Presentation of identical tumor antigen peptides by canine and human MHC class I

Summary

The increasing number of immunotherapies developed in the last two decades presents the need for an appropriate animal model to evaluate the efficacy of these treatments. The spontaneous nature of cancer in dogs and the common features they share with human malignancies make the dog a favorable translational model. The major histocompatibility complex (MHC) molecules in dogs are referred to as dog leukocyte antigens (DLA). Here, we introduce two antibodies for the characterization of the DLA class I immunopeptidome from primary canine tumors. We show that up to 55% of the peptides presented by tumor DLA are identical to peptides reported from common HLA class I molecules, displaying striking similarity in length and anchoring positions. Intriguingly, hundreds of these tumor DLA peptides are derived from well-established cancer-associated antigens. In summary, we demonstrate that canine and human MHC class I molecules are highly homologous in their antigen presentation function and peptide repertoire. These findings exhibit promising implications for advancing cancer immunotherapies and their translation from dogs to humans.

Graphical abstract

Highlights

• •Introduction of antibodies enabling DLA Immunopeptidomics workflow
• •Characterization of the DLA immunopeptidome of primary canine tumors
• •Uncovering DLA and HLA functional homology
• •Detection of identical tumor antigen peptides presented by DLA and HLA

Molecular biology; Immunology; Cancer

Introduction

The last two decades have witnessed the emergence of immunotherapy as an effective treatment for different types of cancer in humans. Great progress has been made in characterizing the human immune response and developing methods to identify immunogenic peptides that are presented to T cells in the context of class I HLA. Today, the immunotherapies that utilize HLA-presented peptides to target tumor cells, such as CAR T cells,¹ personalized cancer vaccines,² and bispecific T cell engagers³ have transformed the treatment landscape of cancers with demonstrated clinical efficacy. However, the main challenge that remains is how to appropriately test these new treatments in animal models before deploying them in the clinic.⁴

Pet dogs provide many advantages over traditional animal models that possess a very short lifespan or dissimilar immune responses.⁵ Dogs and humans present a high degree of homology in many areas of their genomes, including well-established oncogenes and tumor suppressor genes such as PIK3CA, KRAS, BRAF, and TP53 that are involved in tumorigenesis.^6,7 Like humans, dogs develop cancers spontaneously and these neoplastic transformations share many common features with different human cancers such as breast, melanoma, lymphoma and sarcoma.^8,9,10,11 In addition to similarities in the molecular and clinical landscape of cancer, dogs have an intact immune system and share the same environmental risks with humans, such as exposure to carcinogenic chemicals and biologicals. Together, these factors make dogs a promising model in comparative oncology to study naturally developing canine cancers and determine their translational relevance to human cancers.¹²

Despite a considerable increase in our knowledge of molecular characteristics of canine cancers, comparative oncology still faces many challenges within the field of immunotherapy including the annotation of the canine immune system.⁵ While cellular components of the canine immune system such as cytotoxic, helper and regulatory T cells as well as natural killer (NK) cells^{13,14,15,16,17} have been characterized extensively, in-depth analysis of the antigenic landscape of the canine tumors has been hampered due to lack of validated reagents and antibodies. For dogs to become a relevant model in the field of immuno-Oncology, in addition to the characterization of various immune cells, DLA molecules and their ability to present antigens to T cells must be studied so that a comprehensive understanding of the different aspects of the canine immune response is achieved.

DLA-88 is the most polymorphic of the four DLA class I loci (DLA-88, DLA-12, DLA-64, and DLA-79), encoding 139 of 173 DLA class I molecules that are so far identified.¹⁸ In addition, DLA-88 shows significantly higher gene expression level¹⁹ and has a demonstrated antigen presentation function.^20,21,22 The major obstacle in studying the peptide repertoire of the DLA molecules is identifying the antibody that recognizes these molecules specifically. So far, no DLA-specific antibody is available for the immunoprecipitation (IP) of DLA from canine cells. To address this issue, one group has engineered human cells to express DLA and has used anti-human B2m antibody for purification.²⁰ Others have transfected the FLAG-tagged DLA heavy chain gene into the canine cell lines and used Flag affinity purification to pull down the DLA molecules and isolate the peptides from the DLA-peptide complex.^21,22 While these methods have been beneficial for establishing the binding motifs of 3 DLA molecules (DLA-88∗501:01, DLA-88∗508:01, and DLA-88∗034:01) from engineered cells,^20,21,22 they are not applicable to direct DLA epitope discovery from primary canine cells and tumors.

In the present study, we have addressed this problem by introducing 2 new monoclonal antibodies into the DLA class I immunopeptidomics workflow. Using these antibodies, we successfully purified DLA molecules from human cells expressing monoallelic DLA as well as primary canine tumors. We resolved the binding motif of 3 new DLA-88 molecules and confirmed a previously characterized one, and we showed that DLA class I of the tumors present peptides that are identical to peptides reported from common HLA-A, -B, and -C molecules. Our method for the identification and characterization of DLA class I ligands enables future DLA immunopeptidomics studies directly from tumor-bearing dogs and provides a roadmap for the transition of immunotherapeutic agents from canine preclinical studies into human clinical trials.

Results

Preparation of DLA monoallelic cell lines

To determine the binding motifs of DLA Molecules, first, cell lines that express a single DLA class I molecule were established. For this purpose, the HLA class I heavy chains in the HCT116 cell line were knocked out using CRISPR-Cas9 as described in Methods. Lack of expression of HLA class I molecules on KO HCT116 cells was verified by staining with anti-human HLA class I antibody W6/32, Western blot (Figures 1A and 1B), and next-generation sequencing (NGS). Next, the KO HCT116 were transduced with heavy chains of DLA-88∗012:01, DLA-88∗003:02, and DLA-88∗501:01, which are among the common alleles in dogs.²³ The surface expression of DLA molecules was confirmed by flow cytometry using H58A and BB7.6 antibodies (Figures 1C–1E).

Figure 1.

Genetic ablation of HLA class I in HCT116 cells and generation of DLA Monoallelic cell lines

(A) Staining of wild-type (Orange) and HLA class I knockout (KO) HCT116 cells (Blue) with W6/32.

(B) Western blot of wild-type and HLA class I KO HCT116 cells. Rock1 was used as a loading control.

(C) DLA-88∗012:01 transduced HCT116 cells stained with H58A (Green) and BB7.6 (Red).

(D) DLA-88∗003:02 transduced HCT116 cells stained with H58A (Green) and BB7.6 (Red).

(E) DLA-88∗501:01 transduced HCT116 cells stained with H58A (Green) and BB7.6 (Red).

