We report on the identification of autoantigens commonly recognized by sera from patients with breast cancer. We selected ten sera from patients with invasive ductal carcinoma (IDC) of the breast with high titer IgG autoantibodies for biopanning of a T7 phage breast cancer cDNA display library. A high throughput method involved the assembly of 938 T7 phages encoding potential breast cancer autoantigens. Microarrays of positive phages were probed with sera from 90 patients with breast cancer [15 patients with ductal carcinoma in situ (DCIS) and 75 patients with IDC of the breast], with 51 non-cancer control sera and with sera from 21 patients with systemic autoimmune diseases. A 12-phage breast cancer predictor group was constructed with phage inserts recognized by sera from patients with breast cancer and not by non-cancer or autoimmune control sera (P < 0.0001). Several autoantigens including annexin XI-A, the p80 subunit of the Ku antigen, ribosomal protein S6, and other unknown autoantigens could significantly discriminate between breast cancer and non-cancer control sera. Biopanning with three different sera led to the cloning of partial cDNA sequences identical to annexin XI-A. IgG autoantibodies reacting with the amino acid 41–74 sequence of annexin XI-A were found in 19% of all women with breast cancer but in 60% of sera from women with DCIS of the breast. In addition, partial sequences identical to annexin XI-A, nucleolar protein interacting with the forkhead-associated (FHA) domain of pKi-67, the KIAA1671 gene product, ribosomal protein S6, cyclin K, elongation factor-2, Grb2-associated protein 2, and other unknown proteins could distinguish DCIS from IDC of the breast and appear to be potential biomarkers for the diagnosis of breast cancer.
Ductal carcinoma in situ (DCIS) of the breast, the earliest form of clinically recognizable breast cancer has been increasingly detected with the use of mammography screening. Although DCIS of the breast is highly curable and the majority of patients do not develop recurrences after 5–10 years, it is now well recognized that it is a heterogeneous group of lesions with a diverse malignant potential (1, 2, 3) . However, because the risk factors for DCIS are similar to those of invasive ductal carcinoma (IDC) of the breast, it is likely that biomarkers associated with DCIS of the breast might be of value for the early diagnosis of breast cancer (1, 2, 3) . Efforts to diagnose breast cancer based on autoantibodies to the hundreds of individual antigens that have been cloned have thus far been largely unsuccessful. Although the range of possible serological tumor markers for breast cancer reported in the literature is broad (4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19) , few have been incorporated into routine oncologic practice, and none have been thought to be of value for the diagnosis of DCIS of the breast (1, 2, 3, 4, 5) .
In this work we combined procedures designed to minimize the confounding effect of unrelated autoantibodies with high throughput methodology to validate our immunoscreening approach. We report on a group of breast cancer autoantigens that are recognized by sera from multiple patients with DCIS and IDC of the breast with potential value for the early diagnosis of breast cancer.
MATERIALS AND METHODS
Patients and Materials.
Sera and pathological specimens with comprehensive 10-year outcome data from 800 breast cancer patients collected by the Karmanos Cancer Institute Breast Cancer Prognostic Study at Wayne State University during the decade from 1975 to 1985 were available for this study (10) . Sera from 10 patients with IDC of the breast were used for immunoscreening a T7 bacteriphage cDNA library of breast cancer proteins. Positive phages cloned with these 10 sera were used to construct a breast autoantigen microarray. Sera from 15 women with DCIS of the breast and from 75 women with IDC of the breast consecutively enrolled in the Karmanos Cancer Institute Breast Cancer Prognostic Study with biopsy-proven diagnosis were chosen to probe the breast autoantigen microarray. All sera were obtained before treatment and were stored frozen at −70°C until use. Fifty-one non-cancer, non-autoimmune control sera were obtained from women attending the rheumatology clinic of Wayne State University with diagnoses of osteoarthritis (20) or fibromyalgia (21) , having no past or family history of breast or ovarian cancers. Osteoarthritis is also called degenerative joint disease or osteoarthrosis, indicating the inherently noninflammatory nature of this joint disease (22) Fibromyalgia is a chronic musculoskeletal disorder diagnosed mainly in middle-aged Caucasian women characterized by widespread pain, characteristic trigger points and other clinical manifestations such as fatigue, sleep disturbances, and irritable bowel syndrome but not characterized by inflammation. Women with similar age and race (breast cancer, mean age at diagnosis of 59.3 years, 88 Caucasian, 2 African American; control women, mean age of 56.4 years, 48 Caucasian, 2 African American, and 1 Asian), having these conditions were chosen as controls because neither osteoarthritis nor fibromyalgia are related to the systemic autoimmune diseases and are not characterized by immunological abnormalities. Additional non-cancer control sera were obtained from women with rheumatoid arthritis (23) and systemic lupus erythematosus (24) . This study had the approval of the Human Investigation Committee of Wayne State University.
