 3A). CMV⁺ donor grafts showed a trend toward fewer unique reads than their counterparts, although the difference was not statistically significant (1.39 × 10³ versus 4.03 × 10³, Fig. 3B). The CDR3 TRG ratio was low and similar between the groups (0.003 versus 0.004, Fig. 3C). The quantile plots revealed that in CMV⁺ donor grafts, in addition to the multiple clonal expansions that has been shown in all grafts, a single dominant clone can occupy 21% of the TCR repertoire (Fig. 3D).

Grafts from CMV⁺ donors presented significantly reduced diversity, as shown by having the lowest Efron–Thisted estimator (4.09 × 10³ versus 7.71 × 10³, p = 0.02, Fig. 3E) and inverse Simpson D index (30.70 versus 81.21, p = 0.02, Fig. 3F) and the highest clonality (0.42 versus 0.26, p = 0.006, Fig. 3G). The median proportion of top 10 clones was 2-fold higher in CMV⁺ grafts (51.48 versus 22.17%, p = 0.01, Fig. 3H). Corroborating these data, hyperexpanded clones took up 2.5 times more space in the CMV⁺ grafts (49.33 versus 19.38%, p = 0.007) at the expense of those present in small and medium quantities (4.95 versus 15.32% [p = 0.04] and 20.40 versus 40.00% [p = 0.04], respectively, Fig. 3I).

CDR3 spectratyping revealed that CMV⁺ grafts had a significantly decreased proportion of CDR3 sequences having 24, 27, 42, and 51 nt lengths (0.27 versus 0.61% [p = 0.01], 0.57 versus 1.12% [p = 0.02], 5.11 versus 7.54% [p = 0.007] and 0.61 versus 2.1% [p = 0.04], respectively). CMV⁺ grafts also had an increase of sequences with 30–39 nt, and the increase was significant for sequences having 36 nt compared with CMV⁻ grafts (5.32 versus 4.13%, p = 0.03, Fig. 3J). As a result, whereas CMV⁻ grafts present a Gaussian distribution of the CDR3 region, CMV⁺ grafts presented a non-Gaussian distribution (Fig. 3K), indicating that CMV positivity leads to dominant expansion of clones and subsequent reshaping of the grafts’ γδ TCR repertoire.

Taken together, the above data suggest that CDR3 TRG diversity is significantly reduced in the CMV⁺ PBSC grafts and that CMV may lead to TCR reshaping, with certain CDR3γ clonotypes undergoing CMV-driven clonal expansion.

To further explore the TRG repertoire composition, we performed hierarchical clustering of TRGV and TRGJ segment usage profiles (36). Despite observing two cluster formations based on J2 and JP expression, the grafts presented an overall similar V and J gene usage profile regardless of their clinical outcome after HSCT or donor graft CMV status (Fig. 4A).

We also assessed the usage frequency of individual TRGV and TRGJ families. For all groups, the segments TRGV9 and TRGJP were the most frequent. Patients who did not relapse after HSCT received grafts with a bias toward lower usage of TRGV4, TRGV5, and TRGJP2 segments and higher usage of TRGJP1 compared with relapse patients (median 4.13 versus 7.46% [p = 0.04], 1.12 versus 4.77% [p = 0.02], 3.02 versus 5.52% [p = 0.02], and 6.62 versus 3.23% [p = 0.02], respectively). No differences were observed between the aGvHD and non-aGvHD groups (Fig. 4B). The TCR repertoire in CMV⁺ donor grafts presented a lower TRGV2 and TRGJP expression (6.29 versus 8.99%, p = 0.04, and 30.99 versus 60.63%, p = 0.01) as well as a higher TRGJP1 gene usage compared with CMV⁻ counterparts (11.04 versus 3.90%, p = 0.04).

Changes in V-J segment recombination in CDR3 junctions were then assessed (Fig. 4C) (36, 37). The V9-JP combination was the most frequent pairing in all patients. Nonrelapse patients received grafts with lower usage of the pairings V4-J2, V5-J2, and V8-JP2 (median 3.12 versus 6.57% [p = 0.04], 1.06 versus 4.38% [p = 0.02], and 1.33 versus 1.91% [p = 0.03], respectively) and higher usage of the V2-JP1 pairing (1.28 versus 0.36%, p = 0.03) than relapse patients. No differences were observed between aGvHD groups (Fig. 3D). The TCR usage of the sequence pairs V2-J2, V2-JP2, V4-JP2, V9-JP, and V9-JP2 was lower in CMV⁺ than in CMV⁻ grafts (median 4.62 versus 6.94% [p = 0.04], 0.59 versus 1.33% [p = 0.04], 0.31 versus 0.09% [p = 0.02], 30.75 versus 60.30% [p = 0.02], and 0.07 versus 0.29% [p = 0.01], respectively, Fig. 4D). In summary, we show that γδ T cells from the grafts mostly express TRGV9 and TRGJP segments. Additionally, we identified some significant variations in V and J gene usage and V-J pairing between the patient groups that did or did not relapse.