DLA-transduced monoallelic cells enable the identification of the DLA binding motifs

The stably transduced DLA monoallelic HCT116 cells were expanded, and IP was performed using BB7.6 coupled affinity column, followed by Nano LC MS/MS. A total of 591–1465 unique (8-14 aa) peptides were identified from each IP experiment (Data S1). To determine the binding motif for the eluted peptides, they were submitted to GibbsCluster2.0.²⁴ Default configuration and parameters were used to align and cluster the peptide sequences into different groups based on their similarities (Figure 2).

Figure 2.

Characterization of the DLA binding motifs and their length distribution from monoallelic HCT116 cells

(A) Binding motif of DLA-88∗003:02.

(B) Length distribution of peptides eluted from DLA-88∗003:02 transduced HCT116 cells.

(C) Binding motif of DLA-88∗012:01.

(D) Length distribution of peptides eluted from DLA-88∗012:01 transduced HCT116 cells.

(E) Binding motif of DLA-88∗501:01.

(F) Length distribution of peptides eluted from DLA-88∗501:01 transduced HCT116 cells.

In all cases, a cluster with the highest number of peptides and the highest Kullback-Leibler distance (KLD),²⁵ displaying a distinct binding motif, was observed as expected from a monoallelic cell line. Canonical peptide binding anchors were detected at positions P2 and P9 for all three DLA molecules. DLA-88∗003:02 exhibited a strong preference for acidic, negatively charged amino acids such as aspartic acid (D) and glutamic acid (E) at P2, and aliphatic residues such as isoleucine (I), valine (V), and leucine (L) at P9 (Figure 2A). In DLA-88∗012:01, P2 favored aliphatic and hydrophobic amino acids such as proline (P), alanine (A), isoleucine (I), and valine (V), and P9 accommodated the aromatic amino acid phenylalanine (F) in addition to aliphatic residues (L, V, and I). An auxiliary anchor position was also observed at P3, preferring positively charged, basic residues such as lysine (K) and arginine (R) (Figure 2C). Our binding motif for DLA-88∗501:01 from monoallelic HCT116 cells matched the previously published DLA-88∗501:01 motif obtained from engineered human cells (Figure 2E).²⁰ Like HLA class I molecules, the length preference (mode) for all 3 DLA molecules was 9 amino acids, which included the majority (>60%) of ligands (Figures 2B, 2D, and 2F). We have therefore established a system for assigning motif and ligand data to individual DLA class I molecules.

Canine tumors confirm and expand upon the binding motifs that were identified using monoallelic cells

After establishing the binding motif for DLA molecules using transduced human cells, we moved to purifying DLA directly from canine tumors. Information regarding tumor type, dog breed, DLA type, and the antibody used for affinity chromatography can be found in Table 1. (For additional information about canine tumors, please see Table S1).

Table 1.

Canine tumors information

Sample ID	Tumor Type	Dog Breed/Sex/Age	DLA typing	Antibody
Lola	osteosarcoma	Black Russian Terrier/Female/5y	DLA-88∗501:01	H58A BB7.6
163828A	anal sac carcinoma	English Cocker Spaniel/Female/NAa	DLA-88∗004:02	BB7.6
Lily	anal sac carcinoma	English Cocker Spaniel/Female/NAa	DLA-88∗004:02 DLA-88∗501:01 DLA-88∗501:02	H58A
Bogey	oral melanoma	American Cocker Spaniel/Male/10y	DLA-88∗003:02 DLA-88∗017:01 DLA-88∗006:01	BB7.6

^aNot available.

Osteosarcoma tumor (Lola)

H58A antibody is made in mice ascites and is available in unpurified form, requiring an additional purification step before coupling to affinity matrix for the IP of DLA molecules from canine cells and tumors. On the other hand, BB7.6 is commercially available in purified form, which can be readily used for coupling to affinity matrix and IP experiments. Before using BB7.6 and H58A to purify DLA from primary canine tumors, we first compared their immunoaffinity purification performance using an osteosarcoma tumor (Lola) expressing the well-characterized DLA molecule, DLA-88∗501:01 (Table 1).

The purification of DLA from the tumor using H58A and BB7.6 yielded 181 and 275 unique (8–14 aa) peptides, respectively (Data S2 and S3). After the submission of the peptides to GibbsCluster2.0, a cluster including majority of peptides (77.3% for H58A and 81.5% for BB7.6) (Table S2) was identified from both experiments showing a binding motif similar to what had been described previously from monoallelic cells by our team and others (Figures 3A and 3B)²⁰ with an overlap between the two (Figure 3C) indicating both antibodies bind the DLA molecules and can be utilized for the IP of DLA molecules from primary tissues.

Figure 3.

Comparison of H58A and BB7.6 antibodies' performance in the DLA immunopeptidomics workflow

(A) Peptide binding motif of DLA-88∗501:01 from peptides purified from osteosarcoma tumor (Lola) using BB7.6 antibody.

(B) Peptide binding motif of DLA-88∗501:01 from peptides purified from osteosarcoma tumor (Lola) using H58A antibody.

(C) Overlap between the peptides contributing to DLA-88∗501:01 clusters, eluted from Lola osteosarcoma tumor using BB7.6 and H58A antibodies.

Anal sac carcinoma tumor (163828A)

Next, we purified DLA from a canine anal sac carcinoma tumor that was homozygous for DLA-88∗004:02 according to RNA-seq results. BB7.6 affinity column was used to purify DLA from this tumor. A total of 1335 (8–14 aa) peptides were identified from this tumor (Data S4) using our LC-MS/MS approach as described in Methods (Figure 4A). The result of submission of the peptides to GibbsCluster2.0 was one cluster (1057 peptides, 79.2%) (Table S2), showing a clear motif from eluted peptides with 9mers being the preferred length (Figures 4B and 4C). Both P2 and P9 anchor positions showed preference for aliphatic residues such as leucine (L), isoleucine (I), and valine (V), with P9 accommodating phenylalanine (F) and methionine (M) as well. In addition to P2 and P9, an auxiliary anchor position at P3 was observed displaying a strong preference for negatively charged, acidic residues such as aspartic (D) and glutamic acid (E) (Figure 4B).

Figure 4.

DLA immunopeptidomics workflow and characterization of DLA binding motifs from mono- and multiallelic canine tumors

(A) Experimental workflow for purification of DLA molecules from canine tumors and characterization of their binding motifs.

(B) Binding motif of DLA-88∗004:02 from peptides purified from anal sac carcinoma (163828A) using BB7.6 antibody.

(C) Length distribution of peptides eluted from DLA-88∗004:02.

(D) Submission of the Lily anal sac carcinoma peptides to GibbsCluster2.0.

(E) Deconvolution and assignment of the peptide sequences to DLA alleles (DLA-88∗501:01/2 and DLA-88∗004:02).

