Research ArticleCancer

MHC Class I–Associated Phosphopeptides Are the Targets of Memory-like Immunity in Leukemia

See allHide authors and affiliations

Science Translational Medicine  18 Sep 2013:
Vol. 5, Issue 203, pp. 203ra125
DOI: 10.1126/scitranslmed.3006061


Deregulation of signaling pathways is a hallmark of malignant transformation. Signaling-associated phosphoproteins can be degraded to generate cancer-specific phosphopeptides that are presented by major histocompatibility complex (MHC) class I and II molecules and recognized by T cells; however, the contribution of these phosphoprotein-specific T cells to immune surveillance is unclear. We identified 95 phosphopeptides presented on the surface of primary hematological tumors and normal tissues, including 61 that were tumor-specific. Phosphopeptides were more prevalent on more aggressive and malignant samples. CD8+ T cell lines specific for these phosphopeptides recognized and killed both leukemia cell lines and human leukocyte antigen–matched primary leukemia cells ex vivo. Notably, healthy individuals showed robust CD8+ T cell responses against many of these phosphopeptides within the circulating memory compartment. This immunity was significantly reduced or absent in some leukemia patients. This reduction correlated with clinical outcome; however, immunity was restored after allogeneic stem cell transplantation. These results suggest that phosphopeptides may be targets of cancer immune surveillance in humans, and point to their importance for development of vaccine-based and T cell adoptive transfer immunotherapies.


Hematologic malignancies are susceptible to curative immunological therapies, such as stem cell transplantation (SCT), donor lymphocyte infusion, and manipulation of endogenous immunity through immunostimulatory cytokines (14). The clinical utility of cellular immunotherapies provides powerful direct evidence that the adaptive immune response can control and eradicate tumors (59). However, immunotherapies targeting identified tumor antigens have met with limited success, suggesting that the optimal antigens remain undiscovered (1012). Moreover, the vast majority of identified tumor antigens are not derived from oncoproteins orchestrating the transformation process. Efforts targeting these antigens may be compromised by tolerance mechanisms and immune escape.

The importance of signal transduction pathway deregulation in cancer pathogenesis is well established (1317). Small-molecule therapies targeting these pathways have met with considerable clinical success, providing a powerful argument for immunotherapies that target similar deregulated pathways (1820). Protein phosphorylation is the dominant mechanism involved in oncogenic signaling processes (21). We previously showed that phosphorylation is preserved on peptides during antigen processing for presentation by both major histocompatibility complex (MHC) class I and II molecules (2224). This suggests that phosphopeptide antigens derived from cancer-related phosphoproteins could serve as immunological signatures of “transformed self” (2226). Moreover, phosphorylation can enhance the binding of peptides to MHC class I molecules, creating “neoantigens.” Collectively, these results suggest that phosphopeptides are attractive targets for cancer immunotherapy (27).

To date, no studies have examined MHC class I–bound phosphopeptides displayed on primary human tumor samples, nor has the human immune response against MHC class I–restricted phosphopeptides been characterized. Here, we identify phosphopeptides from primary leukemia samples and show that healthy individuals display immune responses with memory characteristics against many of them. Immunity is lost in some patients with leukemia, but can be restored. This unexpected preexisting immunity to phosphopeptides suggests that they could be targets of immune surveillance and play a role in enabling immunological control of malignancies in humans.


Characterization of leukemia-associated MHC class I–restricted phosphopeptides

To identify tumor-associated phosphopeptides, we affinity-isolated human leukocyte antigen (HLA)–A*0201 (HLA-A2) and HLA-B*0702 (HLA-B7) peptide complexes from four primary chronic lymphocytic leukemia (CLL) tumors, a primary hairy cell leukemia (HCL), a primary mantle cell lymphoma (MCL), a primary acute lymphoblastic leukemia (ALL), and a primary acute myeloid leukemia (AML), in addition to normal splenic T and B cells, bone marrow (BM) cells, and cultured B lymphoblastoid cell lines (B-LCLs). Collectively, 10 HLA-A2–restricted and 85 HLA-B7–restricted phosphopeptides were identified (Fig. 1A). All tumor types and normal tissues expressed a greater number of HLA-B7 than HLA-A2 phosphopeptides (Fig. 1B).

Fig. 1. HLA-bound phosphopeptides are differentially displayed on tumor and matched healthy tissue.

