Longitudinal autoantibody responses against tumor-associated antigens decrease in breast cancer patients according to treatment modality

Early diagnosis of breast cancer (BCa) is critical for increased disease survival, both at the time of initial disease as well as for recurrence [1, 2]. Including all races, the 5-year survival rate for women diagnosed with local BCa is 99%. Survival declines to 84% for regional stage and plummets to 26% for distant stage [3–5]. Approximately 6–10% of U.S. women present with metastatic BCa when first diagnosed [5–7], and 20 to 50% of patients initially diagnosed with primary BCa will develop metastatic disease [7]. Incurable metastatic disease is responsible for most BCa-related deaths. Unfortunately, recurrence of BCa is most often found when patients report symptoms, such as shortness of breath, chronic cough, headache, weight loss or bone pain. Once a patient is diagnosed with metastatic BCa, the intent for treatment is no longer curative; instead, the goal is to control the disease for as long as possible [8, 9]. Presently, no alternative screening methods are recommended for asymptomatic patients without clinical findings on physical examination [10]. By the time the metastases are identified by physical symptoms, the patient’s chances of surviving the disease are greatly diminished [11]. Lack of early detection of recurrence for asymptomatic patients is a factor in the inability to cure metastatic BCa.

Numerous previous studies have detected the presence of autoantibodies to various cancer proteins in BCa patients [12, 13]. Measurement of these autoantibodies could offer early detection of BCa recurrence before physical symptoms are apparent, which would have a monumental impact by giving patients the choice of appropriate treatment options at an earlier stage of disease [14]. We previously screened patient plasma samples for autoantibodies against 20 different tumor-associated antigens (TAAs) and compared the profiles to healthy women utilizing an ELISA-based platform. The 20 antigens were selected from a membrane-associated polyribosomal cDNA library (MAPcL), which encodes membrane and secreted proteins highly expressed in BCa and should preferentially induce an antibody response in patients [15]. The conformation of membrane and secreted proteins is particularly important because discontinuous epitopes will only be present for antibody recognition when the antigen is folded properly. Expression constructs were generated to encode the extracellular portion of the TAA fused to rabbit Fc (rFc), and a eukaryotic expression system was developed to produce conformation-carrying antigens that are processed with post-translational modifications [16]. A panel of 7 conformation-carrying TAAs, consisting of ANGPTL4, DKK1, LGALS1, MUC1, GFRA1, GRN and LRRC15, was identified that in combination could discern newly diagnosed BCa patients from healthy controls with a 73% sensitivity and 76% specificity [16]. To continue in the development of this technology, the single-well ELISA-based autoantibody assay was transitioned to a multiplex Luminex platform. The Luminex xMAP microsphere technology (Luminex, Austin, TX) allows measurement of the interaction of patient autoantibodies with a panel of antigen biomarkers enabling quantitation across all biomarkers in a single well [17].

It is unknown what, if any, changes occur to a BCa patient’s autoantibody profile after the tumor is removed and treatment begins. Because the autoantibodies specifically recognize cancer antigens, the autoantibody response may decrease during and after treatment due to the reduced tumor burden in the body. To address this question, the multiplex bead assay was performed on serial blood draws of 200 newly diagnosed BCa patient samples. Autoantibody responses against 32 antigens were determined over the course of both time and treatment, including surgery, radiation, chemotherapy, antibody and hormonal therapies.

Four of the 12 treatment groups received trastuzumab therapy (17 patients total, Table 3). Trastuzumab is a humanized monoclonal antibody that is generally administered every 3 weeks for 1 year to patients with amplification of the ERBB2 gene [30]. Since ERBB2 is one of the TAAs included in our assay, the treatment of these patients with trastuzumab provided a fortuitous internal spiked control. We determined the signal levels of the serial blood draws against ERBB2 for the 17 patients. Two main response patterns were observed correlating with time of trastuzumab administration. First, for patients that had discontinued trastuzumab therapy after 6 months, their anti-ERBB2 antibody responses for the 12-month blood draws decreased (Fig. 1, represented by patients BC-082 and BC-149). The second pattern is represented by patients BC-018 and BC-019 in Fig. 1. These patients continued to receive trastuzumab during the longitudinal blood draws, and the anti-ERBB2 antibody response against the ERBB2 antigen plateaued between the 6 and 12-month visits (Fig. 1).