Next, we checked for TCR overlap and clonotype sharing (35) within the patient groups and whether these could be associated with the observed clinical outcome. Overall hierarchical clustering and multidimensional scaling did not show a clear clustering between the clinical groups (Fig. 5A). Actually, 36–1041 (median of 246) of all the 25,882 unique CDR3 amino acid sequences identified were shared between grafts (Fig. 5B). Therefore, we performed the combined rearrangement analysis of all grafts using the ImmunoSEQ Analyzer, searching for private (i.e., sequences present exclusively in one patient) and public sequences (i.e., sequences shared between at least two patients). The total number of amino acid public sequences was 3886, representing 15.02% of the total number of combined clones between samples. In fact, only one clone (CALWEVQELGKKIKVF) was shared between all samples, whereas the number of clones present exclusively in one sample was 21,996 (84.98%). This indicates that the TRG repertoire in PBSC is mostly composed of private sequences.

Next, we calculated the overlap of full TRG sequences from each patient group pair by the Morisita–Horn similarity index (35). Nonrelapse patients presented a significantly higher Morisita–Horn index than relapse patients (0.67 versus 0.32, p < 0.001), indicating the occurrence of a higher number of public clonotypes shared between nonrelapse patients (Fig. 5C). There was no difference between patients who did or did not develop aGvHD (median 0.43 versus 0.45 respectively). A 4.7-fold lower index was observed in CMV⁺ donor grafts compared with CMV⁻ ones (0.13 versus 0.61, p < 0.0001). This indicates that donor graft CMV positivity leads to a significantly more private TCR repertoire compared with CMV⁻ grafts.

As highlighted in Fig. 5D, the CMV⁺ donor grafts have a notable presence of hyperexpanded clones that can take up to 41.61% of the entire repertoire. These overrepresented clonotypes are not shared between the CMV⁺ grafts and further corroborate the private TRG repertoires observed in these donors. After running a tracking clonotype routine (36) on CMV⁻ grafts, we observed that nonrelapse patients may present a clonotype sharing of up to 23.60% (Fig. 5E). This shows that CMV⁺ donor grafts are associated with private repertoires, whereas patients who did not relapse after HSCT received grafts with more public clonotypes.

We then checked which clonotypes are overrepresented and exclusively present in the different patient groups, as well as the clones shared between patients within the same patient group. We identified four private overrepresented sequences exclusively present in grafts given to nonrelapse patients taking from 2.00 to 6.23% of the TRG repertoire (Table II). All these sequences were longer than 45 nt, corroborating the spectratyping data shown above. Additionally, we identified 12 public clones shared exclusively between the grafts received by nonrelapsing patients.

For patients who did not have aGvHD, we found three preponderant clones in private repertoires and four public sequences shared exclusively between the patients from this group that could be associated with the clinical outcome (Table II), comprising 0.94–9.82% of the TRG repertoire. CMV⁺ grafts presented the highest percentage of repertoire taken by private overrepresented clones identified within the group, ranging from 13.72 to 41.61%. We found one public TCR sequence exclusively present in all patients from this group (CATWDGPYYKKLF) consisting of 39 nt. This indicates that some clones might play a role in clinical outcome after HSCT or in TRG repertoire reshaping in CMV⁺ donor grafts. The identified private and public clones may provide useful information to improve γδ T cell–based therapies in the future.

HSCT is widely applied in clinical practice to treat AML patients, being recognized as one of the most potent curative therapies for this condition (38). To improve transplantation safety, the impact of donor graft T cell composition has been studied and the role of αβ T cells on clinical outcome described (3–6). Research on γδ T cells remains hampered by limited information on their biology and potential role in HSCT (39). To our knowledge, we used for the first time a high-throughput analysis of the γδ T cell TRG repertoire to characterize γδ T cell distribution in the grafts received by AML patients. We show that basic TRG characteristics are not related to aGvHD development and identified public TRG CDR3 sequences that may play a role in AML remission after HSCT. In addition, we describe that CMV positivity is associated with graft TRG repertoire reshaping and overrepresentation of private sequences.

