Autoantibody repertoire profiling in tissue and blood identifies colorectal cancer-specific biomarkers

Abstract

Autoantibodies (AAbs) in the blood of colorectal cancer (CRC) patients have been evaluated for tumor detection. However, it remains uncertain whether these AAbs are specific to tumor-associated antigens. In this study, we explored the IgG and IgM autoantibody repertoires in both the in situ tissue microenvironment and peripheral blood as potential tumor-specific biomarkers. We applied high-density protein arrays to profile AAbs in the tumor-infiltrating lymphocyte supernatants and corresponding serum from four patients with CRC, as well as in the serum of three noncancer controls. Our findings revealed that there were more reactive IgM AAbs than IgG in both the cell supernatant and corresponding serum, with a difference of approximately 3–5 times. Immunoglobulin G was predominant in the serum, while IgM was more abundant in the cell supernatant. We identified a range of AAbs present in both the supernatant and the corresponding serum, numbering between 432 and 780, with an average of 53.3% shared. Only 4.7% (n = 23) and 0.2% (n = 2) of reactive antigens for IgG and IgM AAbs, respectively, were specific to CRC. Ultimately, we compiled a list of 19 IgG AAb targets as potential tumor-specific AAb candidates. Autoantibodies against one of the top candidates, p15INK4b-related sequence/regulation of nuclear pre-mRNA domain-containing protein 1A (RPRD1A), were significantly elevated in 53 CRC patients compared to 119 controls (p < 0.0001). The project revealed that tissue-derived IgG AAbs, rather than IgM, are the primary source of tumor-specific AAbs in peripheral blood. It also identified potential tumor-specific AAbs that could be applied for noninvasive screening of CRC.

Abbreviations
• AAb
• autoantibody
• CRC
• colorectal cancer
• GO
• Gene Ontology
• HIS
• hexa-histidine
• RPRD1A
• p15INK4b-related sequence/regulation of nuclear pre-mRNA domain-containing protein 1A
• SYSUCC
• Sun Yat-Sen University Cancer Center
• TAA
• tumor-associated antigen
• TBST
• 0.05% Tween-20 in TBS
• TIL
• tumor-infiltrating lymphocyte
• TIL-B
• tumor-infiltrating B lymphocyte
• TLS
• typical tertiary lymphoid structure
• TME
• tumor microenvironment

INTRODUCTION

Colorectal cancer stands as one of the leading causes of cancer-related deaths worldwide. It ranks third in terms of incidence rates and holds the second-highest position in mortality rates among all malignant tumors.¹ In developing countries, there has been an increasing trend in CRC incidence attributed to shifts in diet and lifestyle associated with economic growth.² Given its insidious onset, prolonged malignant transformation timeline, and the significant influence of the diagnosis stage on prognosis, the implementation of population-based CRC screening programs can effectively address the growing disease burden.³

Colonoscopy is the gold standard for CRC screening, with a sensitivity of over 95%. However, its invasiveness and high costs hamper its application in asymptomatic individuals at risk for CRC, resulting in a low diagnostic rate for early CRC.⁴ Fecal occult blood tests are the most widely adopted noninvasive screening method, capable of detecting cancer or precancerous lesions through the examination of small blood quantities in the stool samples, but have low sensitivity and specificity for early-stage CRC.⁵ In addition, stool tests might be less favored in population-based screening,⁶ whereas blood-based tests offer a more convenient option for screening large populations.

The importance of tumor-infiltrated B cells in the antitumor immune response has been discovered.^6–8 It is capable of detecting and responding to abnormal TAAs, including overexpression, mutations, misfolding, abnormal protein degradation, or posttranslational modifications,⁹ leading to the production of tumor AAbs. Autoantibodies, which can be produced months to years before clinical manifestations, have a long half-life in the blood and can be easily collected through noninvasive blood sampling. Coupled with their high specificity in antigen–Ab responses, tumor AAbs hold great promise as indicators for precancerous intervention and early screening of CRC.¹⁰

However, the heterogeneity of cancer implies that a single universal source of tumor AAbs in the blood does not exist. Therefore, the sensitivity of individual markers that have been screened is often low, and a panel of AAb biomarkers is usually required.^11,12 Moreover, AAb profiling based on blood samples directly might not be highly specific due to the presence of numerous Abs that recognize other autoantigens or high-abundance proteins in the blood circulation.^13,14 While new candidates have been profiled from blood, they appear to have relatively low detection efficiency, which currently precludes their clinical application.