For the primary exploratory analysis, we excluded the 17 patients who received trastuzumab as part of their therapy. Therefore, the analysis included 183 out of 200 patients and encompassed 8 of 12 patient treatment groups (Table 3). A repeated measure ANOVA was used to model the geometric mean changes from baseline for each autoantibody at 6 and 12 months, and the model included all interactions among indicator variables for radiation, hormonal therapy and chemotherapy with time. If the ANOVA model predicted at least 3 significant changes for an antigen among the 16 estimates (8 groups * 2 time points consisting of 6 and 12 month blood draws), it was chosen for further analysis. Using this variable selection criterion, the model identified 11 antigens that were likely to have significantly modulated autoantibody signals in response to treatment in the 12 months following surgery. The TAAs chosen for further study included: A1AT, ANGPTL4, CAPC, CST2, DKK1, GFRA1, GRN, GRP78, LGALS3, LRP10 and NY-ESO-1.

The actual geometric mean changes from baseline were calculated for the 11 TAAs at 6 months (Additional file 2) and 12 months (Fig. 2). No significant changes in the autoantibody response against the 11 antigens were observed for patients who received surgery alone or were treated with surgery and a single monotherapy: hormonal therapy, chemotherapy or radiation. While NY-ESO-1 was predicted to have significant changes in levels of response in the repeated measures ANOVA model, the observed changes were not significant (Fig. 2). However, significant decreases in levels of response between baseline and 12 month time points against 9 of the 11 TAAs (i.e. A1AT, ANGPTL4, CAPC, CST2, DKK1, GFRA1, GRN, LGALS3 and LRP10) were observed in three treatment groups. Radiation + chemotherapy, radiation + hormonal therapy, and radiation + hormonal therapy + chemotherapy had average geometric mean decreases for the 9 significant TAA of − 11, − 13, and − 18%, respectively (Fig. 2). In the radiation + hormonal therapy + chemotherapy group, A1AT and LRP10 had the greatest decrease in autoantibody response levels (− 26 and − 28%, respectively) compared to the other antigens and regimens. The triple therapy of radiation + hormonal + chemotherapy was more effective at reducing the autoantibody responses against the TAAs than any other combination of treatment. In addition, the decrease in response of the geometric mean was − 15% at 6 months after surgery (Additional file 2) compared to a greater decrease of − 18% at the 12-month time point (Fig. 2). It is apparent that autoantibody levels diminish during the course of treatment, and they continue to decrease with time as the patient is further removed from initial diagnosis. At the 12-month time point, GRP78 was the only TAA to exhibit a significantly increased autoantibody response of 44%, which occurred in the hormonal therapy + chemotherapy group (Fig. 2).

Measurement of autoantibody responses to cancer proteins holds promise as a minimally invasive screening method in cancer diagnostics and prognostics. Numerous TAAs that elicit an immune response have been identified, and findings in this realm continue to expand [31–33]. Although studies have shown the presence of the autoantibodies at the time of BCa diagnosis [13, 34, 35], little is known about how these antibodies are maintained over time and affected by treatment. This study provides important data on longitudinal changes of autoantibody response occurring in BCa patients after surgery. We were able to assess 32 TAAs simultaneously with our unique multiplex bead assay. In addition, rich data on pathology and treatment information allows the findings of this study to be associated with the treatment regimens administered to each patient.