Eleven breast pathological specimens from tumors that were kept frozen (−70°C) at the Karmanos Cancer Institute, corresponding to patient sera used to probe the microarray were available for immunohistochemical studies. Five of these specimens had DCIS of the breast and six had IDC of the breast. Immunohistochemistry was performed using the avidin-biotin peroxidase complex technique on 4-μm formalin-fixed tissue sections. After blocking with Super Block (Skytek, Inc.), the sections were incubated overnight at 4°C with polyclonal goat antihuman annexin XI antibody (L-19, Santa Cruz Biotechnology Inc., Santa Cruz, CA) and immunostained with the Elite ABC kit (Vector Laboratories, Burlingame, CA). Images were obtained with a Sony 970X digital camera interfaced with the MCID 5+ system from Imaging Research, Inc. (St. Catherine, ON, Canada).
Selection of Sera and Probing the T7 Display cDNA Library.
We used immunoreactivity as the main criterion to select breast cancer patient sera for biopanning the T7 phage cDNA display library. All 800 breast cancer sera from the Karmanos Cancer Institute Breast Cancer Prognostic Study were used to probe immunoblots of total protein extracts of a breast cancer cell line (MCF10CA1d, clone 1, a malignant variant of ras transformed MCF10A cells; Ref. 25 ) as described previously (10) . We selected sera from 10 patients with IDC of the breast that were collected at the time of diagnosis to identify phage-encoding autoantigens via biopanning of a breast cancer T7 phage cDNA display library (Novagen, Madison, WI) as per the manufacturer’s instructions. These 10 sera were selected because they exhibited strong signals at a dilution of ≥1:500 on immunoblots. In particular, a band at 55–56 kDa was very prominent in several of the sera (Fig. 1) ⇓ . We hypothesized that sera with high titer IgG antibodies would probably lead to the identification of dominant reactivities.
After the final round of biopanning, phages were plated at low density on a lawn of Escherichia coli, and plaque lifts were immunoscreened using patient sera to identify individual positive phages. We reasoned that random picking of plaques without the guidance of antibody recognition would probably lead to dilution of relevant clones. Briefly, A/G agarose beads were incubated with 5 μl of a 1:20 dilution of serum for 1 h at 4°C, washed with PBS plus 1% Tween 20, and then incubated with the T7 phage display library overnight at 4°C. Beads were then washed with PBS plus 1% Tween 20 and used to infect a culture of isopropyl-β-D-thiogalactopyranoside-induced E. coli strain BLT5616. The mixture was shaken until lysis was observed. The lysed culture was clarified by centrifugation, and the supernatant was taken through 4–10 additional rounds of biopanning.
