Naturally occurring B-cell responses to breast cancer

Abstract

As demonstrated by the effectiveness of trastuzumab, antibodies against breast cancer antigens are a potentially potent mechanism of tumor control. While trastuzumab is administered exogenously, its efficacy suggests that induction of very high titer antibody responses in vivo might also be therapeutic. Both naturally occurring and vaccine-induced antibody responses to some breast cancer antigens are associated with improved survival in some cases. However, the improvement in survival associated with antibody responses to breast cancer is modest, and tumor regression is not known to be associated with the natural antitumor antibody response, indicating a need for improved understanding of the natural antitumor antibody response. Naturally occurring B-cell responses in the form of serum antibody, tumor reactive lymph node B cells, and tumor-infiltrating B cells have been described, and a variety of breast tumor–associated antigens have been identified based on reactivity of patient antibodies. This review discusses current knowledge of humoral immunity to breast cancer with regard to specific antigens and the basis for their immunogenicity, and the contexts (tumor, lymph node, serum) in which responses are observed. With few exceptions, "tumor-associated antigens" identified with naturally occurring antibodies may be overexpressed on tumor but are in fact nonspecific autoantigens. This suggests that while overexpression or aberrant processing can increase immunogenicity in some cases, the immunogenicity of many or even most tumor-associated antigens is a function of expression in tumor or the result of ancillary tumor factors.

Introduction

Naturally occurring and induced antibodies have provided potent protection from viral, bacterial, and parasitic infections. However, antibodies have only recently demonstrated potent antitumor activity in clinical trials. As a single agent, trastuzumab produced an 11–15% overall response rate against metastatic breast cancer in phase II trials [1]. In combination with chemotherapy, trastuzumab produced a response rate of up to 50%, and increased time to progression and overall survival compared with chemotherapy alone (reviewed in [2]). The success of trastuzumab against bulky solid breast tumors suggests that induction of high titer antibodies against tumor-associated antigens might be an effective means of tumor control. There are rare reports of spontaneous regression of breast cancer; some of those regressions are presumed to be immunologic (reviewed in [3]). Ross et al. described 13 cases of "partial or complete tumor regression in the absence of any treatment capable of producing regression." One report suggested an antibody-mediated regression, as subsequent murine and human breast carcinomas were successfully treated with ascites fluid from the primary patient [4]. However, spontaneous regression is rare, despite the presence of serum antibody against breast tumor–associated antigens in approximately 50% of patients [5, 6].

Why are naturally occurring antibody responses against breast cancer not more effective? A variety of factors including lymphocyte trafficking, tumor expression of CD59 and complement regulatory protein, antigen loss, HLA expression, T-cell activation, and dendritic-cell migration contribute to antibody-mediated antitumor immunity and B-cell activation, maturation, and differentiation. However, tolerance has emerged as a constant and formidable obstacle to cellular and vaccine-based immunotherapies of cancer, suggesting that tolerance constitutes a similar obstacle to naturally occurring responses. Most tumor-associated antigens (including HER-2/neu) would more properly be described as tumor-associated autoantigens, since few if any breast tumor–associated antigens or epitopes are truly tumor specific. As such, they are afforded the same protection from immunity as other autologous proteins. Studies have demonstrated that vaccination of breast cancer patients with allogeneic protein induces normal cellular and humoral immunity, but that vaccination with tumor-associated antigens fails to induce high titer responses [5, 7, 8]. This demonstrates that the deficiency in response to tumor antigens is not a function of systemic immunodeficiency, but specific tolerance to tumor-associated antigens.

Given this maintenance of tolerance, how is tolerance to tumor-associated autoantigens broken at all? A wide variety of ubiquitously expressed tumor-associated autoantigens including actin elicit spontaneous antibody responses [9, 10], suggesting that expression in tumor is sufficient to initiate immunity against virtually any protein. However, responses against tumor-associated autoantigens fail to reach high titers because of tolerance mechanisms. A recent murine study provides support for this model. In this study by Bondanza et al., dendritic-cell vaccination of wild-type mice with autoantigen readily induced low titer primary antibody responses but failed to induce high titer secondary responses [11]. In contrast, mice with a genetic background of autoimmune susceptibility mounted high titer antibody responses against autoantigens. An analogous situation may exist in breast cancer. While naturally occurring anti–HER-2/neu antibody responses have been described in 20% of patients with HER-2/neu overexpressing tumors [12], they do not approach concentrations determined to be effective in vivo. Serum anti–HER-2/neu antibodies were ≤8.93 μg/ml in the sera of breast and ovarian cancer patients judged to have "high titer" anti–HER-2/neu antibodies [12, 13]. This titer is approximately 20-fold lower than the optimal therapeutic mean trough serum concentration (~190 μg/ml) of the anti–HER-2/neu antibody trastuzumab [1]. In addition, naturally occurring anti–HER-2/neu antibody titers are higher than those observed for some other breast tumor–associated antigens. Naturally occurring anti-MUC1 serum antibodies in breast cancer patients had a median value of 0.8 μg/ml [14, 15]. Despite multiple vaccinations with MUC1 peptide, anti-MUC1 titers reached a median high of only 18 μg/ml, decreasing to 5.5 μg/ml after 21 weeks. It is likely that antibodies against MUC1 and other tumor-associated targets will require high levels of antibody for effective antitumor activity in vivo. Many naturally occurring antibodies are also directed against cytosolic antigens, and would be unlikely to have any direct antitumor activity.