Patients in this study that were ERBB2-amplified received trastuzumab treatment, which recognizes and binds to a native pocket-like binding region of ERBB2 [36, 37]. As predicted, the levels of trastuzumab against ERBB2 in patient plasma correlated temporally with patient treatment (Fig. 1). To remove the confounding variable of trastuzumab treatment, patients treated with the anti-ERBB2 antibody were removed from the primary analysis. The data collected from the longitudinal blood draws indicated that significant changes in antibody responses were detected at the 6-month time point (Additional file 2), with a more extensive reduction detectable 12 months after the primary tumor was surgically resected and therapy initiated (Fig. 2). The absence of a large tumor mass is one explanation for the reduced production of autoantibodies to TAAs. With the majority of the cancer cells removed, the immune system would be exposed to fewer cancer antigens reducing the levels of autoantibody production. Alternatively, the decrease in autoantibody activity could be due to radiation + chemotherapy and/or hormonal therapy-induced immunosuppression. The fact that a single treatment modality after surgery, including radiation alone, did not alter the levels of autoantibodies indicates that it is not simply due to treatment-induced immunosuppression.

Surprisingly, we also found that treatments administered after the removal of the primary tumor had a profound effect on the autoantibody profile. GRP78 showed an increase in autoantibody signal at the 12-month time point in patients receiving chemotherapy followed by hormonal therapy. This protein is typically localized to the endoplasmic reticulum [38], but it has been shown to translocate to the cell surface in cancer cells [39], particularly in times of cellular stress [40]. The cellular stress imposed on tumor cells by anti-neoplastic treatments would explain a change in localization of GRP78 from the endoplasmic reticulum to the surface, making it more available to the immune system in its aberrant location. Production of anti-GRP78 antibodies could occur in this situation, explaining the outlying antibody increase obtained for this TAA.

There is one confounding factor that is difficult to overcome when attributing the data to the treatment groups: the treatment given to a patient is dictated by the physical characteristics of the tumor, i.e. the presence of the estrogen receptor, amplification of ERBB2, size of the tumor, in situ versus invasive components and lymph node involvement. The changes seen in these samples are attributed to the therapies administered to that patient, but it is acknowledged that these therapies are a function of the characteristics of each subtype of BCa [41, 42].

To that end, the treatment modality groups that had the greatest decrease in autoantibody response levels were radiation + hormonal therapy; radiation + chemotherapy; and radiation + hormonal therapy + chemotherapy (Fig. 2). The common denominator of the three most affected groups for significant changes in autoantibody response levels is radiation treatment. However, radiation treatment alone is not enough to significantly decrease the response levels of the autoantibodies (Fig. 2).

Four antigens (ANGPTL4, DKK1, GFRA1 and GRN) overlapped with the ability to discern BCa patients from healthy [16] and to elicit an autoantibody response against them that was significantly lowered over the course of treatment (Fig. 2). Yet, three of the seven previously characterized antigens (LGAL1, LRRC15 and MUC1) were not altered longitudinally in the current study. It is possible that these antigens only have utility in the initial diagnosis of BCa and do not exhibit a statistically significant reduction in autoantibody levels after treatment. Another explanation is that the 7 antigens utilized to discern health from cancer patients were analyzed as a group and not individual responses against each antigen.

The ultimate goal of detecting autoantibodies in the follow-up setting is to determine if the autoantibody response to a panel of cancer antigens can predict patient response to treatment and detect recurrence of the disease. Future studies will include acquiring serial blood draws before and after the patient presents with physical symptoms of recurrence. The fact that the autoantibody levels of BCa patients decrease over the course of treatment when radiation is used in combination treatment modalities supports the potential of using the detection of these autoantibody levels as a prognostic indication of recurrence.

Adjuvant treatment following the removal of the primary tumor had an extensive effect on the autoantibody profile of the patient. Autoantibody levels of BCa patients against TAAs decreased over the course of treatment when radiation was used in combination treatment modalities. Using a Luminex-based multiplex bead analysis, significant decreases in levels of response against A1AT, ANGPTL4, CAPC, CST2, DKK1, GFRA1, GRN, LGALS3 and LRP10 were observed between baseline and 12 month time points in three treatment groups: Radiation + chemotherapy, radiation + hormonal therapy, and radiation + hormonal therapy + chemotherapy. A significant increase in autoantibody response was observed against GRP78 in the hormonal therapy + chemotherapy group.