(A) Phosphopeptides isolated from HLA-A2 and HLA-B7 molecules from 11 primary tumor samples (ALL1, AML1, CLL1 to CLL4, HCL1, and MCL), EBV-transformed B cells (B-LCLs), and HLA-matched healthy tissue (T cells, B cells, and BM). Red and green dots indicate phosphopeptide antigens selected for further study in patients with CLL and AML, respectively. (B) Comparison of the number of individual phosphopeptides identified between HLA-A2 and HLA-B7 in both normal and malignant tissues. (C) Comparison of the number of phosphopeptides identified between normal, indolent malignant, and aggressive malignant tissues.

The large number of HLA-B7–restricted phosphopeptides enabled us to compare their representation on different tumor types and healthy tissue. On average, more than twice as many were found on aggressive (AML and ALL) as on indolent (CLL and HCL) tumors or normal tissue (Fig. 1C). Of 56 HLA-B7–restricted phosphopeptides identified on AML, 36 were not found on any other leukemia, and only two were found on BM (Figs. 1A and 2A). Of 45 phosphopeptides identified on ALL, 19 were not found on other leukemias, and 13 of 19 were also not observed on T or B cells (Fig. 2A). Seven were derived from oncogenes implicated in leukemogenesis: MYC, EP300, SKI, GFI-1, Bcl-11A, MEF2D, and MLL (Fig. 1A). Twenty-seven HLA-B7 phosphopeptides were identified on two CLL tumors, all of which were shared with either AML or ALL. Sixteen were observed on normal tissue, but seven were common to AML, ALL, and CLL and not normal tissue [CCDC45, GRK2, SETD2, C17orf85, TSC22D4, and SPR(pS)PGKPM, derived from an unknown protein]. These proteins have not been previously associated with leukemic malignancies, and the functions of most remain undefined. Twenty-six of 27 CLL peptides were also found on HCL, which itself expressed one additional phosphopeptide. Twenty-four of these were also shared by B-LCLs (Fig. 2B). These results identify a cohort of phosphopeptides expressed on multiple leukemic malignancies, but not normal tissue, which represent potential immunotherapeutic targets.

Fig. 2. HLA-B7–associated phosphopeptides have characteristics that distinguish them from nonphosphorylated peptides.

(A and B) Euler diagrams depicting the distribution of HLA-B7–restricted phosphopeptides among different leukemias and normal tissues (A) and within different B cell malignancies (B). (C, D, and F) Logo plots of residue frequency at each position of all 9-mer HLA-B7 phosphopeptides (C), 9-mer nonphosphorylated HLA-B7 peptides from the Immune Epitope database (d), and predicted HLA-B7 phosphopeptide binders with a pSer at position 4 (F). (E) Position of phosphoserine for all B7-predicted binders.

Characteristics of the HLA-A2–bound phosphopeptides were similar to those previously reported (22, 23). Of 85 HLA-B7–restricted phosphopeptides, 2 were dually phosphorylated and the remainder monophosphorylated, with 77 of 85 containing phosphoserine and 9 of 85 containing phosphothreonine. The phosphate was found at position 4 in 72% of HLA-B7 phosphopeptides, similar to the distribution in HLA-A2 phosphopeptides (fig. S1, A and B) (27). When compared with 1038 nonphosphorylated HLA-B7–restricted 9-mer peptides in the Immune Epitope database (28), HLA-B7–restricted phosphopeptides showed a similar strong preference for proline at P2 and common hydrophobic C-terminal anchor residues (Fig. 2, C and D). However, they showed an unusual bias for basic residues at P1 and proline at P5 (Fig. 2C and fig. S1C). To test whether these biases were imposed by an underlying kinase recognition motif, we identified potential HLA-B7–binding 9-mer peptides in the PhosphoSite data set (29) of known serine phosphorylation sites. The position of the phosphoserine within 1031 phosphopeptides predicted to bind to HLA-B7 was not skewed toward P4 (Fig. 2E). Thus, these biases likely reflect roles in binding to HLA-B7 analogous to those demonstrated for HLA-A2–associated phosphopeptides (27). Of 164 predicted HLA-B7–binding peptides that had phosphoserine at P4, there was no apparent bias for basic residues at P1 (Fig. 2F). However, 87% contained a proline at P5, suggesting that this bias reflects an underlying kinase motif, rather than being imposed by HLA binding.