In this review, we catalogue the tumor-associated antigens that are known to elicit naturally occurring antibody responses in order to identify tumor-associated antigens with some degree of native immunogenicity. Because antibody responses are a requisite component of effective antitumor immune responses in murine systems [16] and associate with clinical course in human antitumor vaccine protocols [17, 18], selection of naturally immunogenic targets may be key to cancer immunotherapy.

In looking at naturally occurring antibody responses against breast cancer, we also wished to determine how the anatomical and physiological setting of the antibodies (peripheral blood, node, and tumor) affects the immunobiology of the response, the immunogenicity of tumor-associated antigens, and the ability of B cells to respond to tumor-associated antigens. Although the initial repertoire is selected against autoantigen reactivity, it is unlikely that lymph node cells are exposed to every possible autoantigen throughout the life of an individual. Lymph node dendritic and follicular dendritic cells may be exposed and tolerized to antigens expressed in tissues draining to that node, but not to antigens absent or expressed at low levels in local tissue. As a result, antibody responses against autoantigens absent or expressed at low levels in local tissue may be initiated in draining lymph nodes, yet fail to convert to a high titer systemic response because of systemic tolerance to the autoantigen. We also explore the significance of tumor-infiltrating B cells. As we have described previously, breast tumors contain B-cell follicles [19]. However, the cytokine, cellular, and architectural microenvironment in the tumor differs from that of the node. Cytokines are secreted from the tumor itself, from cells in the tumor microenvironment, and from infiltrating leukocytes. Also, it is unclear that tolerance and B-cell differentiation mechanisms would operate normally in an ectopic follicle. Outside of breast cancer, ectopic follicles have previously been described only in the context of autoimmune disease [20, 21, 22, 23], where these germinal centers are etiologic in the disease.

Serum antibodies against breast cancer antigens

One of the earliest studies of serum-antibody responses to breast cancer reported elevated circulating antigen/antibody complexes in the sera of breast cancer patients in comparison to healthy controls [24]. Based on this, a natural immune response against autologous tumor was postulated. Subsequent studies demonstrated that antibody responses to primary tumor occur in approximately 50% of breast cancer patients [5, 6]. Breast cancer patient sera were reactive with cytoplasmic components of tumor cells, and less reactive with tumor cell surface. Subsequently, many tumor-associated antigens have been discovered using antibodies from patient sera. Reactive antigens have been overwhelmingly cytoplasmic, suggesting that cytoplasmic proteins might be more naturally immunogenic than cell surface proteins, and perhaps a bias in the screening methods. SEREX [25] and similar methods employ bacterial expression libraries for the identification of tumor antigens, in which glycosyl and other posttranslational modifications would not be available. These are an important component of cell surface protein antigens and epitopes. Cytoplasmic proteins may be more immunogenic as the result of more abundant expression, or the presentation of cryptic epitopes via proteolysis in apoptotic cancer cells. Cytoplasmic proteins comprise approximately 90% of the expressed proteins by mass, and dominate both in diversity and by levels of expression. In addition, membrane proteins are tightly associated with the membrane (or cell ghost) and are therefore not as available to proteolysis and MHC I presentation as are soluble proteins. Cell surface proteins may also be poorly immunogenic for a number of reasons including shedding, glycosylation, and low copy number. Of over 250 breast cancer antigens identified by reactivity with serum antibodies, a small minority are cell surface proteins (Table 1). These include HER-2/neu and MUC1, which are overexpressed on breast tumors [26, 27]. A third is malignin, which is a cancer-specific and common cell surface immunogen in breast cancers [28, 29]. Other examples include amphiphysin [30, 31], tomoregulin [32], integrin B5 [33], tetraspanin [34], and α-2-Macroglobulin receptor-associated protein [34].

Table 1.

Antibodies from breast cancer patient sera. Partial listing of antigens discovered by reactivity with antibodies from breast cancer patient sera