As such, AAbs derived from TIL-Bs in tumor tissue could serve as valuable additions to AAb biomarkers due to their higher specificity for cancer antigens. However, the repertoire of Abs produced by TIL-B in situ is not well understood,^8,15 and most studies on AAb repertoires have focused on plasma or serum antibodies.^10,13,16 Consequently, it remains unclear how many AAbs in the bloodstream specifically target TAAs and how they differ from Abs produced by TLS in the TME.

We, therefore, aimed to elucidate the characteristics of AAb production derived from both CRC tissues and peripheral blood. Specifically, IgG and IgM isotypes and the difference in AAb repertoires were identified. To this end, we used high-throughput human protein microarray chips (HuProt version 4.0; CDI Laboratories, Inc.) to profile AAbs from TIL supernatant and peripheral sera. Finally, one of the top candidate markers was subjected to testing in an optimized ELISA in a case–control study.

MATERIALS AND METHODS Study cohorts

Patients with malignant tumors were selected from the tumor biobank of SYSUCC. In this study, the inclusion criteria for malignant tumor patients' specimens were as follows: (1) all subjects were histologically confirmed with CRC, and (2) the CRC patients did not receive any clinical treatment such as chemotherapy or radiotherapy before blood collection or surgery. A total of 75 CRC patients who were diagnosed in 2013–2022 were recruited as cases. Subjects with gastroenteritis and healthy subjects without evidence of any cancer were selected from the routine general health check-ups in SYSUCC (Table S1).

Hematoxylin–eosin and immunohistofluorescence staining

Tissues were fixed in 10% formalin, embedded in paraffin, sliced into 3–4 μm slices, and stained with H&E. Tumor tissue sections were observed using a transmission electron microscope (Nikon). For immunohistofluorescence staining, tissue sections were double stained with CD3 (ZA-0503; ZsBio) and CD19 (ZM-0038; ZsBio) Abs, followed by HRP-conjugated anti-rabbit Ab (1007910020; Panovue) incubation. The sections were then stained with fluorescent dyes PPD520, PPD570, and DAPI (1007910020; Panovue), and images were taken by a laser scanning confocal microscope (LSM 880; Zeiss).

Tumor-infiltrating lymphocyte isolation and culture

Tumor and normal adjacent tissue were received within 2 h after resection and immediately processed for lymphocyte isolation. Briefly, tissue was washed three times with PBS, and surrounding fat, connective tissue, and blood were removed simultaneously. Tissue was then cut into small pieces and five to six times the volume of digestive liquid was added. The digestive liquid used was prepared using PBS buffer and two enzymes: collagenase IV (at a concentration of 1 mg mL⁻¹) from G-CLONE (product code EZ0300) and DNase I (at a concentration of 0.1 mg mL⁻¹) from Biosharp. Tissue was digested at 37°C with gentle shaking at 150 rpm for 40–50 min to detach the cells. An equal volume of serum-containing medium was added to terminate the digestion. The cell suspension was let stand for 2–3 min and filtered with a 200 mesh filter. Cells were centrifuged at 140 g for 5 min, and washed with 5 mL PBS buffer. Finally, 40% Percoll (Solarbio, product code: P8370) was added to isolated lymphocytes. Cells were calculated and cultured in RPMI-1640, 10% FBS, and 0.5% penicillin/streptomycin for 48 h.