Phosphopeptide-specific T cell responses in healthy donors

Three HLA-A2 and two HLA-B7 phosphopeptides were derived from LSP-1, a lymphoma marker (30). One, RQA(pS)IELPSMAV, is present on all HLA-A2+ tumor samples at high copy number. RQA(pS)IELPSMAV-pulsed dendritic cells (DCs) were used to prime autologous T cells. Responses to this phosphopeptide could be elicited in three of three healthy individuals. Specific T cells were enriched using HLA-A2 RQA(pS)IELPSMAV tetramers to produce T cell lines (Fig. 3A). These lines secreted interferon-γ (IFN-γ) in response to stimulators pulsed with RQA(pS)IELPSMAV, but not with unphosphorylated RQASIELPSMAV, or to other phosphopeptides, including the closely related RQA(pS)IELPSM (Fig. 3B). Recognition was therefore both phosphate-dependent and peptide sequence–specific. These lines also killed the HLA-A2+ AML cell line THP-1 and the HLA-A2–transfected ALL cell line Jurkat-A2, but not untransfected Jurkat (Fig. 3, C and D). RQA(pS)IELPSMAV-specific T cells killed HLA-A2+ primary AML and CLL tumors, but not an HLA-A2 CLL tumor (Fig. 3E). T cell lines were also elicited against an HLA-B7–restricted phosphopeptide [RPT(pS)RLNRL] derived from NCOA-1 and showed remarkable phosphopeptide-specific killing of three primary CLL tumors (Fig. 3F).

Fig. 3. Phosphopeptide-specific T cells from healthy donors recognize and kill leukemic targets.

(A) Expansion of LSP-1 [RQA(pS)IELPSMAV]–specific T cells from healthy donor PBMCs using DCs and tetramer selection. (B and C) Peptide antigen specificity (B) and recognition (C) of leukemia cell lines by LSP-1–specific T cells. DMSO, dimethyl sulfoxide. (D and E) T cell–mediated killing of leukemia cell lines (D) or primary leukemia samples (E) by phosphopeptide-specific T cell lines but not an HLA-mismatched primary tumor. (F) NCOA-1 [RPT(pS)RLNRL]–specific T cell–mediated killing of three HLA-matched CLL tumor samples was inhibited by adding autologous NCOA-1–loaded cold targets.

We next evaluated immunity in 10 healthy donors against 10 HLA-A2 and 76 HLA-B7 leukemia-associated phosphopeptides (Fig. 4A). Immune responses to 50 of 76 HLA-B7 and 9 of 10 HLA-A2 phosphopeptides were observed. Each individual responded to an average of 16 of 76 (range, 11 to 22) HLA-B7 and 2 of 10 (range, 0 to 4) HLA-A2 phosphopeptides. These responses were observed in 7-day in vitro cultures without addition of exogenous cytokines. No responses of a similar magnitude were observed for several phosphopeptides encoded by well-established leukemia oncogenes, most notably MYC, BCL-11A, and EP300. However, the MYC and BCL-11A phosphopeptides were also present on normal T cells (Fig. 1A). Phosphopeptides from other leukemia oncogenes (MLL, LPP, SKI, GFI-1, and MEF2D) elicited strong immune responses in several individuals. There was substantial donor-to-donor variation as to which phosphopeptides stimulated these unusually strong responses: The GFI-1 phosphopeptide stimulated responses in 9 of 10 individuals; the MLL, LPP, and MAP3K11 phosphopeptides in 8 of 10; and those from LSP-1, SCAP, and RBM15 in 7 of 10. For four phosphopeptides that were recognized by most healthy donors, T cell specificity was confirmed in 14-day cultures using HLA-phosphopeptide multimers (Fig. 4B). Furthermore, these T cells did not recognize the unphosphorylated counterpart peptides (Fig. 4C and fig. S2). Collectively, these data demonstrate that most of the leukemia-associated phosphopeptides elicited surprisingly strong and specific responses in a significant fraction of healthy individuals.

Fig. 4. Phosphopeptide-specific immunity is present in healthy individuals within the circulating memory T cell compartment.

(A) Marked heterogeneity of phosphopeptide-specific immunity in 10 healthy donors against 86 antigens revealed by 7-day enzyme-linked immunospot (ELISpot) assays. (B) ELISpot results for immunodominant phosphopeptides directly ex vivo and after 14-day in vitro expansion when tetramer binding is evident. (C) Expanded T cells do not recognize the unphosphorylated counterpart peptides. SFCs, spot-forming cells. (D) Relative magnitude of phosphopeptide-specific T cells compared against common viral epitopes after overnight culture (n = 3). (E) Analysis of memory flow-sorted CD8+ enriched T cells by 7-day ELISpot reveals that phosphopeptide-specific responses reside in the memory compartment.