Antigen	Reference	Antigen Ca specifica	Overexpressed in Cab	Abs Ca/Ai specificc	Subcellular locationd
Acetoacetyl-CoA-thiolase	[34]	N	?	Ca	?
Adenylosuccinate lyase	[34]	N	Y [217]	Ca	CYT
Alanyl-t-RNA synthetase	[34]	N	?	Ca	CYT
Aldolase A	[34]	N	?	Ca	NUC/CYTO
α-2-Macroglobulin receptor-associated protein	[34]	N	?	Ca	MEM [218]
Amphiphysin I	[30, 31]	Cancer-CNS	Y	Ca/Ai	MEM
βIV spectrin 140	[113]	N	?	Ca/HN	CYT
βIVΣ I spectrin	[113]	N	?	Ca/HN/Ai	CYT
BRACA2	[33]	N	Y [219]	Ca	NUC [220]
Calmodulin 80 K protein	[33]	N	?	Ca	CYT
Cathepsin L2	[33]	Cancer-testis [221]	Y	Ca	CYT
CD151 (tetraspanin)	[34]	N	Underexoressed [222]	Ca	MEM
CENP-F	[33]	N	Y [223]	Ca	NUC [224]
Centromere protein-F	[32]	N	Y [225]	Ca/Ai	NUC
c-myb	[226]	N	Y	Ca	CYT
c-myc	[204, 227]	N	Y [228]	Ca/Ai	CYT
Cyclin B1	[204]	N	Y	Ca/HN	CYT/NUC [229]
Cytochrome oxidase I	[106]	N	Y	Ca/HN	CYT
DNA pol δ subunit	[32]	N	?	Ca/HN	NUC
Epidermal growth factor receptor kinase substrate	[34]	N	?	Ca	?
Ferritin light subunit	[106]	N	Y [230]	Ca	CYT
Fibulin-1	[32]	N	Y	Ca	ECM
GCDFP-15/gp17	[231]	N	Y	Ca	CYT/SEC
Glutamic acid decarboxylase (GAD)	[107]	N	Y [232]	Ca/Ai	CYT
Growth factor receptor bound 7	[32]	N	?	Ca/HN	CYT
HER-2/neu	[26]	N	Y	Ca	MEM
HSP27	[88, 89]	N	Y	Ca/Ai	CYT
HSP70	[88, 89]	N	Y	Ca/Ai	CYT
HSP90	[88, 89]	N	Y	Ca/Ai	CYT
Human HRY gene	[34]	N	?	Ca	NUC
IGF-II mRNA-binding protein I (IMPI)	[204]	Oncofetal	Y [233]	Ca/Ai/HN	CYT [234]
ING1 tumor suppressor gene	[34]	N	N	Ca	CYT/NUC
Int-2 (FGF-3)	[33]	N	Y	Ca	SEC [235]
Int-6 (eIF-3 p48)	[33]	N	Y[236]	Ca	CYT
Integrin B5	[33]	N	N	Ca	MEM
INT-M1-1 (EST A1940776)	[32]	N	?	Ca/HN	?
INT-MI-2 (EST BE079402)	[32]	N	?	Ca/HN	?
Keratin 10	[34]	N	Y/N	Ca	CYT
Koc (IGF-II) messenger RNA binding protein	[204]	Oncofetal	Y	Ca/Ai/HN	CYT
Lactate dehydrogenase-A	[32]	N	N	Ca	CYT
Lactate dehydrogenase-B	[32]	N	Y	Ca	CYT
Malignin	[28]	?	Y	Ca/HN	MEM/CYT
MMTV	[96]	Y	N	Ca/HN	CYT
MUC1	[27]	N	Y	Ca/HN	MEM
NAD(+) ADP-ribosyltransferase	[34]	N	?	Ca	?
Nova	[115]	Cancer-CNS	Y	Ca/Ai/HN	CYT
Nuclear receptor corepressor	[34]	N	?	Ca	NUC
NY-BR-1	[103]	Cancer-testis	Y	Ca	NUC
NY-ESO-1	[34]	Cancer-testis	Y	Ca	CYT
p150 spir	[33]	?	?	Ca	CYT [237]
p53	[83]	N	Y	Ca/Ai/HN	CYT
P62 (IGF-II) messenger RNA binding protein	[204]	oncofetal	Y	Ca/Ai/HN	CYT
Pentraxin (NP1)	[33]	Cancer-CNS	Y	Ca	?
Poly(ADP-ribose) polymerase	[34]	N	?	Ca	CYT
Ran GTPase activating protein	[106]	N	?	Ca/HN	NUC
Replication protein A (RPA)	[104]	N	Y	Ca	NUC/CYT
Retinoblastoma binding protein	[238]	Cancer-testis	Y	Ca	NUC
Retinoblastoma binding protein 6	[34]	N	?	Ca	NUC [239]
Rho-associated coiled-coil forming protein	[34]	N	?	Ca	CYT
Ribosomal protein L35A	[106]	N	?	Ca	CYT
RS/DJ-1	[240]	N	N	Ca	NUC/CYT/SEC
S3 ribosomal protein	[33]	N	?	Ca	CYT
Sialyl-Tn	[75]	N	Y	Ca/HN	MEM
sp100	[106]	N	?	Ca	NUC
SSX2	[34]	Cancer-testis	Y	Ca	NUC [241]
Supervillin	[32]	N	?	Ca/HN	CYT
Survivin	[204]	N	Y [234]	Ca/HN	CYT/NUC
Thyroid homone–binding protein	[32]	Cancer-testis	Y	Ca	ER
Tomoregulin (Tr)	[32]	N	?	Ca/HN	MEM
Transcription termination factor I–interacting peptide 21	[34]	N	?	Ca	NUC
U1snRNP	[34]	N	?	Ca/Ai	NUC
v-fos transformation effector gene	[32]	N	Y [242]	Ca/HN	CYT
Yo (CRD2)	[114]	Cancer-CNS	Y	?	CYT
Zinc-finger helicase	[34]	Brain, cardiac, fetal	?	Ca	NUC

^aAg Ca specific: Y yes, N nonspecific (restricted expression if known indicated), ? no available data

^bOverexpressed in Ca: Y yes, N no, ? no available data

^cAbs Ca/Ai specific: ca antibodies present in cancer, Ai antibodies present in autoimmunity, HN antibodies present in healthy normals. Data provided where known; lack of Ca, Ai, or HN does not necessarily indicate that the absence of antibody in that population has been documented

^dSubcellular location: CYT cytoplasmic, MEM membrane, NUC nucleus, ? no available data

Serum antibodies to cell surface tumor–associated antigens

HER-2/neu, the human epidermal growth factor receptor 2 (HER-2, c-erbB-2) is overexpressed on 25–50% of breast tumors as the result of gene amplification, and is correlated with poor prognosis [35]. While HER-2/neu is expressed in normal adult tissue, including cardiac tissue [36], expression levels are low in normal tissues in comparison to tumor. HER-2/neu is a transmembrane protein that contains a glycosylated extracellular domain, and an intracellular tyrosine kinase domain [37]. HER-2/neu heterodimerizes with other members of the EGFR family to form functional receptors with high affinity for EGF-like ligands including the neuregulins, although a direct ligand for HER-2/neu has not yet been identified. Because stimulation through the HER-2/neu receptor induces cell growth, HER-2/neu overexpression is involved in the pathogenesis of malignancy, and is transforming when transfected into cells in vitro [38].