Free AAb detection

Protein binding microarray chips comprised of 21,356 individual human GST and HIS-tagged full-length proteins (Huprot version 4.0) were blocked with blocking buffer (5% BSA in 1× PBST [pH = 7.5]) for 1.5 h at room temperature, then washed with 1× PBST for 5 min. Each serum sample was diluted 2000-fold in blocking buffer, and the TIL supernatant was diluted to a final IgG concentration of 150 ng mL⁻¹ using 5% BSA. All of the supernatants of TILs were validated for IgG expression by western blotting. Samples were then incubated with the blocked proteome microarrays at room temperature for 1 h. Thereafter, the microarrays were washed three times with PBST, for 5 min each time, then Mouse Anti-Human IgM(H + L)-Cy5 (Jackson, Product Code: 709-605-073) and Mouse Anti-Human IgG(H + L)-Cy3 (Jackson, Product Code: 209-165-082) were added (1:2000 dilution). Following incubation with Anti-Human IgM(H + L)-Cy5 and Mouse Anti-Human IgG(H + L)-Cy3, the microarrays were washed with PBST (three times, 5 min each time). Finally, they were spun dry for 2 min and then scanned with an Axon GenePix 4000B. The GenePix Pro 6.0 software (Axon Instruments) was used to extract data from the recorded microarray images. The signal for each spot was defined as the ratio of the foreground to the background median intensity. Data were normalized among different blocks by the formula: I ̂ _x,y,z = I _x,y,z − Median(I _z) + 1, where x is the column number of each protein in a block, x = 1,2,3… 32, y is the row number of each protein in a block, y = 1,2,3… 60, and z is the number of each block in a chip, z = 1,2,3… 24.

Expression and purification of RPRD1A recombinant fragments

The sequence of RPRD1A was amplified from the cDNA of SW480 cells. The sequences were cloned into the pSKB2 vector with a 6× HIS tag on the N terminus and were expressed in Shuffle T7 Escherichia coli (EC2030S; Weidi) as previously reported.¹⁷ Briefly, bacterial cultures harboring the plasmid were grown overnight and then inoculated into the fresh TB medium. Freshly inoculated cultures were allowed to reach an optical density at 600 nm of 0.6. The cultures were subsequently induced with 0.1 mM isopropyl-β-D-thiogalactoside and grown overnight at 16°C, followed by pelleting and lysing the cells using a cell disruptor (NJ Bio) in suspension buffer (10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.1, 140 mM NaCl, 2.7 mM KCl, 5% glycerol, 2 mM DTT, 1 μM DNase I, 1 mM PMSF, and protease inhibitor cocktail I [C0001; TargetMol]). The lysate was centrifugated at 40,000 g for 1 h. The supernatant was filtered and applied to the Ni-NTA column (GE Healthcare) equilibrated with binding buffer containing 20 mM sodium phosphate, pH 7.1, 150 mM NaCl, 15 mM imidazole, and 2 mM β-mercaptoethanol. After washing with increasing concentrations of imidazole, up to 150 mM, the protein was finally eluted with 300 mM imidazole and desalted with a desalting column (GE Healthcare).

Western blot analysis

Purified proteins were fractionated by electrophoresis on SDS–10% polyacrylamide gels, transferred to nitrocellulose membranes, blocked, and then probed with anti-6× HIS primary Ab (ab213204; Abcam) and HRP-conjugated anti-rabbit Ab (A31573; Thermo Fisher Scientific). Immunoreactive bands were visualized using a Bio-Rad Chemidoc MP imaging system. The integrated density of western blot lanes was analyzed using ImageJ software.

Enzyme-linked immunosorbent assay

Enzyme-linked immunosorbent assay detecting the amount of IgG in the supernatant of lymphocytes from normal and tumor tissues was carried out by using a Human IgG ELISA Kit (70-EK171-24; MultiSciences) following the manufacturer's instructions.

The optimized ELISA was developed by using recombinant RPRD1A (3 μg mL⁻¹) as the coating proteins, and HRP-conjugated Affinipure Goat Anti-Human IgG (H + L) (SA00001-17; Proteintech) as the detecting proteins to quantify the AAbs in the plasma. The antigens were diluted in coating buffer containing 50 mmol L⁻¹ carbonate buffer, pH 9.6 (C3041, Sigma), and 96-well microplates were coated with 50 μL per well of the diluted antigens at 4°C overnight. The coated 96-well plates were washed three times with wash buffer containing TBST, followed by 200 μL blocking buffer (5% BSA in TBST) per well for 2 h of incubation at 37°C. After three washes with the washing buffer, 50 μL plasma (1:400 dilution) was added and incubated at room temperature. The HRP-conjugated Affinipure Goat Anti-Human IgG (H + L) (1:10000 dilution) was added into the wells after five washes with the wash buffer. After 1 h of incubation at room temperature and five washes with the wash buffer, the plates were supplemented with 100 μL tetramethylbenzidine solution (KGP125100; KeyGEN). Finally, the reaction was terminated by adding stop solution (KGP12710; KeyGEN), and optical density measurements were taken at 450 and 630 nm on a microplate spectrophotometer (Epoch 2; BioTek).