Leukemia-associated phosphopeptide-specific immunity in healthy donors

Because responses to phosphopeptides were observed after only 7 days of in vitro culture in the absence of exogenous cytokines, we compared them with responses to immunodominant epitopes from three persistent viruses [cytomegalovirus (CMV), Epstein-Barr virus (EBV), and herpes simplex virus (HSV)] and two nonpersistent viruses (influenza and adenovirus). Responses against four phosphopeptides were even evident in ex vivo ELISpot analysis of peripheral blood mononuclear cells (PBMCs) from some donors (Fig. 4, B and D). These responses were similar to, or higher than, those to both nonpersistent viral epitopes, but lower than those to CMV and EBV epitopes. This high level of phosphopeptide-specific T cells led us to investigate whether the responding cells resided within memory or naïve T cell compartments. Enriched CD8 T cells from peripheral blood of two HLA-B7+ healthy donors were flow-sorted into naïve (TN), central memory (TCM), effector memory (TEM), and terminal effector memory cells (TEMRA) on the basis of expression of CD45RA and CD27 (31). After 7-day in vitro culture, ELISpot analysis demonstrated that T cells responding to individual phosphopeptides were exclusively in the memory compartment and predominantly had a TCM phenotype (Fig. 4E and fig. S3). This suggests that most of the healthy individuals have been previously exposed to a stimulus that establishes immunological memory to tumor-associated phosphopeptides.

Absent phosphopeptide-specific immunity in leukemia patients

We next evaluated the level of phosphopeptide-specific CD8 T cell immunity in immunocompetent patients with early-stage CLL (table S1). Because tumor cells are present at high levels in PBMCs from CLL patients, comparison with responses in PBMCs from healthy individuals is problematic. Therefore, responses against a group of 12 CLL-associated phosphopeptides (identified by red dots in Fig. 1) were assessed by direct ex vivo ELISpot of purified CD8 T cells (fig. S4). As expected, we detected immune responses to most of these phosphopeptides in all 10 healthy HLA-B7+ individuals (Fig. 5A). However, 9 of 14 CLL patients (group 1) had low or absent immunity to all 12 phosphopeptides (less than a combined total of 10 spots for all 12 phosphopeptides/200,000 CD8 T cells). Five CLL patients (group 2) showed breadth of recognition of different phosphopeptides similar to that of the healthy donors. However, average responses to all phosphopeptides were significantly lower in group 2 patients than in healthy individuals (Fig. 5B).

Fig. 5. Leukemia-associated phosphopeptide-specific immunity is lacking in CLL patients.

(A) Comparative ELISpot analysis of enriched CD8+ T cells from patients with CLL and healthy donors against 12 CLL-associated antigens reveals two distinct groupings. (B and C) Analysis of the average patient response against each phosphopeptide between the groups and healthy donors (HD) (B) reveals that both group 1 (n = 12) and group 2 (n = 12) have suppressed phosphopeptide immunity, yet responses to mitogens are intact (C). (D to F) Kaplan-Meier analysis of overall survival (OS) (D), progression-free survival (PFS) (E), and time to first treatment (TTFT) (F) is reduced in group 1 patients, but not statistically significant. **P < 0.01 by Student’s t test comparing average responses for each phosphopeptide between healthy donors and either group 1 or group 2.

This lack of phosphopeptide-specific immunity in CLL patients might have been due to an overall depression of T cell immunity. However, bulk CD8 T cell responses to low-level anti-CD3 were comparable in groups 1 and 2 patients and healthy donors (Fig. 5C). To assess whether the lack of phosphopeptide-specific immunity in group 1 patients reflected T cell anergy, we analyzed patient responses to the pNCOA-1 phosphopeptide, which is immunogenic in 9 of 10 healthy donors, in the presence of interleukin-2 (IL-2) because this cytokine has been shown to reverse the anergic state (32). Responses to anti-CD3 increased in the presence of IL-2, but no anti-phosphopeptide immunity was detected (fig. S5). This suggests that preexisting phosphopeptide immune T cells in group 1 patients had been deleted, or had never developed, rather than being anergized.

Although CLL is a less aggressive tumor than either AML or ALL, the disease is more aggressive in a subgroup of patients. Thus, we assessed whether the level of phosphopeptide-specific immunity was associated with patient outcome. Despite selecting patients with early-stage disease, there were large differences in progression-free survival, overall survival, and time to first treatment between group 1 and group 2. Group 2 patients survived longer and required less treatment than group 1 patients (Fig. 5, D to F). However, because of the small study size, these differences did not reach statistical significance.