Both antibody and cell-mediated responses against HER-2/neu have been documented in breast cancer patients [13, 26, 39, 40]. In a subset of patients with early stage HER-2/neu-positive breast cancer, 20% produce anti–HER-2/neu serum antibodies [12]. Further, HER-2/neu–specific B cells occur in the peripheral blood of some patients for whom a serum anti–HER-2/neu response is not detectable [41]. However, only 7% of patients with stage III–IV HER-2/neu-positive breast cancer have detectable serum IgG antibodies against HER-2/neu [12]. Advanced stage breast cancer patients have normal cellular responses to recall antigens such as tetanus, but still do not generate effective responses against tumor antigens [7]. Normal humoral responses to non-tumor-associated antigens have also been demonstrated by vaccination of breast cancer patients against influenza virus [8]. In contrast, while 58% and 75% of breast cancer patients with stage I or locally advanced disease, respectively, have serum antibodies that are reactive with autologous tumor, only 17% of patients with metastatic disease have tumor-reactive serum antibodies [5]. These studies suggest that a decrease in antibody response against tumor-associated antigens may be a result of tumor/autoantigen-specific tolerance mechanisms rather than generalized disease-related immune dysfunction.

The Disis group found that antibodies against HER-2/neu are highly cancer-specific. Of 200 healthy individuals examined, none had detectable anti–HER-2/neu antibodies [12]. However, another group reported that anti–HER-2/neu antibodies occur in the sera of healthy control individuals with equal frequency [42]. It is unclear what the disparity in these studies might be.

Breast cancer patient antibody titers against HER-2/neu are in many cases "substantial," with titers of >1:5,000 [13], suggesting that HER-2/neu is quite immunogenic in vivo. Antibodies can even be detected in the sera of patients with tumors that have been judged HER-2/neu-negative by immunohistochemistry [13]. As recent studies of HER-2/neu gene amplification by FISH detection have demonstrated, many patients have genetically overamplified HER-2/neu and benefit from antibody immunotherapy, yet are judged HER-2/neu-negative by immunohistochemistry [43]. This suggests that some individuals may have genetically amplified HER-2/neu, and express low levels of HER-2/neu protein that are nevertheless sufficient to elicit an antibody response and be susceptible to antibody therapy.

The extracellular domain of the HER-2/neu receptor is proteolytically shed, and can be detected in the sera of 15% of breast cancer patients, but not healthy normal donors [44]. Levels of circulating HER-2/neu extracellular domain correlate with both tumor burden and anti–HER-2/neu immune response [12, 41, 44]. Although the specific target epitopes have not been determined, at least some serum antibodies and antibodies from peripheral B lymphocytes bind the extracellular portion of HER-2/neu [12, 44]. Furthermore, antibodies from peripheral B lymphocytes bound the cell surface of HER-2/neu-positive cultured cell lines, indicating recognition of naturally occurring epitopes [26]. The antibody response against this shed protein challenges some theories of immunogenicity, as the presence of soluble antigen can delete reactive B cells in the germinal center reaction [45, 46]. It is possible that systemic levels of HER-2/neu extracellular domain are insufficient to delete reactive B cells. Naturally occurring serum antibodies are not restricted to extracellular domain epitopes. One patient with a strong serum antibody response had antibodies against the intracellular domain [40]. Antibodies against HER-2/neu were of both the IgG and IgM isotypes [13, 26, 41].

In spite of the fact that HER-2/neu is expressed in tissues other than breast cancer [36], anti–HER-2/neu antibody immunotherapy with trastuzumab has produced clinical responses with minimal toxicity (reviewed in [1]). Elimination of antibody-dependent cellular cytotoxicity (ADCC) through knockout of the FcγRIII receptor reduced antitumor activity in murine models approximately 80% (G. Fyfe, Genentech, AACR 2002, San Francisco, CA), supporting a leading role for ADCC. In mice as compared to humans, the dosage of the humanized anti-Her-2/neu antibody had to be increased fourfold in order to achieve equivalent antitumor effect [47], presumably because humanized antibodies do not mediate ADCC efficiently in mice. Trastuzumab also mediates tumor cell killing by complement fixation in murine models [48]. Direct mechanisms of tumor control by trastuzumab include HER-2/neu down-regulation, inhibition of HER-2 heterodimerization, induction of G1 arrest, and inhibition of VEGF (reviewed in [49]).