Statistical analyses

For array data, the cut-off of the A + 6SD (mean + 6× SD of protein signals) was applied to determine a positive hit on protein.¹⁸ In panel 1, the selection criteria were set as a fold change >1.5 between case and control in the serum samples and detection in the cultured supernatant of TILs. To improve the data strength of the selected panels, we combined panel 1 with another database, that is, candidates selected from one pair of TIL–supernatant and peripheral serum from a CRC patient to form panel 2. The AAbs discovered in both panel 1 and panel 2 were identified as candidates.

The enrichment analysis, including GO, was carried out using the clusterProfiler package in RStudio. Mann–Whitney U-test was used to compare the control and case AAb levels.

RESULTS Identification of tumor-driven immune response in TME of CRC

To investigate whether the TME of CRC hosts TAA-driven immune response, microscopy images of H&E and immunohistofluorescence staining showed intratumoral T/B lymphocytes forming TLS, whereas no obvious aggregation was observed in paratumor tissue (at least 3 cm away from a tumor) (Figure 1A,B). Western blot analysis was undertaken to test the reaction between peripheral serum and corresponding tissues from two CRC patients. As shown in Figure 1C, peripheral serum showed a stronger response to tumor tissues than normal paratumor tissue, suggesting an autoimmune reaction present in the microenvironment of CRC tissue. To further investigate the abundance of IgG AAbs, TILs from primary and paratumor tissue were isolated by creating cell suspension within hours of resection. The cultured supernatants were analyzed by western blot and ELISA. As shown in Figure 1D, IgG levels were detected in both cultured supernatants and pellets of TILs, and the expression of IgG was upregulated with the prolongation of culture duration (1–7 days). Immunoglobulin G was highly expressed in supernatants of lymphocytes from tumor tissues compared to adjacent normal tissues of CRC patients, as shown in Figure 1E. These results indicated that there exists TAA-driven humoral immune response in the TME of CRC.

Immunoglobulin M the predominant AAb in both TME and peripheral blood

To investigate the characterization of AAb profiles in antitumor responses, we screened IgG and IgM Abs by high-throughput human protein chip array (HuProt version 4.0) in cultured supernatants of TILs and paired serum, as shown in Figure 2A. The AAbs bound to their target proteins were simultaneously detected using two different fluorescently labeled secondary Abs. Quality control of the protein microarrays was verified by detecting anti-human IgG, human IgM, anti-human IgA, and BSA (Figure S1). As shown in Figure 2B, the number of positive AAb targets ranged from 172 to 358 IgG and 853 to 1218 IgM in TIL supernatants. The number of reactive antigens for IgG and IgM AAb identified in peripheral serum ranged from 202 to 426 and 524 to 1313, respectively. All of the TIL supernatant and serum samples had more distinct IgM AAb targets identified than IgG AAbs, with a difference of approximately three- to fivefold. Signals for both IgG and IgM AAbs to the same antigen observed in each sample were relatively uncommon, with an average of only 156 and 175 in TIL supernatant and serum samples, respectively.

In the TIL supernatant, the total number of the same antigen against IgG among three CRC cases was 115, while the number of common antigens against IgM was 447, and signals for the shared antigens against both IgG and IgM were 47. In peripheral serum, the total number of common IgG and IgM AAbs was 213 and 457, respectively, and there were 81 antigens for IgG and IgM were shared (Figure 2C). These results indicated that both the tissue microenvironment and peripheral blood showed a higher prevalence of IgM as the primary AAb.