A library of 938 T7 phages encoding potential breast cancer autoantigens was assembled. Plaque-pure phages were grown to high titer in bacterial cultures that were incubated until complete lysis. Supernatants collected after a 10 min × 10,000 × g spin were arrayed in 384-well microtiter dishes. The entire 938 phage library was spotted in duplicate onto nitrocellulose-coated fluorescent array surface technology (FAST) slides (Schleicher and Schuell) using a Flexys robot (Genomic Systems). Each slide was probed either with sera from patients with DCIS, IDC of the breast or control sera, and with a mouse monoclonal antibody directed against a non-variable T7 phage coat protein (Novagen). Patient reactivity was detected using CY3-labeled antihuman secondary antibodies, and the amount of phages present in each spot was measured by binding of CY5-labeled antimouse antibodies. Phage slides also contained 96 “blank” spots (buffer only) to measure background, and 32 “control” spots, which consisted of identical phage clones spotted in 32 equidistant positions to assess positional variability across each slide.
CY3 and CY5 signals were read using a GenePix 4000A slide array reader (Axon Instruments) and quantified using ImaGene software (ImaGene 5.1; BioDiscovery Inc., 1997–2002, standard version). The CY3/CY5 ratios were calculated for each phage spot, and values for duplicate spots were averaged. The difference between duplicates was <5% for 98% of the spots. Although positive phages were clearly visible, we determined a positive cutoff value for each patient serum as a CY3/CY5 ratio >3 SDs above the mean ratio for the 938 signals. We determined the identity of the autoantigens that were recognized by multiple breast cancer patient sera or by control sera on the microarray by sequencing the phage cDNA inserts. After identifying informative phages, the corresponding cDNA inserts were amplified by PCR using primers flanking the insertion site and sequenced by the core facility of the Center for Molecular Medicine and Genetics at Wayne State University. After eliminating phage-related sequences, the GenBank database was searched for sequence similarities to our identified cDNA sequences using the Basic Local Alignment Search Tool program (26) . The majority of the positive phage inserts had unique sequences, but PCR and sequence analysis of some of the reactive phages revealed more than one sequence, and consequently those reactivities were scored as the result of unknown proteins. Searching for sequence identities of the cloned phages in the human expressed sequence tag database showed that some of the cloned phages were closely related, and sequence homologies among the cDNA of the identified positive phages were confirmed by multialignment of the cloned partial sequences using the Basic Local Alignment Search Tool program (26) .
Autoantigens Distinguish Sera from Patients with Breast Cancer from Non-Cancer Control Sera.
The first step to determine whether the phages cloned with sera from patients with IDC of the breast were relevant to breast cancer involved probing the autoantigen microarray with a training set of 7 sera from patients with DCIS, 38 sera from patients with IDC of the breast, and sera from 26 women without cancer (Table 1) ⇓ . Each positive phage was examined separately for a potential association with the diagnoses of DCIS or IDC of the breast by χ2 determination. Starting with the phages most significantly associated with cancer versus the non-cancer group, a 12-phage breast cancer predictor group was constructed stepwise from selected phages that had the ability to increase the set of correctly predicted cancer sera (Table 2) ⇓ . We eliminated from further consideration any phages that reacted with the secondary reagents and gave a positive signal in control experiments in which no patient sera were used in the primary incubation. We also eliminated any phages that were recognized equally by both cancer and non-cancer patient sera for the construction of the cancer predictor. These may encode epitopes similar to those found on common infectious agents or environmental allergens. In a test of the accuracy of our approach, we retrospectively verified that identical phages were each scored as positive by any given patient serum.
In a second step, the group of 12 phages identified in the training set was used as a predictor of breast cancer in an independent group of 45 sera (8 sera from patients with DCIS of the breast and 37 sera from patients with IDC of the breast) and sera from 25 women without cancer. The number of patients of the independent group was set to closely match the training group in the proportions of DCIS and IDC of the breast patients. As a confirmation of the significance of the predictor group, both the sensitivity and specificity were high for the independent group (Table 1) ⇓ . When we combined data obtained from both the training and the independent sets, the predicting ability for breast cancer of the 12-phage predictor group remained intact (P < 0.0001). The results suggest that the breast autoantigens identified have potential predictive value for both DCIS of the breast as well as for IDC of the breast (Table 1) ⇓ . Sera from 12 systemic lupus erythematosus and 9 rheumatoid arthritis patients tested negative with the 12 phages of the selected predictor group for breast cancer.