There is no published evidence that naturally occurring antibody responses against HER-2/neu are protective. This is likely because naturally occurring levels of anti–HER-2/neu antibody are subtherapeutic. In a study of breast and ovarian cancer patients, the maximum naturally occurring level of anti–HER-2/neu antibody was 8.93 μg/ml, although more typical levels were 0–3.4 μg/ml [13]. In contrast, phase I and phase II trials of trastuzumab determined that mean serum levels were 116±62.0 μg/ml (reviewed in [1]). Most current trastuzumab regimens employ a dosing schedule of an initial 4-mg/kg dose, followed by 2-mg/kg weekly maintenance. After 20 weeks, this dosage schedule results in a mean trough serum concentration of ~190 μg/ml. Although efficacy at low doses has not been determined in humans, mean trough concentrations of trastuzumab at 7 weeks were higher in complete (70.3 μg/ml) and partial (58.4 μg/ml) responders than in nonresponders (44.3 μg/ml) in one study [50]. While phase I studies are not designed to study clinical response, a phase I dose escalation study of a second anti–HER-2/neu antibody, MKC-454, failed to produce objective responses below 4-mg/kg dosing [51]. In a phase I trial of trastuzumab + IL-2, doses escalated from 1 mg/kg every 2 weeks to 8 mg/kg weekly. Four major responses were observed, all in breast cancer patients treated with trastuzumab doses of at least 4.0 mg/kg [52]. Murine models have also established unequivocally that trastuzumab is dose-dependent in vivo. Nude mice injected with HER-2/neu–overexpressing ovarian tumor cells were given varying doses (25–150 mg/kg) of 4D5, the murine parent antibody of trastuzumab, in a single injection [47]. Tumor growth was reduced ~80% at 75–150 mg/kg, 60% at 25 mg/kg, and 40% at 5 mg/kg. Imaging studies suggest that higher sustained doses result in better tumor distribution of antibody [53]. Inhibition of tumor cell growth by 4D5 was also dose-dependent in vitro, with no inhibition of HER-2/neu–transfected C13 cells at <15 μg/ml [47]. In summary, spontaneous antibody responses against this naturally immunogenic protein are subtherapeutic. Although HER-2/neu vaccine trials have been pursued [54], no data regarding induced antibody titers has been published.

Antibody responses against tumor antigens may not always be helpful. While some anti–-HER-2/neu antibodies promote differentiation of breast epithelium, induce tumor cell apoptosis, or inhibit tumor cell growth, some murine monoclonal antibodies against the HER-2/neu ectodomain stimulate tumor cell growth in vitro (reviewed in [55]), suggesting that some antibody responses may also be stimulatory of tumor growth in vivo. In addition, antibodies against shed tumor cell surface antigens can promote tumor invasion and metastasis through FcR-induced release of angiogenic cytokines in the tumor microenvironment [56].

Another immunogenic breast cancer antigen is MUC1 (polymorphic, epithelial mucin, PEM), which is immunogenic in the underglycosylated form occurring on tumors [14]. MUC1 is a high molecular weight transmembrane protein with a central domain that consists of heavily O-glycosylated core tandem 20-amino acid peptide repeats [57, 58]. While MUC1 is overexpressed on a variety of epithelial tumors, MUC1 is also expressed by most glandular epithelial cells and many hematologic cell types (reviewed in [59]). MUC1 is involved in cellular adhesion and immune regulation, and may contribute to immune evasion [60, 61, 62]. Underglycosylated MUC1 is overexpressed on the cell surface of a variety of epithelial malignancies, and is shed into the serum. This is the basis of the CA 15.3 clinical diagnostic test [63]. On breast tumors, MUC1 glycans are truncated and sialylated [64, 65], exposing the underlying protein epitopes. While this aberrant posttranslational processing contributes to MUC1 immunogenicity by increasing both MHC class I and class II presentation [66], the MUC1 epitopes recognized by serum antibodies are diverse, and not limited to the underglycosylated core peptide motif. In a study by von Mensdorff-Pouilly et al. [15], naturally occurring serum antibody responses to MUC1 were studied, and the epitopes against which responses occur mapped. Using a combination of serially mutated naked peptides, a common minimal epitope (RPAPGS) was mapped within the core peptide repeat of MUC1. Naturally occurring anti-MUC1 antibodies from breast cancer patients reacted more strongly with GalNAc-modified MUC1 peptides than naked MUC1 peptide, indicating epitopes associated with glycosyl side chains. IgG responses were as common as IgM, and mapped to similar epitopes. While equal numbers of polyclonal IgG and IgM sera were observed, IgG epitopes could be mapped with finer precision using whole sera, while IgM had broader epitopes. IgG were predominantly of the IgG2 isotype, which is the dominant isotype induced by vaccination with polysaccharide antigens (reviewed in [67]). A small number of IgG1 and IgG3 responses, the more common isotypes in anti-protein immunity, were also observed. In summary, this study demonstrated that MUC1 glycosyl moieties are an integral component of the humoral epitope rather than the underglycosylated core peptide.

There is some evidence that naturally occurring humoral responses to MUC1 can be protective in breast cancer. An antibody response against MUC1, as manifested by the presence of MUC1-containing circulating immune complexes (PEM.CIC) [14], was inversely correlated with the extent of disease. While 32% of patients with node-negative local disease were PEM.CIC-positive, only 20% of node-positive patients had PEM.CIC, and no patients with distant metastases had PEM.CIC. The naturally occurring antibody response was also of higher titer in patients with localized disease. The natural humoral immune response also correlated with a slightly improved disease-specific survival, and a decrease in distant recurrence [68]. Anti-MUC1 serum antibodies also occur in both lactating and nonpregnant healthy women [69, 70, 71]. Conversely, elevated levels of circulating MUC1 are associated with pregnancy, in which individuals, anti-MUC1 antibodies are less frequent and of lower titer. The presence of MUC1-containing circulating immune complexes (PEM.CIC) [14, 68] correlated inversely with the extent of disease in breast cancer, and by implication the concentrations of shed and circulating MUC1, as levels of circulating MUC1 correlate positively with extent of disease in MUC1-positive breast cancer [72]. The inverse relationship between MUC1 levels and antibody titers may be a result of deletion of autoreactive B cells by the presence of soluble antigen [45, 46]. It is also possible that higher levels of circulating MUC1 are associated with lower antibody responses due to the immunoinhibitory effects of MUC1.