Immunoglobulin G is the dominant tissue-derived AAb in peripheral serum

To investigate the relationship between the AAb repertoire in TIL supernatant and peripheral blood samples, we compared the distribution characteristics and differences of AAb profiles. An average of 1273 and 1236 AAb targets were identified in the TIL supernatant and the corresponding serum, respectively (Figure 3A), targeting approximately 6% of the proteome (a total of 21,356 human proteins). The distribution of IgG AAb targets was relatively uniform across different individuals and sample types, whereas the distribution of IgM antigens was more heterogeneous (Figure 3A). Among the different sample types, we found the average number of IgG AAb targets was higher in serum (n = 377) than in TIL supernatant (n = 336). In contrast, the average reactive antigens for IgM AAb were more abundant in the TIL supernatant (n = 1153) than in the paired serum (n = 897). A range of AAbs present in TIL supernatant that were also present in the corresponding serum ranged from 432 to 780 AAb targets, with an average of 53.3% in common among CRC patients (Figure 3A). The proportion of tissue-derived IgG AAb to total IgG in serum was much higher than that of IgM (Figure 3B), indicating that IgG was the predominant antitumoral Ig in peripheral circulation.

Moreover, most of the detected AAbs were also presented in noncancer individuals, with only a few being cancer-specific AAbs. Among three paired TILs and serum samples from cancer patients and three samples from noncancer individuals, only 23 IgG (4.7%) and 2 IgM (0.2%) cancer-specific AAbs were identified (Figure 3C).

Two tumor-specific antigens identified in TIL supernatants against IgG and IgM, PRH1 and AVEN, were selected to show the representative images. The IgM AAb against AVEN antigen identified in TIL supernatant had a stronger signal in corresponding serum than in noncancer gastroenteritis serum. Similarly, the IgG AAb to PRH1 antigen identified in TIL supernatant also had a stronger signal in paired serum than in gastroenteritis serum (Figure 3D).

Specific tumor-derived AAb candidates found in corresponding peripheral serum

We compared serum from subjects with gastroenteritis matched to CRC, which showed 30% of antigen-reactive AAbs in difference. We next sought to determine specific panels of tumor-derived AAbs present in the peripheral serum. The schematic of the workflow is shown in Figure 4A. Promising AAbs were analyzed by the approach that identifies markers from three cases versus three controls, and validates them on the fourth one, as the fourth case was not tested in the same batch. A top list of 34 IgG AAb candidates (including 19 antigens) was finally selected for tumor-specific candidates, which were specifically presented in TILs and serum of CRC patients and expressed at low levels in noncancer controls (Figure 4B). In particular, there were no IgM AAbs screened for CRC-specific biomarkers according to the aforementioned criteria.

We subsequently undertook GO analyses to identify potential biological processes, cellular components, and molecular functions associated with the autoantigens associated with CRC development. Candidate genes were found evenly involved in the B-cell receptor signaling pathway, complement activation, humoral immune response, membrane invagination, phagocytosis, B cell activation, and pyruvate metabolic process with a score of 2. The cellular components of these candidates were significantly enriched in axons and the side of the membrane, followed by blood particles, Ig complex, and mitochondrial protein-containing complex. Moreover, these candidate genes were found to be involved in molecular functions of antigen binding, Ig receptor binding, transmembrane transporter binding, and voltage-gated channel activity (Figure 4C).

Despite CRC having one of the highest tumor mutational burdens, very few CRC tumors had nonsilent mutations in our TAAs. Of 380 CRC patients with publicly available data,¹⁹ 16 TAAs were matched. The rate of nonsilent somatic mutation ranged from 0% to 3.7% (Table S2). This low rate of mutation suggested that the TAAs targeted by AAbs were likely occurring at the protein level.