(F) Length distribution of the peptides eluted from Lily anal sac carcinoma.

(G) Submission of the Bogey oral melanoma DLA peptides to GibbsCluster2.0.

(H) Length distribution of the peptides eluted from Bogey oral melanoma.

Anal sac carcinoma tumor (Lily)

Finally, we applied our method to purify DLA-presented peptides from multiallelic canine tumors. The immunopeptidome of multiallelic cells is a mixture of peptides presented by different alleles, and therefore, the knowledge of the DLA binding motifs is valuable for the correct assignment of the peptides to DLA molecules. For this reason, multiallelic tumors that expressed at least one previously characterized DLA allele were selected for peptide ligand characterization in this study.

The first multiallelic tumor was an anal sac carcinoma tumor expressing DLA-88∗501:01, DLA-88∗501:02, and DLA-88∗004:02 by RNA-seq. DLA purification from this tumor using H58A affinity column resulted in the identification of only 54 (8–14 aa) peptides (Data S5) due to the small size of the tumor (Table S1). Despite expressing 3 different DLA molecules, submission of the tumor peptides to GibbsCluster2.0 resulted in only one dominant cluster (Figure 4D), including 46 peptides (85.2%) (Table S2). Binding motifs of DLA-88∗501:01 and DLA-88∗004:02 have already been characterized in this study using monoallelic DLA-transduced human cells (Figure 2E) and homozygous canine tumor (Figure 4B). The motifs for these 2 alleles are very similar except for the residues at position 3 for which DLA-88∗004:02 has a strong preference for acidic, negatively charged amino acids such as aspartic (D) and glutamic acid (E) while DLA-88∗501:01 accommodates small residues such as alanine (A), asparagine (N) and serine (S) which can be explained by the presence of arginine (R156) in the pseudo sequence²⁶ that contributes to P3 binding pocket, which is absent from DLA-88∗501:01 (Figure S1A). On the other hand, DLA-88∗501:01 and DLA-88∗501:02 share the same pseudo sequence, meaning that the specific residues in the binding pockets that directly interact with peptide antigen are identical, and consequently, the binding motifs remain the same. Therefore, both DLA-88∗501:01 and DLA-88∗501:02 can bind and present similar peptides (Figure 4E).

Despite the great potential of unsupervised bioinformatic methods such as GibbsCluster for the deconvolution of the peptide specificities present in complex immunopeptidomes, they also have inherent limitations, such as underestimating the number of specificities when dealing with multiple MHC molecules with highly overlapping motifs and a limited peptide repertoire,²⁴ which were both challenges we faced in the case of this tumor. Here, due to the small number of peptides purified from this tumor and highly similar binding motifs of the DLA alleles expressed on the surface of tumor cells, the peptide sequences were inspected and manually assigned to the most likely DLA allele based on the binding motif of those 2 DLA molecules that had already been characterized (Figure 4E).

Oral melanoma (Bogey)

The second multiallelic tumor was an oral melanoma tumor expressing 3 DLA-88 molecules (DLA-88∗003:02, DLA-88∗006:01 and DLA-88∗017:01) with only one characterized DLA allele (DLA-88∗003:02) (Figure 2A). The BB7.6 affinity column was used to purify DLA from this tumor, yielding 7520 non-redundant (8–14 aa) peptides, including 2573 nonamers (Data S6). It is noteworthy that the DLA-88∗003:02 and DLA-88∗017:01 molecules are expressed together as a haplotype at the DLA-88 and DLA-88L Loci. It is one of the 8 major allelic haplotypes in DLA class I and is the most frequent compared to others. DLA-88L positive dogs are thought to have a higher gene expression level and a higher peptide presentation ability than the DLA-88L negative dogs.²⁷ This is a reason why the peptide yield for this tumor was higher than the other tumors in this study despite having a relatively similar size (Table S1). Due to the presence of 2 uncharacterized DLA molecules with unknown binding motifs, only 9 mers were used for motif deconvolution to reduce background noise. Using nonamers, which represent the majority of ligands, mitigates the interference of non-ligands and co-eluted contaminants.

The submission of the Bogey tumor DLA peptides to GibbsCluster2.0 resulted in 5 clusters (Figure S2A), 3 of which demonstrate distinct binding motifs, including one that was identical to DLA-88∗003:02 (Figure 4G). To assign the other 2 motifs to DLA-88∗006:01 or DLA-88∗017:01, the pseudo sequences of these DLA molecules were used as a guide. Both clusters show lysine (K) and arginine (R) at P3 (Figure 4G). However, while one prefers small aliphatic residues such as leucine (L), isoleucine (I), and valine (V) at P2, the other favors bulky aromatic residues, such as phenylalanine (F) and much less tyrosine (Y) and tryptophan (W). Aspartic (D156) and glutamic acid (E156) contribute to the P3 pocket of DLA-88∗006:01 and DLA-88∗017:01, respectively, which explains the preference of the P3 pocket for basic amino acids, such as K and R, in both DLA molecules (Figure S1B). However, the presence of the larger methionine (M45) at P2 of DLA-88∗017:01 makes this pocket small so that it favors small aliphatic residues, while the small threonine (T45) in DLA-88∗006:01 creates a larger P2 pocket that can accommodate the bulky aromatic residues (Figures 4G and S1B). Even though applying a combination of a fully unsupervised approach such as GibbsCluster and the DLA pseudo sequence represents a practical approach to assign the binding motifs to the corresponding DLA molecules, data from monoallelic cells expressing DLA-88∗006:01 and DLA-88∗017:01 must be generated to confirm these DLA motif assignments.

DLA and HLA class I are functional homologs

Similarity between the binding motif of DLA-88∗501:01 and HLA-A∗02:01 has been previously reported.²⁰ Here, we sought to determine whether there was similarity between other DLA molecules characterized in this study and HLA molecules and identify the human homologs for these DLA molecules. For this purpose, we first searched for the HLA molecules with similar peptide binding motifs to our DLA in http://mhcmotifatlas.org/.²⁸ For each DLA molecule, we identified multiple HLA molecules encoded at the HLA-A, -B, and -C loci that display highly similar binding motifs (Figure 5). For example, we found HLA-B∗40:02 and HLA-B∗40:01 to show similar motifs to DLA-88∗003:02 (Figure 5A). HLA-B∗07:04, HLA-B∗07:02 and HLA-B∗42:01 showed motifs analogous to DLA-88∗012:01 (Figure 5B) and HLA-A∗02:07, HLA-C∗05:01 and HLA-C∗08:02 displayed very similar motifs to DLA-88∗004:02 (Figure 5D). For DLA-88∗501:01, in addition to previously reported HLA-A∗02:01, we identified other HLA molecules, including HLA-A∗02:02 and HLA-C∗03:04, to present a comparable binding motif (Figure 5C).