In one study of MUC1 vaccination of breast cancer patients, initial naturally occurring anti-MUC1 antibodies had a median value of 0.8 μg/ml with a range of 0.6–6.5 μg/ml, and reached a transient peak of only 18 μg/ml after injection with a MUC1-KLH vaccine [15]. The relatively low antibody titers achieved and the transient nature of the response suggest an active suppression of immunity against this autoantigen. Perhaps as a result, clinical trials targeting MUC1 in breast cancer have thus far produced essentially negative results [73].

Sialyl-Tn (STn) is one mucin-associated carbohydrate antigen resulting from the abnormal glycosylation of MUC1 [74]. Naturally occurring serum antibodies against sialyl-Tn were reported in breast cancer patients [75]. However, others reported that anti–sialyl-Tn antibodies are lower in breast and colorectal cancer patients than in healthy controls [76]. Breast cancer patients with poorly differentiated breast cancer had lower serologic antibody titers compared with those with other histologic grades. The Theratope vaccine (Biomira) consists of the mucin sialyl-Tn conjugated to keyhole limpet hemocyanin. In phase II trails of Theratope, patients with high antibody titers against Theratope showed a trend toward prolonged survival [74, 77, 78]. Both cell-mediated and humoral responses against Theratope were observed. Breast cancer patients had a median anti-OSM IgG titer (a measure of anti-mucin STn) of 1:40 after vaccination. However, patients treated with high-dose chemotherapy, autologous stem cell rescue, and Theratope had a serum antibody titer of 1:160 [79]. This was attributed to decreased tumor burden.

Serum antibodies against malignin have been found in a wide range of malignancies, including breast cancer [28, 29]. Malignin is a 10-kDa polypeptide occurring in both the cytosol and on the cell surface of many cancer cell types [28]. Unfortunately, there is little information on its possible function or tissue-specific expression. The sequence of malignin has not been published in Genbank. One study reported serum IgM antibodies against malignin in 97% (31/32) of breast cancer patients [80]. In contrast, only 2% (3/154) of healthy controls tested positive. Of these, two patients were subsequently diagnosed with nonbreast malignancies. A subsequent double-blinded study of antimalignin antibody testing in over 300 patients demonstrated sensitivity and specificity for detecting tumors including breast cancer was 95% (>99% upon repeat test) [81]. Thus, testing for antimalignin antibodies has been suggested as a tool for the detection of breast and other cancers. The presence of antibodies against malignin correlate with improved survival [28].

Serum antibodies to cytoplasmic tumor-associated antigens

A large number of cytoplasmic proteins elicit serum antibody responses, including antigens that are nearly restricted to tumor in expression and those that are expressed ubiquitously. (We use the term cytoplasmic to refer to all proteins that are not cell surface membrane proteins.) p53 is a well-characterized cytoplasmic breast cancer antigen to which there is a native humoral response [82]. Anti-p53 antibody responses have been of great interest because of the role of p53 as a tumor suppressor. Approximately 15% of breast cancer patients have circulating anti-p53 antibodies [83], with no correlation with disease stage. The presence of serum antibodies to p53 in breast cancer patients is correlated with a poor prognosis, possibly because p53 is overexpressed primarily on high-grade tumors [84]. In one study, 40% of breast cancer patients overexpressed p53, and 9% had serum antibodies against p53 [85]. Curiously, 40% of patients with an anti-p53 antibody response did not have p53 accumulation in tumor, demonstrating that p53 antibody responses apparently do not absolutely require p53 protein overexpression. Consistent with this, serum anti-p53 antibodies were detected 5% (1/20) of healthy donors in a study that detected anti-p53 antibodies in 10% (2/20) of breast cancer patients [32]. Although p53 is frequently mutated in breast and other tumors, humoral immunity to p53 is not directed against mutated epitopes [84], and mutation of the p53 gene is not necessary for the generation of an antibody response. In fact, p53 gene mutations were no more frequent in antibody-positive than antibody-negative patients [85]. Serum antibodies recognize both mutant and wild-type p53, and are rarely reactive with epitopes in the mutational hot spot regions [84]. Immunodominant epitopes are located at both the amino and carboxy termini of the p53 protein [86].

As we have suggested for other tumor-associated proteins, the source of immunogenicity of p53 may lie in the tumor rather than in some inherent property of the p53 protein itself. It has been suggested that the enhanced immunogenicity of tumor p53 may be due to heat shock protein binding. All p53-positive breast tumors that elicited a serologic anti-p53 antibody response (7/7) contained p53–70-kDa heat shock protein (HSP) complexes [86]. In contrast, tumors that did not elicit an antibody response did not contain these complexes (0/8) despite having mutated copies of p53. This suggests that HSPs serve as an adjuvant for p53. The native immunogenicity of HSPs has been demonstrated in many studies, and antitumor vaccination strategies have been proposed that utilize HSPs as adjuvant (reviewed in [87]).

Antibody responses against heat shock proteins have also been described in breast cancer patients [88, 89]. Antibodies against the 27-kDa heat shock protein (HSP27) are associated with improved survival [89], while antibodies against HSP90 correlate with poor survival [88]. Because heat shock proteins are completely cytosolic, an antibody response is unlikely to directly provide protective immunity. Antibodies against HSP90 and HSP27 are detectable only in breast cancer patients, while antibodies to HSP70 are detectable in approximately one third of both healthy controls and breast cancer patients [89].