An intriguing result, that can probably be explained by the method used for selecting the screening sera, was the cloning of 17 identical partial sequences encompassing bp 205–418 of the annexin XI-A cDNA ( 27 ; CB331917) from 3 of the 10 sera from IDC of the breast used for biopanning the T7 phage cDNA display library. Three of these phages were included in the 12-phage breast cancer predictor group because their addition increased the sensitivity and specificity (Table 2) ⇓ . Although the deduced amino acid (aa) sequences were identical, the lengths of the sequences cloned were not the same. Twelve of the 17 annexin XI-A phages contained peptides with the amino acid sequence identical to residues 41–74 of annexin XI-A. Of these 12 phages, 11 were able to significantly distinguish DCIS from IDC of the breast (P from < 0.0001 to 0.05), and only one phage failed to distinguish DCIS from IDC of the breast. The five remaining phages lacked a part of the aa 41–74 sequence and had peptides of unequal lengths (four were shorter, aa 41–58, 41–70, 45–74, 47–74, and one was longer, aa 47–111). None of these five phages were reactive with any of the sera from patients with DCIS of the breast. These findings suggest that the ability to significantly distinguish DCIS from IDC of the breast depends critically on the 41–74 sequence of annexin XI-A. If the sequences of aa 40–43 or 71–74 are missing, the ability of the phage to distinguish DCIS from IDC is lost. Examination of the total group of 90 patients with breast cancer showed that 60% of those with DCIS but only 7% of those with IDC of the breast were reactive with the 12 annexin XI-A phages containing the aa sequence 41–74 (P < 0.00001 with Yates correction for small values). None of the annexin XI-A phages lacking the complete aa sequence 41–74 were recognized by sera from DCIS patients and reacted with only 4% of the sera from patients with IDC of the breast (Table 3) ⇓ .
Immunohistochemical staining of tumor sections with a polyclonal antibody to the NH2-terminal moiety of annexin XI showed that with one exception, the protein expression in the tumors corresponded with positive serum reactivity (DCIS of the breast, 4 of 4, IDC, 1 of 1), and a serum negative for annexin XI-A antibodies corresponded to lack of tissue expression of the protein (DCIS, 0 of 0, IDC, 5 of 5). The exception was one specimen from DCIS of the breast that stained positive for annexin XI whereas the corresponding serum was negative on the microarray. Fig. 2 ⇓ shows examples of DCIS and IDC breast cancer tissue with positive and negative staining for annexin XI.
Multialignment of the partial sequences of the 12 phage predictor group showed that three phages had the 41–74 aa sequence of annexin XI-A, whereas seven phages had homologous sequences suggesting reactivity toward an unknown common antigen without significant homology with any known protein in the GenBank database. The nature of this phage remains unknown and thus we called this antigen UPX. Because the other two phages had no significant homology in the GenBank database (CF751973–4 and CF751975–6), it appears that the 12 phage predictor is based on reactivity toward epitopes located on annexin XI-A and on a small number of unknown proteins.
The phage reactivities of sera from patients with DCIS or IDC of the breast were not significantly associated with the tumor grade, amount of necrosis, or lymphocytic infiltration, although the numbers of patients and particularly of tumor specimens available after stratification by these parameters were relatively small.
Autoantigens May Distinguish DCIS from IDC of the Breast.