Figure 5.

Identification of DLA homologs in human

(A) DLA-88∗003:02 demonstrates homology in its binding motif with the members of the B44 supertype family.

(B) DLA-88∗012:01 binding motif displays homology to the members of the B07 supertype family. Both DLA-88∗501:01

(C) and DLA-88∗004:02 (D), show similarity in their binding motifs to the members of A02 supertype as well as certain HLA-C molecules.

Based on the similarity between the peptide binding motifs of DLA and HLA molecules, we hypothesized that peptides presented by DLA molecules could bind their human homologs. To test this hypothesis, peptides eluted from each of the monoallelic cell lines and the homozygous canine tumor expressing DLA-88∗003:02, DLA-88∗012:01, DLA-88∗501:01, and DLA-88∗004:02 were submitted to NetMHCpan4.1,²⁹ and their predicted bindings to the corresponding HLA molecules were determined. For this analysis, the rank thresholds of 0.5 for strong binders and 2.0 for weak binders were applied on peptides of 9aa length that include the majority of ligands. Up to 86.2% (DLA-88∗004:02 and HLA-C∗05:01) (Table S8) of the peptides from different DLA molecules were predicted to bind their human homologs, confirming the functional similarity between DLA and HLA molecules (Tables S3–S8).

The observation that the peptide binding motif of each DLA molecule displays homology to more than one HLA molecule suggests that the DLA functional similarity goes beyond only one HLA molecule. To examine this, the predicted binding of the peptides from each DLA molecule to different members of the A02, B07, and B44 supertypes to which HLA-A∗02:02, A∗02:07, B∗07:02, B∗07:04, B∗42:01, and B∗40:02 belong was determined. Supertypes are groups of HLA molecules that bind and present largely overlapping peptides, meaning they are functionally related.^30,31 In each supertype, alleles with high frequency in the world population (>0.01%) were selected.³² In almost all cases, the majority of the peptide repertoire of the DLA molecules was predicted to bind to multiple members of each HLA supertype, further supporting the functional similarity between DLA and HLA-A and -B molecules (Tables S3–S6). In addition to members of HLA-A and B supertypes, several HLA-C molecules, including HLA-C∗03:04, HLA-C∗02:02, HLA-C∗05:01, and HLA-C∗08:02 that had displayed similar binding motifs were included in this analysis (Tables S7 and S8).

Our results reveal that the peptides presented by DLA-88∗004:02 and DLA-88∗501:01 have a high predicted binding affinity to multiple members of HLA-A02 supertype as well as several HLA-C molecules (Tables S3, S4, S7, and S8). On the other hand, DLA-88∗003:02 and DLA-88∗012:01 display a preference for binding to the members of HLA-B07 and B44 supertype families (Tables S5 and S6). It is worth mentioning that DLA-88∗004:02 peptide repertoire showed an equally high predicted binding affinity for HLA-A and -C molecules, while DLA-88∗501:01 presented a higher affinity to HLA-A molecules compared to HLA-C (Tables S3, S4, S7, and S8).

Identical peptides are presented by the DLA of canine tumors and HLA

The high predicted binding affinity of DLA ligands to HLA-A, -B, and -C molecules suggests that dog and human leukocyte antigens (DLA and HLA) might present the same peptides. To test this, we searched the peptides identified from the DLA of the canine tumors against the peptides identified from HLA class I molecules reported in IEDB. The length criteria of 8–14 aa was applied to both DLA and HLA presented peptides. To eliminate any bias, peptides purified from the DLA-transduced human cell lines were not included in this analysis, and only DLA ligands from dog tumors were analyzed. Interestingly, we found that 30 to 55% of the peptides identified from canine tumors were identical to HLA class I peptides reported in IEDB. The percentage of identical peptides for Lola Osteosarcoma was 54.7% (99/181) when H58A was used for immunoaffinity purification and 53.1% (146/275) when BB7.6 was used. For the Lily and 163828A anal sac carcinomas, the percentage of identical peptides was 44.4% (24/54) and 42.5% (568/1335), respectively. Finally, for the Bogey oral melanoma tumor, 30.3% (2280/7520) of DLA ligands were identical to the peptides reported from HLA class I molecules (Data S2–S6).

Moreover, the HLA molecules that presented peptides identical to a DLA molecule were the same HLA molecules we identified as DLA homologs based on binding motif similarity. It is noteworthy that the Osteosarcoma and anal sac carcinoma tumors that expressed DLA-88∗004:02 and DLA-88∗501:01, presented a higher number of HLA-like peptides (42.5%–54.7%) compared to oral melanoma tumor (30.3%) that expressed DLA-88∗003:02. This is consistent with our finding that higher proportion of the DLA-88∗004:02 and DLA-88∗501:01 ligands bind to HLA-A and -C alleles in comparison with the lower proportion of the DLA-88∗003:02 peptides that bind mainly HLA-B molecules (Tables S3, S4, S6, S7, and S8).

DLA and HLA class I present identical peptides from shared tumor antigens

Finally, when we searched the DLA/HLA shared peptides against a list of known tumor antigens compiled from different databases^33,34 we found that DLA molecules extracted from canine tumors presented 273 peptides from 179 cancer-associated antigens. Tumor antigens sampled by the DLA of tumors include well-established oncogenes and tumor suppressor genes such as Janus kinase 1 (JAK1),³⁵ Myc proto-oncogene protein (MYC),³⁶ signal transducer and activator of transcription (STAT),³⁷ DNA methyltransferase 1 (DNMT1),³⁸ DNA topoisomerase 2-alpha (TOP2A),³⁹ BRCA1 associated protein-1 (BAP1),⁴⁰ nucleophosmin 1 (NPM1),^41,42 speckle-type POZ protein (SPOP),⁴³ stromal antigen 1 and 2 (STAG 1 and 2),⁴⁴ B-cell lymphoma 6 protein (BCL6),⁴⁵ catenin beta-1 (CTNNB1),⁴⁶ DEAD-box RNA helicase DDX5 (p68) (DDX5),⁴⁷ mitogen-activated protein kinase kinase kinase kinase 4 (MAP4K4),⁴⁸ Forkhead box protein P1 (FOXP1),⁴⁹ AT-rich interactive domain-containing protein 1A (ARID1A),⁵⁰ melanoma-associated antigen D1 (MAGED1),⁵¹ and transforming protein RhoA (RHOA),⁵² to name a few (Data S2–S6). These data demonstrate that DLA molecules can present peptides identical to HLA ligands from shared tumor-associated antigens. However, it must be emphasized that sequence identity between DLA and HLA presented peptides does not imply shared immunogenicity or T cell recognition, which requires additional testing.