Initial evaluations of TRG repertoire by unique CDR3 ratio showed that CDR3 diversity is reduced overall in PBSC grafts, and there is high presence of multiple clonal expansions. This is in accordance with a recent study showing that γδ T cells from adult peripheral blood present much lower TRG diversity than cord blood samples, a result of age-driven clonal expansion (15). Additionally, a skewed peripheral blood TRG distribution is observed in adulthood (15, 40), indicating that the reduced diversity in the grafts might be related to expansion of TRG clones during life and harvested within the hematopoietic stem cells. Additional data are required to further show the reduced TRG diversity in grafts when compared with cord blood and adult samples. Furthermore, G-CSF used in hematopoietic stem cell mobilization may have affected the distribution and clonality of the TRG repertoire. Despite the fact that earlier literature has shown that G‐CSF mobilization does not alter the total amount of the γδ T cell fraction, which remains unchanged between peripheral blood of healthy donors and PBSC grafts (8), a previous work showed some TRG families expanding, presenting no expression, or having an oligoclonal distribution after mobilization (41). Which TRG clonotypes are expanded or depleted during mobilization should be addressed in future studies. These studies should be performed in the light of different mobilization regimens, such as G-CSF versus plerixafor, considering that different mechanisms of action can be associated with specific cells’ mobilization and potentially associated with posttransplant events.

The reduced diversity and high presence of overrepresented clones was more prominent in CMV⁺ graft donors. These grafts presented high frequency of hyperexpanded clones and non-Gaussian distribution of the TRG CDR3 length when compared with CMV⁻ counterparts, indicating TRG disturbance and reshaping. CMV infection leads to T cell inflation and clonal expansion, resulting in reduced TCR repertoire diversity (27, 42). CMV reactivation after HSCT impacts γδ T cell reconstitution, driving the proliferation of γδ T cell clones (15, 16, 29). There is no information about its influence on graft γδ T cell distribution. In this article, we show that donor CMV positivity is associated with a high clonality, increased frequency of hyperexpanded clones, and a reduced usage of the V9-JP pair, the most abundant rearrangement present in peripheral blood (15, 43). These changes are probably due to adaptive clonal expansion of distinct TRG clones in response to CMV (15). A recent study showed that few private CMV-induced TRG sequences underwent massive proliferation after CMV reactivation post-HSCT (15). Although it is not possible to compare the amino acid sequences with that study because of methodological differences, it did show specific clonotypes taking up more than 60% of the TRG repertoire after CMV reactivation (15). In the same way, we show private, hyperexpanded clonotypes taking up around 40% of the TRG repertoire. This indicates that donor graft CMV positivity is associated with the expansion of CMV-specific clones, reshaping the TRG repertoire.

After HSCT, an increased number of γδ T cells is associated with a better clinical outcome in terms of higher disease-free and event-free survivals (9, 10, 12, 14), making these cells promising candidates for cellular therapy (39). As γδ T cells exert a potent antileukemic effect (16) and can mediate a graft-versus-leukemia effect without causing GvHD (44), ongoing trials are investigating the benefits of grafts enriched in γδ T cell numbers (13, 14) and their role in donor lymphocyte infusion (45). The impact on outcome of γδ T cell donor graft composition is most likely not solely related to the absolute number of γδ T cells present in the graft, underlining the importance of a full qualitative characterization of these cells in the PBSC grafts. TRG NGS allowed us to show that basic characteristics of the repertoire, such as diversity, clonality, and number of reads, are not related to clinical relapse after HSCT. In contrast, we found substantial changes on the occupied homeostatic space, spectratyping, and V and J segment usage, as well as bias in V-J pairing frequencies on grafts given to nonrelapse patients that could be related to nonrelapse disease. Importantly, we also identified public TRG CDR3 amino acid sequences present exclusively in grafts given to patients who did experience clinical relapse after HSCT. Those clonotypes might play a role in clinical improvement after HSCT, given that previous works have shown that γδ T cells exert strong antileukemic effect against AML cell lines in vitro (16–18) and possibly in vivo (46). Alternatively, these clonotypes might reflect hidden features of these grafts with a protective effect against relapse.

Although the selection of grafts with specific TRG clones might be difficult, the TRG sequences identified in this study might be used for the generation of a γδ TCR library that can be used to treat AML relapse. These libraries can be constructed by the rapid and efficient methods currently available for cloning and expression of specific TCRs by T cells, allowing prompt generation of leukemia/Ag-specific cells (47). These clonotypes can be transfected into donor lymphocyte infusion products and used to boost leukemia killing in cases of disease relapse after HSCT, allowing the production of leukemia-specific cellular products (48). In vitro studies are still required to show that the clonotypes described in this article exert high killing of AML blasts and that they can be used as a platform for TCR editing.