The identities of other autoantigens recognized by sera from multiple patients with DCIS and IDC of the breast are listed in Tables 4 ⇓ 5 ⇓ 6 ⇓ . Some of these phage inserts showed complete identity with partial cDNA sequences encoding for known proteins (27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40) , whereas other cloned sequences did not show significant homology with known proteins in the GenBank database and were recorded as unknown proteins. A number of phages showed predominant or exclusive reactivity with sera from patients with DCIS, whereas the majority of the autoantigens reacted with sera from both DCIS and IDC of the breast. The finding of 60% reactivity of annexin XI-A phages with sera from DCIS patients but only 11% positivity with sera from patients with IDC (Table 3) ⇓ , and the ability of certain autoantigens to significantly distiguish between DCIS and IDC of the breast (Table 6) ⇓ suggested the possibility that serum reactivity toward breast cancer autoantigens might reveal different antigen phenotypes. In addition, a number of potentially diagnostic autoantigens were recognized by multiple breast cancer sera and had negligible reactivity with control sera, but the differences did not reach significance (Table 5) ⇓ . Among other known and unknown proteins, the ribosomal protein S12 (30 , 31) , the nucleolar protein interacting with the forkhead-associated domain (NIFK) of pKi-67 (34) , the p80 subunit of the Ku antigen (28) , and cyclin K33 appear to be promising potential markers. Both cyclin K and the Ku antigen were represented by two clones, but the reactive antigen(s) in one of the phages (UP785) is uncertain because both sequences were cloned from this phage.
Some Autoantigens Cloned with Breast Cancer Sera Are Irrelevant to Breast Cancer.
Although most of the antigens listed in Tables 2 ⇓ 3 ⇓ 4 ⇓ 5 ⇓ exhibited negligible reactivity with normal and autoimmune sera, a number of control sera reacted significantly (p from 0.02 to 0.0003) with phage inserts cloned with sera from patients with breast cancer, including those with partial sequences of gelsolin (Ref. 41 ; CB331926, NM_000177), α-2-macroglobulin (Ref. 42 ; CF751963, AAA51552), UP889 (CF751990), UP949 (CF751991-CF751992), and UP275 (CF751994-CF751995). None of 90 breast cancer patient sera reacted with any of these phage inserts on the microarray.
Hundreds of autoantigens have been cloned via recognition by antibodies in cancer patient sera; however, efforts to predict malignant disease based on autoimmunity to individual antigens or to groups of antigens have thus far been largely unsuccessful. This failure may in part reflect problems inherent to the specificity of the immunoscreening procedure as well as to the lack of systematic methods for the identification of informative autoantibodies from the scores of antibodies that are either patient-specific or irrelevant to the disease. The use of SEREX (serological analysis of recombinant tumor cDNA expression libraries) and proteomics methodologies (10, 11, 12, 13, 14, 15, 16, 17, 18, 19 , 43) led to the identification of a large group of autoantigens in breast cancer patient sera. Although in aggregate these studies strongly suggest that autoantibodies have potential as biomarkers, thus far, they have not resulted in serological markers with definitive predicting ability for breast cancer in the clinical arena (4 , 5) , and none have been thought to be of value for the diagnosis of DCIS of the breast (1, 2, 3, 4, 5) . Because autoantibodies are part of the normal immune response (44) , one important problem inherent to autoantibody-based methods for identifying tumor-related antigens is demonstrating their tumor relevance. Here we report that a collection of breast cancer autoantigens cloned by screening a T7 phage cDNA library of breast cancer proteins are recognized by multiple sera from patients with DCIS and IDC of the breast but not by sera from non-cancer controls. Moreover, some of these antigens were able to distinguish DCIS from IDC of the breast. The sera we selected for screening the T7 phage cDNA library exhibited high titers of an IgG antibody reacting with a 56 kDa antigen on immunoblots of human breast cancer proteins (Fig. 1) ⇓ . We hypothesized that the gene encoding this antigen would be over-represented in the phages cloned with these sera, thus allowing the identification of dominant reactivities. As we expected, several phages cloned using sera from three different patients with IDC were identical, and sequence analysis of the phage inserts showed identity with a partial sequence of annexin XI-A (27) . The findings of a large number of breast autoantigens exhibiting the ability to differentiate breast cancer sera from normal sera (Tables 1 ⇓ 2 ⇓ 3 ⇓ 4) ⇓ and yet, other antigens that react preferentially with DCIS of the breast (Table 6) ⇓ suggest that our strategy is effective, allowing the identification of autoantigens relevant to breast cancer. We have previously used this strategy to clone RPA32, using a serum from a patient with breast cancer who exhibited a high titer of 32 kDa (10) . Because the 56-kDa autoantigen has also been cloned with a serum from a patient with lung cancer using SEREX (45) , it is possible that autoantibodies to annexin XI-A may also be found in other malignancies.