Discussion

Recent advances in cancer immunotherapy and the efficacy they have offered in the treatment and management of different malignancies has transformed the cancer therapy landscape. However, one challenge that has emerged with the increasing number of immunotherapeutic agents is the high rate of failure in early-phase human studies despite showing efficacy in preclinical testing. This is in part due to the inability of animal models to accurately reflect the human immune system, tumor microenvironment, and the interaction between the two.⁴ The spontaneous nature of the canine cancers, which provide advantages over induced cancer models and the strong similarity with human cancers, suggests including dogs in translational cancer research.^12,53,54 In order to understand the interplay between the tumor and canine immune system, uncovering the tumor antigenic repertoire in the context of DLA and characterization of the immune cells are both equally important. For over a decade, the absence of DLA-specific monoclonal antibodies has limited the study of antigen presentation by DLA molecules from biological samples. In the present study, we addressed this shortcoming by introducing two monoclonal antibodies, H58A and BB7.6, for canine immunopeptidomics applications. Using these antibodies, we first purified DLA molecules from human cells expressing monoallelic DLA on their surface. We next moved on to purifying DLA directly from canine tumors expressing the same DLA molecules. The knowledge of the binding motif of these individual DLA molecules enabled us to then deconvolute the mixture of peptides purified from multiallelic tumors. The length preference and anchor positions, aka binding motif, of the peptides identified by these antibodies from engineered human cells, mono- and multiallelic tumors, were identical with 9mers being the dominant length, demonstrating that our antibodies can reliably and reproducibly purify the DLA molecules from canine tumors.

Using these validated monoclonal antibodies, we defined the peptide binding specificities of 3 new DLA molecules (DLA-88∗003:02, DLA-88∗012:01, and DLA-88∗004:02), confirmed a previously reported one (DLA-88∗501:01) and tentatively assigned the peptide binding motifs for two new molecules (DLA-88∗006:01, DLA-88∗017:01). These molecules displayed striking functional similarity to HLA class I molecules both at individual allele and supertype level by presenting peptides with the same length preference and binding motif that can bind to their human equivalents with high affinity. These findings were further supported by the fact that 30–55% of the peptides eluted from the DLA of the canine tumors were identical to previously reported HLA-presented peptides in IEDB, including peptides from cancer-associated antigens. This confirms that the DLA molecules characterized in this study are truly homologous to HLA class I molecules and can present the same peptides.

To understand how this level of DLA/HLA similarity compares to other animal models, we searched the published MHC class I immunopeptidome of mouse, the most widely used animal model in biomedical research, acquired from 19 normal tissues and 4 tumor cell lines⁵⁵ against human HLA presented peptides reported in IEDB using the same approach. From the total H2Db and H2Kb presented peptides (8–14 aa), 18% of the peptides were identical to humans. When the murine peptides were limited to only include unique, high confidence H2Db and H2Kb ligands, the percentage of identical peptides dropped below 10% (9.76% and 9.4% for H2Kb and H2Db, respectively). This comparison reveals the high level of similarity between canine and human leukocyte antigen presentation function compared to other animal models, which is in part due to similarities between the dog and human genome and proteome.

Furthermore, these DLA molecules that are highly frequent in the canine population^19,23 display functional similarity with the members of the common HLA supertypes, such as HLA-A02 and HLA-B07, in the human population. This suggests that any novel immunotherapeutic target that is presented by these DLA molecules has the advantage of binding to other members of the same supertype and covering a large population in both dogs and humans. This feature facilitates conducting appropriately sized preclinical testing for efficacy and toxicity, and a safe transitioning to human clinical trials, which are the challenges currently facing the immunotherapy field.⁴

Our study lays the groundwork for future large-scale immunopeptidomics studies in tumor-bearing dogs in which immunogenic antigens and neoantigens can be discovered from the canine-human shared tumor antigens and be tested in canine preclinical studies. The outcome could inform the development of new cancer treatments that can be safely used in a clinical setting for both dogs and humans. However, there are still challenges ahead that must be addressed. The DLA molecules characterized here are encoded by the DLA-88 locus, which is known to have high expression on the cell surface and display antigen presentation capacity such as HLA class I molecules.^20,21,22,27 Next step would be the characterization of additional DLA-88 molecules as well as non-classical DLA class I molecules, including DLA-12, DLA-64, and DLA-79⁵⁶ and verification of their antigen presentation capacity. As more DLA immunopeptidomics data become available using the workflow described here, training and development of algorithms that accurately predict and assign ligands to their cognate DLA will be expedited, which will significantly improve the identification of actionable cancer-associated antigens and neoantigens in canine tumor samples.

In summary, the strong similarity between antigen presentation of DLA of canine tumors and HLA reported here offers a great opportunity to include dogs in comparative and translational cancer research. This will further support the transition of immunotherapies that target tumor antigenic peptides presented by both DLA and HLA from preclinical testing to clinical settings, ultimately contributing to the advancement of cancer immunotherapy in both dogs and humans.

Limitations of the study

In the present study, only DLA-88 molecules have been studied, and non-classical DLA class I molecules such as DLA-12, DLA-64, and DLA-79, and their antigen presentation capacity were not investigated. The binding motif of the DLA molecules (DLA-88∗006:01 and DLA-88∗017:01) from the multiallelic tumor (Bogey) remains tentative and must be validated by data generated from a monoallelic cell line or tumor.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, William Hildebrand (william-hildebrand@ou.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data: The raw mass spectrometry data for DLA-transduced monoallelic cells, as well as all tumors processed in this study, have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD074485. All canine tumor RNA-seq data are publicly available, as listed in the key resources table.
• •Code: This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We sincerely thank the pet owners who entered their dogs in this study. This study was funded by a grant from National Cancer Institute (NCI) R01CA252713 (SZ).