We extended these studies to a cohort of 12 HLA-B7+ AML patients in complete remission (table S2) using a panel of 12 AML-specific phosphopeptides (green dots in Fig. 1). Responses to most of these phosphopeptides were detected in 12 healthy donors after 7-day in vitro culture (Fig. 6A and fig. S6). However, as with CLL patients, the average responses of AML patients to individual phosphopeptides were significantly lower (Fig. 6B, P < 0.0001). AML patients could also be stratified: Group 1 patients (10 of 12) showed a profound lack of phosphopeptide-specific immunity, whereas group 2 (2 of 12) showed responses that were similar to those of normal donors. Again, responses to low-dose anti-CD3 were similar between the groups, indicating that the lack of preexisting anti-phosphopeptide responses in patients is not due to immune incompetence (fig. S7). In keeping with this, absolute lymphocyte counts for all 12 AML patients were within the normal range before transplant (table S2). Collectively, these findings demonstrate that most of the patients with either CLL or AML have reduced preexisting immunity specific for leukemia-associated phosphopeptides.

Fig. 6. Phosphopeptide-specific immunity is lacking in patients with AML and restored after SCT.

(A) AML-associated phosphopeptide-specific immunity is lacking in AML patients in remission compared against healthy individuals using a panel of 12 phosphopeptide antigens. (B) Analysis of the average patient response against each phosphopeptide (n = 12) between the groups and healthy donors reveals patients have suppressed immunity. (C) Analysis of total immunity against all 12 antigens between healthy donors (n = 12) and each patient before and after transplant (n = 12) reveals the recovery of AML-associated phosphopeptide immunity after SCT. (D and E) Immune reconstitution of donor anti-phosphopeptide immunity in two AML patients demonstrating large expansions of phosphopeptide-specific T cells. (F) T cell line specific to MLL(EPR) generated from patient AML4 can kill the HLA-B7–transfected KG-1a leukemia cell line. ***P < 0.001, **P < 0.01 by Student’s t test.

Restoration of phosphopeptide-specific immunity after SCT

A graft-versus-leukemia (GvL) response after allogeneic SCT correlates with a positive clinical outcome in AML patients. Although it is believed that GvL is directed to minor histocompatibility antigens (mHAgs), the targets have been only partly identified (32). We hypothesized that SCT might also reconstitute potentially protective immunity against leukemia-associated phosphopeptides in AML patients. All 12 AML patients went on to have allogeneic SCT. Immunity against the 12 AML-specific phosphopeptides was at least partly restored in most of the patients studied (Fig. 6C and fig. S6). Marked expansion of immune responses to six different phosphopeptides was observed in patient AML2 (Fig. 6D), whereas patient AML4 showed marked expansion of responses to MLL(EPR) and LSP-1(QPR) phosphopeptides, and modestly increased responses to C17orf85 and SKI (Fig. 6E). Some of these responses were as large as those against an immunodominant CMV epitope (pp65), and the resulting T cells bound HLA-phosphopeptide tetramers. Furthermore, in vitro–expanded MLL(EPR)-specific T cells from patient AML4 killed AML cells (Fig. 6F), confirming their functional relevance. These data reveal a linkage between immunity to leukemia-associated phosphopeptides and GvL after allogeneic SCT.


Posttranslationally modified antigens are increasingly being shown to play important roles in human disease (3337). Here, we have tested the hypothesis that phosphopeptide neoantigens could be identified on primary human malignant tissue and would be immunogenic in healthy donors or patients with cancer. We now report 95 phosphopeptides displayed on the surface of primary hematological malignant tissue in association with two dominant human MHC class I molecules. We observed substantial differences in the number of phosphopeptides presented by different HLA alleles, although the underlying mechanism for these differences is unclear. Our results establish that MHC class I–displayed phosphopeptides are overrepresented on multiple leukemic malignancies and that more aggressive malignancies display a greater diversity of them. Many of these tumor-associated phosphopeptides are derived from oncogenes linked to leukemogenesis, making these of particular interest as immunotherapeutic targets.

Unexpectedly, many of these phosphopeptides were the targets of preexisting immunity, based on high levels of responding CD8 T cells with a predominantly central memory phenotype. The levels of responding cells were similar to those directed against immunodominant epitopes from some nonpersistent viruses. These T cells, when expanded in vitro, bound to HLA-B7–phosphopeptide tetramers, and recognized and killed primary tumor cells. These observations contrast with those reported for other tumor-associated antigens. Immunity to cancer-testis antigens is generally not present in healthy individuals and becomes detectable in patients with cancer, but is associated with a poor prognosis (38, 39). Immunity to tissue-associated differentiation antigens, particularly those defined as targets for melanoma-specific T cells, is also elevated in melanoma patients (40, 41) but is believed to be compromised by self-tolerance (42, 43). None of the healthy donors in which these memory compartment–associated immune responses were evident showed any signs of autoimmune disease. This suggests that these phosphopeptides are displayed on normal tissue at levels that are substantially lower than those on tumors. In keeping with this, T cells specific for LSP-1 (RQAV) and NCOA-1 recognize tumors, but not resting B cells, despite the presence of these epitopes on the latter cells.