Annexin XI is a member of the annexin superfamily of Ca2+ and phospholipid-binding, membrane-associated proteins implicated in Ca2+-signal transduction processes associated with cell growth and differentiation (46, 47, 48, 49, 50, 51, 52, 53) . Annexin XI may have a role in cellular DNA synthesis and in cell proliferation as well as in membrane trafficking events such as exocytosis and has been found to be identical to a 56-kDa antigen recognized by antibodies in 3.9% of patients with systemic autoimmune diseases (54) . Misaki et al. (27) showed that antiannexin XI positive sera from patients with systemic autoimmune diseases recognize an epitope(s) residing in the NH2-terminal moiety of the molecule. They showed that mutants containing only part of the annexin XI aa 1–123 were recognized by all autoimmune sera, but a mutant containing only the NH2-terminal 32 aa(s) was not reactive with any of the sera tested, and that removal of aa 1–49 eliminated the reactivity of four of five sera. One of the sera tested was still able to immunoprecipitate this short peptide, indicating that at least one reacting epitope is located in this region of the molecule. The precise location of the other epitope reacting with autoimmune sera is uncertain, but from the study of mutants it might be located in the region spanning positions 50–123 of the annexin XI sequence (27) . The partial sequence of annexin XI-A cloned by breast cancer sera spans residues 41–111, but the sequence aa 41–74 appears critical for distinguishing DCIS from IDC of the breast (Table 3) ⇓ . None of the cloning sera used in this work were obtained from patients with systemic autoimmune diseases, and sera from 21 patients with systemic lupus erythematosus and rheumatoid arthritis did not react with the phage inserts containing annexin XI-A. However, neither our data nor the previous studies (27) eliminate the possibility that both cancer and autoimmune sera may react with identical epitopes.
There is a parallel between breast cancer and autoimmune diseases in reference to serum reactivity to annexin XI and RPA32. The prevalence of anti-RPA32 was reported to be 11% and that of annexin XI is approximately 19% in breast cancer sera, whereas the frequency of these antibodies in the systemic autoimmune diseases has been estimated to be 2%–3% and 3.9%, respectively (10 , 54) . It is pertinent that both systemic lupus erythematosus and Sjögren’s syndrome are known to be associated with a tendency to develop lymphoid malignancies (55, 56, 57) . There are reports on the cancer-predicting ability of several members of the large annexin family that are suspected to be involved in the process of carcinogenesis (46, 47, 48, 49, 50, 51, 52) . Thus, it is possible that the antibodies to RPA32 and to annexin XI in the sera of a small proportion of patients with systemic autoimmune diseases may represent early markers of malignancy.
To our knowledge this is the first report on annexin XI-A cloned with sera from IDC of the breast and recognized as an autoantigen by sera from multiple patients with breast cancer. It may be of interest that the sequence that we identified as critical for the recognition of DCIS of the breast spanning aa 41–74 is located in the regulatory NH2-terminal moiety of annexin XI-A, which contains the nuclear localization signal (58) , PEST sequences believed to be signals for rapid intracellular degradation (59) , as well as the aa residues essential for binding the annexin XI-A isoform with calcyclin (S100A6; Ref. 60 ).