Author contributions

Conceptualization: S.K. and W.H.; methodology: S.K., H.Y., S.C., W.B., M.Z., and H.L.C..; investigation: S.K., H.Y., S.C., W.B., M.Z., and H.L.C.; visualization: S.K., H.Y., and S.C.; funding acquisition: S.Z. and W.H.; project administration: W.H.; supervision: W.H.; writing – original draft: S.K. and W.H.; writing – review and editing: S.K., W.H., and S.Z.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-MHC Class I (HLA-A, HLA-B, HLA-C) – human, Clone W6/32	BioXCell	Cat# BE0079; RRID: AB_1107730
Mouse anti-MHC class I monoclonal antibody, Clone DG-H58A	Monoclonal Antibody Center, Washington State University	Cat# DG-BOV2001; RRID: AB_3718131
Anti-Human HLA A2, B7 (MHC Class I), Clone BB7.6	Leinco Technologies	RRID: AB_2893725
Bacterial and virus strains
OneShot Top10 Chemically Competent	ThermoFisher	Cat# C404003
Biological samples
Canine tumor samples	Collected from veterinary hospitals with owner informed consent.	N/A
Chemicals, peptides, and recombinant proteins
RPMI 1640	Gibco	Cat# 11875119
DMEM/F-12	Gibco	Cat# 11320082
PBS	Gibco	Cat# 20012027
FBS	Gibco	Cat# A5256701
CNBr-activated Sepharose 4 Fast Flow	Amersham Pharmacia Biotech/Cytiva	Cat# 17-0981-01
Tris pH 8.0	NELS	Cat# E199-500 ML
IGEPAL® CA-630	Sigma-Aldrich	Cat# 18896
Complete protease inhibitor cocktail	Roche	Cat# 11836145001
Glacial Acetic acid	Sigma-Aldrich	Cat# AX0077
iRT peptides	Biognosys	Cat# 1900615
Transporter 5	Polysciences	Cat# 26008
Xho1	NEB	Cat# R0146S
Salt-T4 DNA Ligase	NEB	Cat# M0467S
EcoR1	NEB	Cat# R3101S
LR Clonase II	ThermoFisher	Cat# 11791020
Polybrene	Sigma-Aldrich	Cat# TR-1003-50UL
Blasticidin	ThermoFisher	Cat# A1113903
Critical commercial assays
Melon™ Gel IgG Spin Purification Kit	Thermo Scientific	Cat# 45206
Zeba™ Spin Desalting Columns	Thermo Scientific	Cat# 89893
QIAfilter Plasmid Kits	Qiagen	Cat #12243
Monarch Plasmid Spin Miniprep Kit	NEB	Cat #T1110L
Monarch Spin DNA Gel Extraction Kit	NEB	Cat #T1120L
QIAamp DNA Blood Mini Kit	Qiagen	Cat #51106
Rapid Barcoding Kit 96 V14	Oxford Nanopore Technologies, UK	Cat #SQK-RBK114.96
Deposited data
Mass spectrometry data for DLA-transduced mono-allelic cells and all canine tumors	This paper	PXD074485
Canine tumors RNA-seq data	The SRA database	PRJNA1428288
Experimental models: Cell lines
HCT116 (Human Colorectal Carcinoma)	ATCC	RRID: CVCL_0291
MDCK-II (Canine Normal Kidney)	ATCC	RRID: CVCL_0424
Oligonucleotides
sgRNA HLA Class I: GATGTAATCCTTGCCGTCGT	IDT	N/A
Recombinant DNA
pMD2.G	Addgene	Cat #122259
psPAX2	Addgene	Cat #12260
pENTR (ref B)	Addgene	Cat #17396
pLenti PGK Blast DEST(ref B)	Addgene	Cat #19065
pLenti eCas9 (ref A)	Addgene	Cat #140237
Software and algorithms
GibbsCluster 2.0	Andreatta et al. 201724	https://services.healthtech.dtu.dk/services/GibbsCluster-2.0/
NetMHCpan4.1	Reynisson et al. 202029	https://services.healthtech.dtu.dk/services/NetMHCpan-4.1/
PEAKS Studio 11.5 software	Bioinformatics Solutions, Waterloo, Canada	https://www.bioinfor.com/peaks-studio/
MinKNOW 24.02.8	Oxford Nanopore Technologies, UK	https://nanoporetech.com/
NanoTYPER™ 2.2	Omixon, Hungary	https://www.omixon.com/
Other
R10.4.1 flow cell	Oxford Nanopore Technologies, UK	Cat #FLO-MIN114
NanoTYPE™	Omixon, Hungary	Cat #NT2411v2

Experimental model and study participant details

Canine tumor samples

Information regarding canine tumor samples including tumor type, dog breed, sex and age have been listed in Table 1. Briefly, 4 canine tumors were used in this study. First tumor (Lola) was an Osteosarcoma tumor from a 5-year-old female Black Russian Terrier. Second tumor (163828A) was an Anal Sac Carcinoma from a female English Cocker Spaniel dog. Third tumor (Lily) was another Anal Sac Carcinoma from a female English Cocker Spaniel dog. The exact age for these two dogs is unknown. Forth tumor (Bogey) was an Oral melanoma from a 10-year-old male American Cocker Spaniel. Dog sex did not influence the results of this study.

Ethical statement and sample collection

Flash-frozen spontaneous tumor samples were collected from client-owned dogs that develop the disease naturally, under the guidelines of the Institutional Animal Care and Use Committee for use of residual diagnostic specimens and with owner-informed consent. The research received the ethical approval from the Institutional Animal Care and Use Committee of the University of Georgia (A2023 08-029-Y1-A0, approved on 12-04-2023).

Cell lines

The human colorectal carcinoma cell line HCT116 and MDCKII (an epithelial-like cell line isolated from normal canine kidney tissue) were obtained from American Type Culture Collection (ATCC). The cells were grown in complete RPMI and DMEM media (Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco/Invitrogen Corp). HCT116 cells were used for transduction with the canine MHC class I alleles DLA-88∗003:02, DLA-88∗012:01 and DLA-88∗501:01. Cells were expanded up to 5e8 or 1e9 and trypsinized. The cells were next washed with PBS twice and spun down at high speed for 10 min at 4°C. The cell pellets were immediately frozen in LN₂ and stored at −80°C until downstream processing. The HCT116 cell line (A∗01:01, A∗02:01, B∗45:01, B∗18:01, C∗07:01, C∗05:01) was subjected to NGS-based HLA typing prior to large-scale culture and data collection for authentication. Cells were also tested for mycoplasma contamination and found to be negative. RNA-seq was used for DLA typing of canine tumors.

Antibodies

W6/32 antibody

W6/32 monoclonal antibody that reacts with the human major histocompatibility complex (MHC) class I (BioXCell) was used for monitoring the expression of the MHC class I in wild type HCT116 cells or lack thereof after knocking out MHC class I.

H58A antibody

H58A, a mouse anti-MHC class I monoclonal antibody that recognizes a highly conserved epitope on MHC class I molecules in various species⁵⁷ including dogs, (Monoclonal Antibody Center, Washington State University) was used for flow cytometry and detection of surface expression of DLA molecules on canine cells and DLA-transduced human cells. This monoclonal antibody is made in ascites from mice and therefore is unpurified. The H58A antibody was also used for immunoprecipitation of DLA molecules from canine cells and tumors after an additional purification step using Melon™ Gel IgG Spin Purification Kit (Thermo Scientific) according to manufacturer’s instructions. Melon™ Gel Monoclonal IgG Purification Kit is a high-yield, mild purification system for monoclonal antibodies from ascites or hybridoma cell culture supernatant, or for large-scale antibody purification from serum.