Although we did not evaluate the reconstitution of these clonotypes after HSCT, it is known that γδ T cells infused within the graft persist after transplantation and undergo fast immune reconstitution (14, 15). In particular, donor CMV-specific clones can be found in the peripheral blood of HSCT recipients for long times after transplant (49). Altogether, these studies show that clones identified in the graft persist after HSCT, indicating that the clones identified in this study might be related to AML remission. Future works comparing host and donor γδ T cell reconstitution and their role in clinical relapse or aGvHD development are still required.

A controversial study showed that patients receiving PBSC with high γδ T cell levels have both an increased cumulative incidence and higher grades of aGvHD (8). Additionally, G-CSF use in mobilization not only influences the distribution and expression levels of the γδ T cell repertoire but also changes γδ T cell clonality and might play a role in mediating aGvHD (41). Using deep sequencing and in-depth analysis, we show that TRG diversity metrics, V and J family usage, V-J pairing, clonal distribution, and spectratyping were not related to aGvHD development. Additionally, the clonality and TCR overlap were similar between patients who did or did not develop aGvHD, supporting the idea that γδ T cells are not involved in GvHD development (9, 10, 14). Future work to assess γδ T cell phenotype and functionality will help to further clarify their role in GvHD.

Ravens et al. (15) described that TRG repertoires in peripheral blood consist mostly of public sequences. In contrast, our work showed that around 15% of the TRG repertoire is composed of public sequences shared between the PBSC grafts, indicating that a large fraction of graft TRG repertoire consists of private sequences. This change from public TRG on peripheral blood to private TRG on grafts may occur during HSC mobilization by the use of G-CSF, once its use has been associated with TRG repertoire changes from a polyclonal distribution before mobilization to an oligoclonal distribution after mobilization (41). The most prevalent public Vγ9 clonotype (CALWEVQELGKKIKVF) (50), described to be persistent in fetus blood (51), cord blood (52, 53), and adults (34, 52, 54), was the only sequence found in all samples analyzed. This shows the presence of lifelong clonotypes in the graft, although their role is still unknown.

We found five private, overrepresented clones exclusively found in CMV⁺ donors that could be used in the future to develop CMV-specific cell therapies using γδ T cells. Additionally, public sequences were especially found in grafts given to patients who did not relapse after HSCT, sharing a large amount of their repertoire and presenting 12 exclusive sequences. CMV⁺ grafts showed a high presence of private clonotypes, probably due to TRG reshaping, indicating that the donor CMV status could be associated with AML outcome after HSCT. A beneficial effect of CMV reactivation after HSCT has been recently demonstrated, as several studies reported reduced risk of relapse in AML patients who develop CMV reactivation (55–57). Additionally, donor CMV positivity alone was also associated with a lower relapse rate after HSCT (58). The immunological background underlying this complex interplay between CMV and relapse is still unclear and represents an interesting topic for ongoing research. γδ T cells can be an important player in this context. Identifying clones related to CMV infection and clinical remission of AML patients might aid in the choice of suitable γδ TCR sequences for cellular therapy (39, 47, 48). Larger validation cohorts are necessary to confirm our data.

Donor graft αβ T cell characterization has been extensively performed. In summary, grafts containing higher percentages of regulatory and CD4⁺ T cells and lower frequencies of CD27⁺, CD127⁺, CD28⁺, and of naive CD8+ T cells have been associated with reduced aGvHD incidence (3–6) and low relapse rates (4, 5). Oppositely, grafts enriched with γδ T cells were associated with higher aGvHD development but not relapse (8). These cells can interplay and influence clinical outcomes in a complementary or contrasting way (59). Future works evaluating the in-depth αβ TCR composition of the grafts and the detailed phenotype of graft γδ T cells are required to clarify these issues and their joint influence on clinical outcomes.

CMV reactivation, primary disease relapse, and GvHD represent the main causes of transplant-related morbidity and mortality following allogeneic HSCT. The immunological background underlying this complex interplay is still unclear, and γδ T cells can be an important player in this context. To our knowledge, our work provides the first deep sequencing results allowing characterization of γδ T cells infused within the grafts in a matched, unrelated HSCT setting. Using several complementary, in-depth evaluations, we show that none of the γδ T cell TRG repertoire characteristics evaluated (e.g., diversity, clonality, distribution, segment usage, and overlap) were associated with aGvHD development. We also identified public clones associated with nonrelapsing disease that may exert a graft-versus-leukemia effect. This opens the window for large-scale manipulation to enrich, isolate, and expand specific γδ T cell clones to explore the full biology of those cells or to assess their role in HSCT for AML patients. Additionally, we identified several private clones associated with donor graft CMV status. The clonotypes identified in this study could be used in the future to better understand the adaptive responses of γδ T cells and to improve the efficacy and safety of HSCT.

The authors have no financial conflicts of interest.