Antibodies to the ribosomal protein S6 (29) , nucleolar protein interacting with the FHA domain of pKi-67 (34) , elongation factor 2 (37) , Grb2-associated binding protein 2 (40 , 41) , the KIAA1671 gene product (32) , and other autoantigens listed in Tables 4 ⇓ 5 ⇓ 6 ⇓ have not been reported previously in the sera of cancer patients. The reactivities of the phage inserts with multiple sera from patients with breast cancer sera are listed in Tables 2 ⇓ 3 ⇓ 4 ⇓ 5 ⇓ . We propose that a number of the autoantigens reported here, including annexin XI (27) , the ribosomal proteins S6 (29) and S12 (30 , 31) , the p80 subunit of Ku (28) , and cyclin K33 are potential biomarkers of breast cancer, whereas annexin XI (27) , the nucleolar protein interacting with the FHA domain of pK-67 (NIFK; Ref. 34 ), the KIAA1671 protein (32) , elongation factor 2 (37) , the ribosomal protein S6 (29) , and other known and unknown proteins (Table 6) ⇓ showed potential in their ability to differentiate patients with DCIS from those with IDC of the breast, suggesting that these autoantigens are candidates as biomarkers for the early detection of breast cancer.
Our previous work (10 , 61 , 62) and the work of Tan (63) suggested that screening autoantigen expression libraries with cancer patient sera containing high titers of autoantibodies has the potential of revealing a number of proteins that may be involved in cellular functions related to tumorigenesis. Because expressed sequence tags represent a copy of a part of the genome that is being expressed, we expected that their identification by immunoscreening a T7 display library of breast autoantigens would allow us to obtain gene expression data in breast cancer. We speculated that this approach might allow us to clone cognate genes products, perhaps related to the signal transduction mechanism(s) that may be activated in cancer.
The identification of annexin XI (27) , cyclin K (33) , ribosomal proteins S6 (29 , 64 , 65) and S12 (30 , 31) , Grb2-associated protein 2 (39 , 40 , 66) , and elongation factor 2 (37 , 67) , as autoantigens recognized by breast cancer sera could be the expression of molecular alterations in the signal transduction mechanism in breast cancer. Annexin XI interacts specifically with calcyclin, a member of the S100 subfamily of elongation factor-hand calcium-binding proteins, which are involved in Ca2+-regulated signaling pathways and found in abundance in certain breast cancer cell lines (68 , 69) . Because calcyclin is overexpressed in tumor cells with the mRNA specifically elevated in the G1 phase of the cell cycle of stimulated cells, it has been suggested that annexin XI-calcyclin complexes may play a role in cell proliferation and cell division (69) . Cyclin K is an RNA polymerase II-associated cyclin involved in transcriptional activity (70 , 71) . The p80 subunit of the Ku antigen has been reported to be recognized by autoantibodies from some patients with the scleroderma-polymyositis overlap syndrome (28) . The heterodimeric Ku protein is the DNA-targeting component of a DNA-dependent protein kinase that plays a critical role in mammalian DNA double-strand breaks repair and has been widely implicated in tumor biology (72 , 73) . The finding of an autoimmune reaction directed toward the Ku antigen and our previous report of autoantibodies to RPA32 in breast cancer patient sera (10) suggests that the molecular changes leading to autoimmunity of proteins involved in DNA repair may be important in breast carcinogenesis. Grb2 is an adapter protein that binds activated growth factor receptor molecules such as receptor tyrosine kinases and transduces signals leading to the activation of the Ras-mediated signal cascade activated in most cancers (66) . Elongation factor 2 is phosphorylated by a calmodulin-dependent protein kinase, CaM K III, which is selectively activated in proliferating cells, and its activity is elevated in human breast cancer (67) . Phosphorylation of ribosomal protein S6 is a common effect of mitogenic stimulation of cells (29 , 64 , 65) , and its kinase is thought to be a downstream target of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK) pathway (65) .