Mouse mammary tumor virus (MMTV)–like sequences have been detected in breast tumors but not healthy breast tissue [90, 91, 92], suggesting that some breast tumors may have a viral origin. Antibodies against MMTV proteins have been found in the sera of breast cancer patients [93, 94, 95, 96, 97, 98, 99]. Anti-MMTV virolytic activity was also found in the sera of breast cancer patients [100]. However, reports of MMTV in human breast tumors remain controversial. A study of 111 breast cancer patients and 122 healthy age-matched controls failed to detect higher levels of sera reactive with most MMTV proteins in breast cancer patients [101], although more breast cancer patient sera were reactive with the p18 MMTV protein. In another large study, the sera of 300 donors were tested for reactivity with a panel of MMTV proteins [102]. While the sera of breast cancer patients had some reactivity with the p27 and gp52 MMTV proteins, the authors concluded that there was "no significant correlation between specific antibodies to mouse mammary tumor virus and human cancer."

In 1995, Sahin et al. first described SEREX technology, in which recombinant tumor cDNA expression libraries were screened with patient sera [25]. This allows the simultaneous identification of many tumor-associated antigens against which a serum antibody response exists. SEREX has resulted in the discovery of over 250 breast cancer antigens (listed in the Ludwig Institute for Cancer Research SEREX database at http://www.licr.org/05_pro/serex.htm). A partial listing of the antigens discovered in breast cancer by SEREX technology can be found in Table 1. Only those antigens about which some protein information exists (rather than uncharacterized ESTs), and for which information has been published in the peer-reviewed literature, are listed. Proteins discovered by SEREX are overwhelmingly cytoplasmic. SEREX is biased for the selection for cytosolic proteins through screening of bacterial libraries. Because many or even most membrane proteins are posttranslationally modified, these epitopes would not be available in a bacterial library. Further, many membrane proteins are not properly expressd in bacterial libraries.

In a series of serologic analyses of breast cancer patients, a primary breast tumor cDNA expression library was screened with both autologous and allogeneic breast cancer patient sera [103]. Thirty genes were identified, including the "cancer-testis" genes NY-ESO-1 and SSX2. (The term cancer-testis refers a group of genes whose expression is relatively restricted to tumor and testis.) Although other cancer-testis genes including MAGE-1, MAGE-3, and MAGE-4 were expressed by the primary breast tumor and were present in the cDNA expression library, serum antibodies to these proteins were not detected. Screening of the same library with allogeneic breast cancer serum also identified the NY-ESO-1 gene. Reactivity of healthy donor sera with NY-ESO-1 was not determined. Other cancer-testis proteins have also been identified by SEREX. The putative DNA-binding protein NY-BR-1 was identified by SEREX utilizing breast cancer patient sera and an autologous tumor cDNA expression library [103]. Extensive testing of mRNA expression of this previously unidentified protein determined expression in breast tumor, breast, and testis. Sequence analysis of NY-BR-1 identified no mutations in comparison to the Genbank genomic sequence. Antibodies against thyroid hormone–binding protein (THBP) were also discovered by SEREX in 45% of breast cancer patients and 10% of normal donors. THBP is expressed in testis, and at very low levels in other tissues [32].

Screening of a HeLa cDNA expression library with a breast cancer patient sera yielded one reactive clone, identified as replication protein A (RPA) [104]. Over 10% (87/801) of breast cancer patients but no normal donors had serum antibodies against RPA. Curiously, anti-RPA antibodies also occur in SLE patients [105], suggesting that this protein may have some natural immunogenicity in addition to immunogenicity conferred by tumor expression. While RPA is overexpressed in autologous breast tumor cells in comparison to adjacent healthy breast, RPA is not tumor-specific. The relative levels of RPA in breast tumor and other replicating tissues were not determined. Other ubiquitous proteins identified by SEREX include poly(ADP-ribose) polymerase, keratin 10 [103], lactate dehydrogenase–A and lactate dehydrogenase–B [32], cytochrome oxidase I, sp100, Ran GTPase-activating protein [106], DNA pol δ subunit [32], and adenylosuccinate lyase [103].

Serum antibody responses and immunogenicity

Common properties emerge from the study of naturally occurring antibody responses, giving clues to the immunogenicity of some tumor-associated antigens. Restricted expression such as with the cancer-testis antigens may increase the immunogenicity of specific proteins. Antigens such as HER-2/neu (cardiac) and amphiphysin (central nervous system) p62 (oncofetal) are also restricted in expression. The same concept would apply to antigens that are expressed only at low levels in the normal tissue of origin of the primary malignancy, but which are overexpressed on tumor. Lymphocytes and dendritic cells in the locoregional lymph nodes may not be tolerized to these antigens, possibly allowing the initiation of a humoral response. While naturally occurring antibodies do not appear to be directed at mutated epitopes, mutation may lead to increased protein stability as with the p53 protein, in effect creating overexpression. Aberrant processing may also increase the immunogenicity of a tumor-associated antigen, as is the case with MUC1. Thus, there are plausible explanations for the immunogenicity of some individual tumor proteins, but the extreme diversity of immunogenic proteins and antibody responses against ubiquitously expressed proteins suggests that the basis of immunogenicity for most proteins lies in the tumor rather than in the properties of specific proteins. For example, p53 may be immunogenic as the result of association with heat shock protein [86].