BB7.6 antibody

Bw4 and Bw6 epitopes are defined by amino acid residues (77, 80–83) on the α-1 α-helix of all HLA-B and certain HLA-A and HLA-C molecules and are recognized by a wide range of antibodies. BB7.6 is an anti-Bw6 monoclonal antibody that recognizes Bw6 epitope (residues 80–83 NLRG sequence) on HLA-B molecules from which R82 and G83 are essential for its binding.⁵⁸ All DLA-88 alleles, except for DLA-88∗020:01 (residue 80–83 TLHG sequence), have the Bw6 epitope (LRG) and therefore could be recognized by BB7.6 (Figure S3A). Detection of DLA molecules by anti-Bw6 monoclonal antibody was confirmed in DLA-transduced HCT116 cells (Figure 1) and canine cell line MDCKII using BB7.6 (Leinco Technologies) and flow cytometry (Figure S3C). This antibody was also used for immunoprecipitation of DLA molecules from DLA-transduced cells and canine tumors.

Method details

Transduction of human cells and preparation of DLA monoallelic cells

Genetic ablation of HLA class I using CRISPR/Cas9

Single guide RNA (sgRNA) was designed using www.benchling.com targeting nucleotides 435–444 located in exon 3, which codes for the heavy chain, to disrupt the expression of endogenous HLA class I. sgRNA sequence 5′ GATGTAATCCTTGCCGTCGT 3′ was inserted into Lenti-eCas9 (Addgene#140237) using molecular biology methods⁵⁹ and generation of recombinant lentivirus has been previously described.⁶⁰ Loss of surface expression of endogenous HLA class I molecules in HCT116 cells was verified using flow cytometry and staining with the HLA class I pan-antibody W6/32 on a BD FACSCalibur (Becton Dickinson) and Western blotting (Figures 1A and 1B). Additionally, genetic alteration of HLA class I was confirmed by gDNA extraction (QIAamp DNA Blood Mini Kit Part# 51106) and PCR amplification (NanoTYPE™ assay kit, NT2411v2 Omixon), followed by next-generation sequencing (NGS). NGS was performed using Oxford Nanopore Technologies (ONT) sequencing platform consisting of the R10.4.1 flow cell (Part #FLO-MIN114), a MinION sequencer, and MinKNOW sequencing software. Library preparation of PCR amplified products was according to manufacturer protocol (Rapid Barcoding Kit 96 V14 (ONT) Part #SQK-RBK114.96). Sequencing data analysis used NanoTYPER™ software (Omixon, Hungary).

Generation of viral vectors to express DLA-88 alleles

DLA-88∗003:02, DLA-88∗012:01 and DLA-88∗501:01 (AA21 P > L) were selected based on full coding sequence availability and allele frequency, by Miyamae et al.^19,27 Codon-optimized sequences including EcoR1 and Xho1 restriction enzymes sites were synthesized as gBlocks™ by Integrated DNA Technologies (IDT). gBlocks were cut with EcoR1 and Xho1 and ligated into pENTR plasmid (Addgene #17396) cut with the same restriction enzymes. Following confirmation by Sanger Sequencing, pENTR plasmids with DLA sequences were recombined into pLenti PGK Blast DEST using Gateway cloning technology according to previously published methods.⁶¹ Cells were selected using 10 μg/ml of Blasticidin (ThermoFisher) and then the cell surface expression of DLA molecules were verified using flow cytometry using H58A and BB7.6 antibodies. To establish a clonal population of cells expressing monoallelic DLA molecules, cells went through a final round of bulk sorting after staining with H58A or BB7.6 (BD FACSCalibur, Becton Dickinson) (Figures 1C–1E).

Isolation and purification of DLA bound peptides

DLA molecules were purified from the transduced human cells (HCT116) and canine tumors by immunoaffinity chromatography using purified mouse anti-MHC class I monoclonal Ab (clone H58A) (Monoclonal Antibody Center − Washington State University) or BB7.6 (Leinco Technologies). Immunoaffinity columns were generated by coupling 2 mg of the purified antibody to 1 mL of matrix (CNBr-activated Sepharose 4 Fast Flow, Amersham Pharmacia Biotech, Orsay, France). Frozen cell pellets containing 5e8 to 1e9 cells were pulverized using Retsch Mixer Mill MM400, resuspended in lysis buffer comprised of Tris pH 8.0 (50 mM), Igepal, 0.5%, NaCl (150 mM) and complete protease inhibitor cocktail (Roche, Mannheim, Germany) and incubated at 4°C for 1 h on a rotary shaker. Lysates were centrifuged in an Optima XPN-80 ultracentrifuge (Beckman Coulter, IN, USA) at 4°C for 90 min (200,000 xg). Cleared supernatants were filtered using a 0.45 μm filter and loaded on immunoaffinity columns overnight at 4°C. Columns were washed sequentially with 10 cv of wash buffers at pH:8.0 and were eluted with 0.2 N acetic acid. The DLA was denatured, and the peptides were isolated by adding glacial acetic acid (up to 10%) and heat (76°C for 10 min). The mixture of peptides and DLA molecules was subjected to reverse phase high-performance liquid chromatography (RP-HPLC).⁶²

Fractionation of the DLA/peptide mixture by RP-HPLC

Reverse-phase high-performance liquid chromatography (RP-HPLC) was employed to simplify the peptide mixture eluted from the affinity column and isolate the peptides from DLA protein. Initially, the eluate was dried under vacuum using a CentriVap concentrator (Labconco, Kansas City, Missouri, USA). The resulting solid residue was dissolved in 10% acetic acid and fractionated over a 150-mm long Gemini C18 column (110 Å pore size, 5 μm particle size; Phenomenex, Torrance, California, USA) using a Shimadzu Nexera instrument (Shimadzu Scientific Instruments, Pittsburgh, Pennsylvania, USA). An acetonitrile (ACN) gradient was run at pH 2 with a two-solvent system: Solvent A (2% ACN in water) and Solvent B (5% water in ACN), both containing 0.1% trifluoroacetic acid (TFA). The column was pre-equilibrated with 2% Solvent B. The sample was then loaded at a flow rate of 120 μL/min, followed by a two-segment gradient at 160 μL/min. Fractions were collected every 2 min using a Gilson FC 203B fraction collector (Gilson, Middleton, Wisconsin, USA), and the ultraviolet (UV) absorption profile of the eluate was recorded at a wavelength of 215 nm.⁶³

Nano LC-MS/MS

Peptide-containing HPLC fractions were dried and resuspended in a solvent composed of 10% acetic acid, 2% acetonitrile (ACN), and iRT peptides (Biognosys, Schlieren, Switzerland) as internal standards. The fractions were individually injected into either a Sciex 5600 TripleTOF mass spectrometer coupled with an Eksigent nanoLC 415 nanoscale RP-HPLC (AB Sciex, Framingham, Massachusetts, USA) or a Sciex 7600 ZenoTOF mass spectrometer coupled with a Waters Acquity M Class UPLC (Waters, Milford, Massachusetts, USA). Different liquid chromatography (LC) gradients and information-dependent analysis (IDA) acquisition methods were used for each instrument. To fragment precursor ions, collision-induced dissociation (CID) with dynamic collision energy (CE) to adjust the CE based on the mass and charge of the ions were used in both instruments.