The finding of annexin XI-A reactivity in the sera of 60% of patients with DCIS and in only 11% of those with IDC of the breast, although statistically significant, could possibly be an artifact attributable to the relatively small sample of patients with DCIS. Alternatively, we speculate that it could be the reflection of antigenic heterogeneity in breast cancer. Ductal carcinoma in situ seems to be a heterogeneous group of lesions, and the relation between in situ and invasive carcinomas of the breast is not clear. The study of histological grade and tumor markers has identified heterogeneity in breast cancer (74 , 75) . The higher reactivity to annexin XI-A phages with sera from patients with DCIS than with those with IDC of the breast as well as the ability of certain autoantibodies to significantly distinguish DCIS from IDC of the breast (Table 6) ⇓ could be interpreted as an indication that some invasive carcinomas may not go through the DCIS stage characterized by annexin XI-A reactivity. In view of the long life of IgG molecules that are the expression of an established immune response, we would expect to find IgG antibodies in their sera if they were preceded by a DCIS stage characterized by annexin XI-A positivity, even if the gene for annexin XI-A had been presumably turned off in some patients with IDC in their transition from DCIS.
Autoantigens cloned by immunoscreening cDNA expression libraries by breast cancer patient sera are not necessarily related to breast cancer. Results obtained in the study of the non-cancer control sera are of interest to interpret the significance of phage inserts cloned with certain breast cancer patient sera. A number of phage inserts cloned with sera from patients with IDC of the breast, including partial sequences identical to gelsolin (41) and to α-2 macroglobulin (42) , were significantly recognized by normal control sera and not by cancer patient sera. In the case of gelsolin, the recognition of this sequence by autoantibodies present in the serum of a patient with breast cancer could falsely be attributed to molecular changes in gelsolin known to occur in breast cancer (76 , 77) . However, a role for gelsolin has also been suggested in the pathogenesis of Parkinson’s disease and in the Finnish-type familial amyloidosis and related conditions (78 , 79) , and the pan-proteinase inhibitor α-2-macroglobulin has been widely implicated in the pathogenesis of Alzheimer’s disease (80 , 81) . The significance of the presence of these autoantibodies in the non-cancer control sera is unknown, but it is clear that despite the origin of the cloning sera obtained from patients with breast cancer, these autoantigens are irrelevant to breast cancer, and might be related to autoimmunity in the degenerative brain diseases or to other causes. The identification of other phage inserts recognized by normal sera will be of great interest because these antigens are undoubtedly irrelevant for breast cancer but may reflect autoimmune phenomena related to a host of other conditions affecting the general population, may be associated with aging, or may be a part of the normal autoimmune response.
In view of our findings, it is likely that probing this autoantigen microarray prospectively with sera from a large cohort of breast cancer patients may allow the identification of biomarkers with diagnostic significance and perhaps may allow the identification of discrete antigen phenotypes with clinical significance. The high prevalence of IgG autoantibodies in the sera of patients with DCIS and IDC of the breast suggests that they are potentially excellent candidates as biomarkers for the early diagnosis of breast cancer.
We thank Dr. John E Tomkiel for collaboration in the initial phase of this work and Drs. Gloria Heppner, Bonnie Sloane, and Michael Tainsky for helpful comments and advice.
Grant support: NIH R21 AR-99-128 (CA-87759), grants from the Michigan Chapter of the Lupus Foundation, the Karmanos Cancer Institute, Mary Webber Parker, and the Flora Temple Fund.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Felix Fernandez-Madrid, University Health Center, 4H, Department of Internal Medicine, Division of Rheumatology, Wayne State University School of Medicine, 4201 St. Antoine Blvd., Detroit, MI 48201. Phone: (313) 577-1133. E-mail:
- Received April 8, 2003.
- Revision received April 20, 2004.
- Accepted May 28, 2004.
- ©2004 American Association for Cancer Research.