Autoantibody RPRD1A significantly increased in sera of CRC patients

To investigate whether specific tumor-derived AAb candidates have potential in CRC detection, one of the top antigens, RPRD1A, with an average fluorescence intensity ratio of AAb of 1.83 in case/control serum (Table S3), was selected for further analysis. RPRD1A is highly expressed in tumors and regulates the nuclear location of transcriptional activators, counteracting oxidative stress and promoting cancer development.^20,21

As previously described, purified recombinant RPRD1A protein, with HIS in the N-terminal, was produced using the E. coli expression system.¹⁷ A band of approximately 37.8 kDa is compatible with the calculated molecular weight of HIS-RPRD1A in the public database (Figure 5A). To efficiently test AAbs in serum samples, we optimized an ELISA with recombinant RPRD1A protein as capture molecules to capture AAbs, and HRP-conjugated anti-IgG Ab to quantify the amount of IgG. The schematic of the ELISA workflow is shown in Figure 5B. We found that levels of anti-RPRD1A AAb in 53 CRC patients were significantly increased compared to 119 noncancer healthy controls (Figure 5C; p < 0.0001). However, in 12 benign colorectal polyps, no significance was observed compared with controls (Figure 5C), indicating the specificity of AAbs against TAAs. The area under the receiver operating characteristic curve of anti-RPRD1A AAb was 0.7 (95% confidence interval, 0.61–0.78) (Figure 5D), with a sensitivity of 87% and specificity of 47.1% at a cut-off of 0.764. These findings implied that the candidate antigens hold promise for CRC screening and detection.

DISCUSSION

In this study, we characterized the IgG and IgM AAb repertoire derived from TME and blood circulation and identified the potential of tumor-specific AAbs in CRC detection. Our rationale for assessing tissue- and blood-derived IgG and IgM AAbs repertoire was: (1) our current understanding of AAb profiles in the TME is limited, and (2) the relationship between AAbs in TME and their presence in peripheral circulation remains elusive. Understanding the tumor-derived AAb repertoire will enable opportunities for developing potential tumor biomarkers and therapeutic targets.

We found that IgM was the predominant AAb isotype in both the TME and the blood circulation. The distribution of IgG AAb targets was relatively uniform in sera, whereas the distribution of IgM showed greater heterogeneity. This is consistent with the characteristics of IgM as the predominant Ig isotype, possessing high valence and low affinity, enabling its rapid recognition and binding to diverse antigens.²² A spectrum of AAbs in TIL supernatant, which also appeared in the corresponding serum, ranged from 432 to 780, with an average of 53.3% in common. A similar study on lung cancer found IgG and IgM AAbs present in B-cell lysate that were also present in the corresponding plasma with 56% ± 6.36% SEM in common.²³ These findings corroborate previous reports indicating that TIL-B responses often function independently of systemic humoral responses.²⁴ Overall, the average number of IgG targets in serum was higher compared to that in TILs, while the average number of IgM antigens was higher in TILs than in serum. Additionally, the proportion of tissue-derived IgG AAb to total IgG in serum greatly surpassed that of IgM. These differences could reflect the local production of IgM in response to tumor antigens.²⁵ As cancer progresses, IgM Ab titers tend to decline, whereas IgG responses gradually increase through isotype switching, providing long-term protection.²⁶ These findings suggested that IgG mainly functions in antitumor humoral immunity, while IgM participates in early local immune responses, corroborating prior reports and offering insights into the dynamics of the immune response during cancer progression.²⁷

In our study, only a small percentage of tissue-derived AAb targets (less than 5%) were specific to TAAs, and 15%–30% of AAbs persisted in the peripheral blood of noncancer controls. These results highlight the fact that tumor-derived AAbs alone cannot fully encapsulate the systemic humoral antitumor response. It is worth noting that well-established markers like p53, typically associated with tumors, can also be present in noncancerous individuals, as reported in previous studies.^28,29 This implies that relying solely on individual tumor AAbs might not offer a comprehensive view of the immune response against cancer, and panels consisting of multiple Abs are likely to play a pivotal role in tumor detection.^8,16,22,26