For the Sciex TripleTOF 5600 system, the fractions were individually submitted to an Eksigent nanoLC 415 nanoscale RP-HPLC (AB Sciex, Framingham, Massachusetts, USA) equipped with a 5-mm long, 350 μm internal diameter ChromXP C18 trap column (3 μm particles, 120 Å pores) and a 15-cm long ChromXP C18 separation column (75 μm internal diameter) packed with the same medium (AB Sciex, Framingham, Massachusetts, USA). An acetonitrile (ACN) gradient was run at pH 2.5 using a two-solvent system: Solvent A (0.1% formic acid in water) and Solvent B (0.1% formic acid in 95% ACN in water). The column was pre-equilibrated with 2% Solvent B. Samples were loaded at a flow rate of 5 μL/min onto the trap column and run through the separation column at 300 nL/min with two linear gradients: 10%–40% Solvent B over 70 min, followed by 40%–80% Solvent B over 7 min. The column effluent was ionized using the Digital PicoView ion source (New Objective, Littleton, MA, USA) of an AB Sciex TripleTOF 5600 mass spectrometer (AB Sciex, Framingham, MA, USA) with the source voltage set to 2,400 V. For the information-dependent acquisition (IDA), a survey scan of positive ions over the range of 300 to 1,250 m/z for 0.25 s was performed. Following each survey scan, up to 22 ions with a charge state of 2–5 and an intensity of at least 200 counts per second (cps) were subjected to collision-induced dissociation (CID) for tandem MS (MS/MS) analysis over a maximum period of 2.5 s. Selection of a particular ion m/z was excluded for 30 s after three initial MS/MS experiments. PeakView Software version 1.2.0.3 (AB Sciex, Framingham, MA, USA) was used for data visualization.⁶³

The Sciex ZenoTOF 7600 system was equipped with an Acquity M Class Waters UPLC (Waters, Milford, Massachusetts, USA), including a 20-mm long, 180 μm internal diameter NanoEase M/Z Symmetry C18 trap column (5 μm particles, 100 Å pores) and a 15-cm long NanoEase M/Z HSS C18 T3 column (75 μm internal diameter, 1.8 μm particles, 100 Å pores; Waters, Milford, Massachusetts, USA). The column was pre-equilibrated with 2% Solvent B. Individual fractions were loaded for 4 min at the flow rate of 5 μL/min onto the trap column and run through the UPLC column at 500 nL/min using the following gradient program: 1% Solvent B for 2 min, 2%–40% Solvent B over 40 min, 40%–80% Solvent B over 7 min, maintaining 80% Solvent B for 2 min, reducing to 1% Solvent B over 1 min, followed by re-equilibration at 1% Solvent B for 27 min. The column effluent was ionized using the Optiflow 1–50 μL ion source at the source voltage set to 4,350 V. IDA survey scans were performed over a range of 350 to 1,250 m/z for 0.25 s, and up to 50 ions with a charge state of 2–4 and with the minimum intensity of 100 cps were subjected to fragmentation over 0.05 s accumulation time. Ions were excluded from fragmentation for 12 s after one occurrence. Sciex OS version 3.4.0 (AB Sciex, Framingham, MA, USA) was used for data visualization.

Peptide data analysis

Peptide sequences were identified using PEAKS Studio 11.5 software (Bioinformatics Solutions, Waterloo, Canada). For the data acquired on the Sciex 5600 instrument, the precursor mass error tolerance was set to 30 ppm, and the fragment ion mass tolerance was set to 0.02 Da. For the data acquired on the Sciex 7600 instrument, the precursor mass tolerance was set to 10 ppm.

The identified sequences were searched against the iRT peptides and the UniProt Canis lupus familiaris protein database (taxon identifier 9615; https://www.uniprot.org/) as well as a merged database comprised of the following individual databases: CanFam3 (http://ftp.ensembl.org/pub/release-104/fasta/canis_lupus_familiaris/pep/), CanFam4 (https://genome.ucsc.edu/cgi-bin/hgGateway?db=canFam4), Basenji Ensembl Release 104 (http://ftp.ensembl.org/pub/release-104/fasta/canis_lupus_familiarisbasenji/pep/), and Great Dane Ensembl Release 104 (http://ftp.ensembl.org/pub/release-104/fasta/canis_lupus_familiarisgreatdane/pep/). For DLA-transduced human cells, a database composed of SwissProt Homo sapiens (taxon identifier 9606; https://www.uniprot.org/) and iRT peptides was used as the reference for database search. Variable post-translational modifications (PTM) including acetylation, deamidation, pyroglutamate formation, oxidation, sodium adducts, phosphorylation, and cysteinylation were included in database search. Identified peptides were further filtered at a false discovery rate (FDR) of 1% using PEAKS decoy-fusion algorithm.

Supplemental information

Data S2. List of all 8–30 aa non-redundant peptides identified at 1% FDR, following H58A immunoaffinity purification of the DLA from Lola Osteosarcoma tumor

List of canine tumor peptides identical to human and their tumor-associated antigen (TAA) source protein.

Data S3. List of all 8–30 aa non-redundant peptides identified at 1% FDR, following BB7.6 immunoaffinity purification of the DLA from Lola Osteosarcoma tumor

List of canine tumor peptides identical to human and their tumor-associated antigen (TAA) source protein.

Data S4. List of all 8–30 aa non-redundant identified at 1% FDR, following BB7.6 immunoaffinity purification of the DLA from anal sac carcinoma tumor (163828A)

List of canine tumor peptides identical to human and their tumor-associated antigen (TAA) source protein.

Data S5. List of all 8–30 aa non-redundant peptides identified at 1% FDR, following H58A immunoaffinity purification of the DLA from Lily anal sac carcinoma tumor

List of canine tumor peptides identical to human and their tumor-associated antigen (TAA) source protein.

Data S6. List of all 8–30 aa non-redundant peptides identified at 1% FDR, following BB7.6 immunoaffinity purification of the DLA from Bogey oral melanoma tumor

List of canine tumor peptides identical to human and their tumor-associated antigen (TAA) source protein.