Although a small study, these findings suggest that most of the healthy individuals have been previously exposed to a stimulus that establishes immunological memory to substantial numbers of phosphopeptides. For a small number of these phosphopeptides, immunological memory is prevalent among most individuals examined. The mechanism that underlies the development of phosphopeptide-specific memory CD8 T cells is of great interest. Although it is possible that a distinct, cross-reactive stimulus could be responsible for the development of a generic phosphate-directed immunity, all of the T cells studied to date are specific for the phosphate, the specific peptide sequence, and the MHC molecule that presents it. On the other hand, hematolymphoid transformation by EBV is ubiquitous in humans, and T cell–immunosuppressed individuals commonly develop EBV-related malignancies (44). This points to ongoing immune surveillance directed against virally transformed, and not just virally infected, cells. Although much of this is directed at latent EBV gene products, T cells directed against EBV-transformed cells also recognize EBV-negative tumors (45). We hypothesize that phosphopeptides that are shared by different kinds of malignancies are among the targets of these T cells. In addition, molecular techniques have shown that 69% of healthy individuals have detectable myeloid-transforming transcripts in peripheral blood (46), and 5 to 12% of healthy individuals show a monoclonal B cell lymphocytosis, a recognized precursor of CLL (47, 48). Finally, because some of these phosphopeptides are also found on cell lines from solid tumor malignancies (22), immune surveillance against a broad range of early-stage nonhematopoietic cancers may also lead to the development of phosphopeptide-specific immune memory. The extensive variation among individuals in which phosphopeptides are the targets of this preexisting immunity is consistent with immune surveillance operating on the individual tumors that arise in each person. Although it is difficult to prove definitively in humans, we hypothesize that phosphopeptide-specific memory cells are evidence of previous encounters with nascent tumors that have deregulated some of their phosphorylation-based signaling cascades.

Most of the patients with CLL and AML lacked evidence of preexisting immunity to substantial subsets of phosphopeptides, and the clinical outcome was more favorable in CLL patients where phosphopeptide immunity was present. It is possible that the lack of phosphopeptide-specific immunity reflects a lack of exposure during the immune surveillance phase preceding tumor development. Alternatively, it may reflect the operation of immunosuppressive mechanisms, or the occurrence of clonal exhaustion, in conjunction with tumor outgrowth. Regardless of the exact mechanism, the results point to a role for phosphopeptide-specific immunity as a component of tumor recognition and control.

Although the antigenic targets for the GvL response include mHAgs (4951), multiple studies have revealed that SCT between identical twins, where mHAg differences are not present, can lead to curative outcomes (5254). In AML, deficient phosphopeptide-specific immunity could be restored after adoptive transfer of T cells from a stem cell donor. This indicates that immunity against phosphopeptide tumor antigens may define a component of the GvL response (1, 5). It is known that survival of CMV+ patients can be enhanced by preselecting donors for SCT who have preexisting CMV-specific immunity (55). It is therefore possible that SCT outcome may also be improved by matching donors with preexisting immunity to particular phosphopeptides based on the display of phosphopeptides on leukemia cells.

In this present study, we have been unable to demonstrate a direct causal relationship between the lack of cellular immunity against phosphorylated antigens in patients, and disease development and/or progression. Neither have we been able to connect, at the molecular level, deregulated oncogenic signaling present in each primary tumor to the phosphorylation events present within the defined antigens. Larger, more complex, and extensive studies will be needed to define these relationships. Collectively, however, our results suggest that the display of MHC-associated phosphopeptides may play a role in preventing either the development or the progression of malignant disease. Thus, enhancing immunity to these tumor-associated antigens should become a focus of future cancer immunotherapeutic strategies.


Tumor samples and cell lines

Blood or leukapheresis samples were taken from patients with high-burden leukemia in heparin, and tumor cells were isolated on Ficoll density gradients. The purity of the tumor was >98% pure in all cases as assessed by flow cytometry. Healthy T and B cell populations were isolated from normal spleen, and tonsil samples were processed by mechanical disruption, followed by density gradient separation and enrichment with anti-CD19 or anti-CD3 microbeads (Miltenyi Biotec) to >98% purity. BM sample was obtained from an elective orthopedic procedure and red cell–depleted with hypotonic lysis. All cell lines were grown at 37°C with 5% CO2 in medium consisting of RPMI 1640 supplemented with 10% fetal bovine serum (FBS) and 2 mM l-glutamine (all from Sigma-Aldrich).