Among the tumor-specific AAbs, 19 tissue-derived IgG AAbs emerged as the primary candidates. The tumor-specific antigens are mainly composed of axon and membrane proteins. Recent evidence has shown that neurogenesis and axonogenesis play a crucial role in tumor development and cancer progression,³⁰ as confirmed by our results that axons are part of TAAs, highlighting their potential as targets for cancer diagnosis and treatment. In addition, most of the TAAs identified displayed a low nonsilent somatic mutation rate (0%–3.7%), similar to previous studies that reported a low rate of mutation (0%–5.8%),³¹ suggesting that the antigens targeted by AAbs were likely to represent proteins that undergo posttranslational modifications, such as glycosylation.³² Notably, one of the top candidate antigens, RPRD1A, showed expression in the cytoplasm of Ab-positive tumors and adjacent normal tissues (Figure S2). This reinforces the notion that the induction of AAbs against TAAs in cancer patients can be triggered by various factors beyond overexpression, including misfolding, aberrant protein degradation, or alternative posttranslational modifications of the proteins. It is essential to highlight that, although RPRD1A is predominantly distributed in the nucleus, its cytoplasmic expression underscores its role beyond the known function of regulating the nuclear location of transcriptional activators.²⁰ The substantial difference in anti-RPRD1A levels between CRC cases and healthy controls validates the potential of tumor-specific AAbs in cancer detection. In addition, our study revealed that RPRD1A AAb levels were higher in pathological stage I disease than in stage III, and remaining stable in stages II–IV (Figure S3). This reinforces the potential of tumor-specific AAbs for early detection of CRC. In particular, the specificity of anti-RPRD1A Abs among tumors appears to correlate with the frequency of antigen expression. We found that in lung cancer, with an RPRD1A expression frequency of 73%, AAbs could effectively distinguish lung cancer from healthy controls (p = 0.04), whereas no such effect was observed in liver cancer, where the RPRD1A expression ratio stood at only 25% (Figure S4).

Despite the care taken in characterizing the IgG and IgM AAb repertoires and assessing their potential in cancer detection, this study has several limitations. First, although we undertook rigorous pairwise screening of samples, the sample size used for high-throughput protein chip profiling was limited. For a comprehensive understanding of AAb profiles, it would be prudent to include a larger number of matching samples in future studies, if feasible. Second, due to the use of different detection batches for the samples from case 4, we were only able to apply the approach that identified markers from three cases and three controls, subsequently validating them on the fourth case for screening. To reduce experimental variation in high-throughput protein array experiments in the future, it would be more practical to test all of the samples in the same batch. Furthermore, our utilization of pooled noncancer serum in the screening process resulted in stringent screening criteria. This led to a limited number of candidate panels being identified and restricted the enrichment of genes in the GO analysis. Finally, additional investigations are required to thoroughly evaluate the efficacy of the candidate biomarkers. This can be accomplished by creating personalized antigen chips aimed at screening for more specific panels and confirming the tumor detection performance of these candidate antigen panels in larger study cohorts. In particular, combining these panels with well-known tumor antigens like p53³³ could yield a more comprehensive panel of screening markers for CRC.

Nevertheless, the identification of the CRC-specific AAb repertoire from TME and peripheral blood holds promise for potential CRC biomarkers, offering a cost-effective alternative to the current diagnostic tools.

AUTHOR CONTRIBUTIONS

Wei-Hua Jia: Conceptualization; data curation; funding acquisition; project administration; supervision; writing – original draft. Pei-Fen Zhang: Data curation; formal analysis; methodology; resources; validation; writing – original draft; writing – review and editing. Ziyi Wu: Formal analysis; resources. Ting Zhou: Resources. Da-Wei Yang: Resources. Quan-Kai Mu: Resources. Wen-Bin Zhang: Resources. Long Yu: Resources. Shao-Dan Zhang: Resources. Ye-Zhu Hu: Resources. Jianbing Mu: Formal analysis; writing – review and editing.

ACKNOWLEDGMENTS

We are grateful to the patients and healthy controls involved in this study. We are also deeply thankful for the support of the Biobank of SYSUCC.

FUNDING INFORMATION

This study was supported by the National Key Research and Development Program of China (grant nos. 2021YFC2500400 to W.H.J., 2020YFC1316902 to Y.S., and 2016YFC1302700 to W.H.J), the Basic and Applied Basic Research Foundation of Guangdong Province, China (grant no. 2021B1515420007 to W.H.J.), and the National Natural Science Foundation of China (grant no. 81973131 to W.H.J.).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICS STATEMENT

Approval of the research protocol by an institutional review board: The study was approved by the ethics committees of SYSUCC (B2021-190-Y01). Informed consent: Written informed consent was obtained from all participants to use their tissue and serum samples. Registry and registration no. of the study/trial: N/A. Animal studies: N/A.