Isolation of HLA-associated peptides

Class I MHC molecules were immunoaffinity-purified from samples, and their associated peptides were extracted, as described (22). Briefly, 1.2 × 109 to 14 × 109 cells were lysed in 20 mM tris-HCl (pH 8.0), 150 mM NaCl, 1% CHAPS, 1 mM phenylmethylsulfonyl fluoride, aprotinin (5 μg/ml), leupeptin (10 μg/ml), pepstatin A (10 μg/ml), calyculin A (1 μg/ml) (Sigma-Aldrich), and phosphatase inhibitor cocktails I and II (Sigma-Aldrich). For primary tumors or the HLA-A2, B7 homozygote B-LCL JY, the mixture was centrifuged at 100,000g for 1 hour, and the resulting supernatant was passed over protein A–Sepharose preloaded with the HLA-A2–specific antibody BB7.2 or HLA-B7–specific antibody ME1. For the HLA-A3, B7 homozygous B-LCL GM03107 columns were loaded with the HLA-B, C–specific antibody B123.2 to recover HLA-B7 molecules. Peptides were eluted from the purified class I MHC molecules with 10% acetic acid and separated by ultrafiltration (Ultrafree-MC, Millipore).

Sequence analysis of HLA-associated phosphopeptides

Immunoaffinity-purified class I peptides were converted to d0- or d3-methyl esters and subjected to Fe+3-immobilized metal affinity chromatography to isolate phosphopeptides, as described (22). Phosphopeptide methyl esters were then analyzed by a combination of nanoflow high-performance liquid chromatography (HPLC), microelectrospray ionization, and collision-activated dissociation on LTQ/FT or Orbitrap tandem mass spectrometers (Thermo Scientific) (22). Electron transfer dissociation (ETD) spectra were acquired on an in-house modified LTQ mass spectrometer (56). Peptide sequences were determined by manual interpretation of CAD and ETD spectra recorded on the above peptide esters. If necessary, phosphopeptide sequences were confirmed by recording tandem mass spectra on the corresponding synthetics.

Epitope prediction

Epitope prediction was performed with SYFPEITHI and a threshold score of 20.


Peptides used in this study were synthesized with Fmoc chemistry, isolated by HPLC to >90% purity, and validated with mass spectrometry (EZBiolab and GenScript).

HLA-phosphopeptide tetramers

HLA tetramers were produced as described (57).

Generation of human phosphopeptide-specific CD8 cytotoxic T cell lines

HLA-A2–restricted RQA(pS)IELPSMAV phosphopeptide-specific cytotoxic CD8 T cells were generated from healthy donors as described previously (58). All cytokines were from PeproTech except where stated. Briefly, PBMCs were cultured in flat-bottom six-well plates at 107 cells per well in RPMI 1640 plus 10% heat-inactivated human AB serum (Biosera) (10% medium). Granulocyte-macrophage colony-stimulating factor (GM-CSF) (800 IU/ml) and IL-4 (1000 U/ml) were added on day 0 to generate DCs. On day 1, a maturation cocktail containing tumor necrosis factor–α (TNF-α) (100 ng/ml), IL-1β (100 ng/ml), IL-6 (10,000 IU/ml), GM-CSF (8000 IU/ml), and prostaglandin E2 (PGE2) (10 μg/ml) (Sigma-Aldrich) was added. On day 2, DCs were harvested and loaded with phosphopeptide (20 μg/ml) for 4 hours in the absence of FBS or human serum. After three washes, T cells were added at a ratio of five T cells per DC in 10% medium. Recombinant IL-7 (10 ng/ml) and IL-15 (10 ng/ml) were added on day 5. On day 9, T cells were harvested and restimulated by adding 107 irradiated PBMCs and 106 irradiated autologous LCLs pulsed with phosphopeptide with IL-7 (5 ng/ml), IL-15 (5 ng/ml), and IL-2 (20 IU/ml). Cultures were restimulated every 7 days thereafter in the same manner. At each restimulation, T cells were enriched either with anti-CD8 microbeads (Miltenyi Biotec) or by labeling with HLA-phosphopeptide tetramers and with anti-PE (phycoerythrin) microbeads (Miltenyi Biotec).

HLA-B7–restricted anti-phosphopeptide T cells were grown in the absence of DCs by plating 5 × 106 PBMCs in 48-well plates in 10% medium with individual phosphopeptides at 10 μg/ml for 7 days without cytokines. Restimulations with irradiated phosphopeptide-pulsed autologous PBMCs took place every 7 days with cytokines added 3 days after each restimulation (final concentration: IL-2, 20 IU/ml; IL-7, 5 ng/ml; IL-15, 5 ng/ml). Functional and cytotoxicity assays were then performed from day 13.

T cell recognition assays

For ELISpot analysis, PBMCs or CD8 T cells were freshly isolated from heparinized blood taken from healthy donors and patients. PBMCs (1 × 106) were isolated from both AML patients and healthy donors and resuspended in AIM-V medium (Invitrogen) with 10% human AB serum (Biosera) in a 96-well plate. For 7-day assays, peptide or phosphopeptides were added individually at 10 μg/ml and placed at 37°C in CO2 incubator for 7 days. For some experiments (Fig. 4B), the 7-day culture was restimulated by adding irradiated phosphopeptide-pulsed autologous DCs for a further 7 days. For the ELISpot, cells were then harvested, washed four times in AIM-V, and incubated for 16 hours with either phosphopeptides (10 μg/ml), peptides (10 μg/ml), or anti-CD3 (OKT3, 100 ng/ml, Mabtech). Cytokine-producing cells were identified as dark spots after a 15-min reaction with 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium by means of an alkaline phosphatase conjugate substrate (Mabtech). Spots were counted with an automated reader (AID Diagnostika), and results were displayed as number of spot-forming cells per 105 CD8 T cells or 106 PBMCs.

For CLL patients, a different approach had to be taken because PBMCs contain largely tumor cells. Therefore, CD8 T cells were magnetically enriched with anti-CD8 microbeads (Miltenyi Biotec) to a purity of >99% for both healthy donors and patients with CLL. CD8 T cells (200,000) from both patients and healthy donors were used for ELISpot analysis as described above.

Cytotoxicity assays

Cytotoxic activity was evaluated in a standard 4-hour 51Cr release assay, as previously described (57).

Statistical analysis

We analyzed the data with the Kaplan-Meier for the survival rate and the unpaired t test for two-group comparisons. We performed the statistical analyses with Prism version 5 (GraphPad). P values <0.05 were considered significant.

Ethical approval

The University of Virginia Institutional Review Board approved all protocols. Patients with CLL were recruited from specialist clinics at the University Hospital NHS Trust and Heart of England NHS Trust (Birmingham, UK). The study received approval from the local ethics committees at South Birmingham, Birmingham East, North, and Solihull, and informed written consent was obtained in accordance with the Declaration of Helsinki in all cases.


Fig. S1. Detailed HLA-binding motif analysis.

Fig. S2. Immunity against phosphopeptides in healthy donors is phosphate-dependent.

Fig. S3. T cell memory subset mapping of anti-phosphopeptide responses in healthy donors.

Fig. S4. Leukemia-associated phosphopeptide-specific immunity is lacking in CLL patients.

Fig. S5. Deletion (rather than anergy) of anti-phosphopeptide immunity in five patients with CLL.

Fig. S6. Leukemia-associated phosphopeptide-specific immunity is lacking in AML patients but restored after SCT.

Fig. S7. Immunocompetence of patients with AML in complete remission before transplantation.

Table S1. CLL patient characteristics.

Table S2. AML patient characteristics.


  1. Acknowledgments: We thank P. Nightingale for his valuable input on statistical analysis. Funding: Supported by NIH grants AI33993 (to D.F.H.) and AI20963 and CA134060 (to V.H.E.), Kay Kendall Leukaemia Research grant KKL3227 and Leukaemia & Lymphoma Research grant 08038 (to M.C.), and NIH Cancer Center Support Grant P30 CA45579 (to the University of Virginia). Author contributions: M.C., V.H.E., and D.F.H. formulated the hypothesis and designed the experiments; M.C., H.D.L.P., A.N., J.M.P., J.Q., A.M.E., J.E.T., J.C., J.G.A., S.A.M., H.-W.H., S.P., and O.G. performed all the experiments; M.E.W., C.C., G.P., and S.D.F. provided patient samples and clinical input into the study; A.N., J.M.P., J.Q., A.M.E., J.C., J.G.A., and J.S. performed and analyzed mass spectrometry studies; M.C., H.D.L.P., S.P., J.E.T., H.-W.H., and O.G. performed and analyzed human in vitro experiments; and M.C., K.L.C., and A.L.Z. conducted MHC-peptide extraction. Competing interests: D.F.H., J.S., V.H.E., A.L.Z., K.L.C., A.N., J.C., and M.C. have equity interest in PhosImmune Inc. A patent application, entitled “Class I MHC phosphopeptides for cancer immunotherapy and diagnosis,” has been filed with PCT International Application No. PCT/US11/37699 based on U.S. Patent Application No. 61/347,559. The other authors declare no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article