Antitumor antibodies and autoimmunity

Antibody responses against tumor-associated autoantigens can result in harmful autoimmunity. For example, stiff-person syndrome manifests itself in some breast cancer patients. Stiff-person syndrome, a disease of the central nervous system, is associated with autoantibodies against glutamic acid decarboxylase (GAD) in a high percentage of non–breast cancer patients with the disease. While there is a report of one breast cancer patient who presented with stiff-person syndrome and anti-GAD antibodies [107], autoantibodies against amphiphysin I are more commonly reported in association with the syndrome in breast cancer [30, 31]. Amphiphysin I is a nonintrinsic membrane protein present in the presynaptic compartment of neurons, and plays a role in synaptic vessel recycling [108]. Amphiphysin is also expressed in a number of breast cancer cell lines, at low levels in normal tissues, and primarily in the brain [109]. The amphiphysin protein contains a COOH-terminal Src homology 3 (SH3) domain that interacts with dynamin 1 and synaptojanin 1 [110, 111]. Epitope mapping of naturally occurring antibodies against amphiphysin in a breast cancer patient with stiff-person syndrome determined that the carboxy-terminal cytosolic region contains an immunodominant epitope [112]. Other "cancer-CNS" proteins are also the target of antibody responses in paraneoplastic neurologic syndromes in breast cancer. In the case of paraneoplastic lower motor neuron syndrome, anti-axonal antibodies against βIVΣ I spectrin and ΒIV spectrin-140 were discovered [113]. Following removal of the primary breast tumor, autoantibody titers were markedly reduced in this patient, consistent with a tumor-associated autoimmune etiology. In another example, the 52-kDa Yo antigen (CRD2) is expressed on Purkinje cells and on breast tumors [114]. Antibodies against Yo in breast cancer patients are associated with autoimmune cerebellar degeneration. Paraneoplastic opsoclonus-myoclonus ataxia has also been observed in association with breast cancer, and is associated with antibodies against Nova-1, a protein expressed in breast tumors and CNS neurons [115]. Glutamic acid decarboxylase, amphiphysin, Yo antigen, Nova-1, βIVΣ I spectrin, and βIV spectrin-140 have the common property of expression almost exclusively in the central nervous system (CNS). The CNS is an immune-privileged site, by virtue of both physical barriers and active regulation in the CNS microenvironment [116]. It is speculated that "a state of immunological ignorance exists" with regard to CNS-specific antigens, allowing the development of potent immunity against cancer-CNS antigens [114]. Individual case reports indicate that breast cancer has also presented in association with a number of other autoimmune syndromes including immune thrombocytopenic purpura [117], palmar fascitis [118], arthritis [118], and scleroderma [119]. The link between antitumor immunity and neurologic disease sounds a cautionary note for the design of anticancer vaccines. Thus far, neurologic autoimmune syndromes have not been observed in vaccine trials.

Tumor-reactive B cells from lymph node

There is a long history of study of lymph nodes in breast cancer. In 1950, a report from Nadel and Ackerman described "granulomatous sarcoid-like changes in the regional lymph nodes" of cancer cases [120]. In 1953, Black et al. described "marked sinusoidal and follicular histiocytic transformation" of lymph nodes in patients with breast cancer [121]. The Black report implied a general knowledge that locoregional lymph nodes from breast tumors are often enlarged, even in the absence of metastatic tumor cells, suggesting a native immune response against tumor.

Sinus histiocytosis

Black et al. first described a correlation between sinus histiocytosis and survival [121]. Sinus histiocytosis referred to filling of the nodes with large finely granular histiocytes, often to the exclusion of secondary follicles and plasma cells. Histiocytes were described as a set of cells resembling those observed in the delayed-type hypersensitivity (DTH) skin reaction [122], now defined as monocyte-derived tissue macrophages. It was later confirmed that sinus histiocytosis was associated with improved survival [123, 124, 125, 126]. In a series of 405 cases of breast cancer, 5-year survival was 76% when both sinus histiocytosis and tumor-infiltrating lymphocytes were present, versus only 49% when neither factor was present [127]. Furthermore, sinus histiocytosis was associated with fewer metastases [128], and was more often observed in node-negative patients [126]. Sinus histiocytosis also correlated negatively with the presence of bilateral disease [129]. Although sinus histiocytosis was most frequently observed in older patients with less aggressive tumors [130], the prognostic significance of sinus histiocytosis was greatest in the most aggressive cases [127, 131].

It is unclear what the mechanism(s) responsible for this improvement in survival might be. Tumor-draining and axillary lymph nodes revealed expansion of lymph node cellular populations in distinct patterns, including germinal center predominant and sinus histiocytosis [132]. Sinus histiocytosis, if limited to instances of proliferation of "histiocytes" might best describe a T_h1-type response, while a germinal center predominant histology might indicate a T_h2-type response. When specific patterns of cellular proliferation in the axillary nodes were compared, a lymphocyte-dominant pattern was associated with best survival, although a germinal center–dominant pattern was also associated with improved survival in comparison to unstimulated or lymphocyte-depleted nodes [132]. It was also suggested that a germinal center–dominant histology was associated with decreased survival [125], suggesting that B-cell proliferation in tumor-draining nodes does not constitute an effective antitumor response.

In comparison to nodes from healthy controls, the tumor-draining lymph nodes of breast cancer patients contained high numbers of IgG⁺ B cells, indicating an affinity-matured response [133]. Overall, higher total numbers of B cells were observed in the lymph nodes of stage II breast cancer patients in comparison to stage I patients [134], and elevated numbers of B cells were representative of greater numbers of germinal centers [135]. Nodes with IgG⁺ follicles were associated with high-grade tumors and the presence of lymph node metastases [130], and high numbers of node plasma cells correlated with weight of primary tumor [136]. From this, it might be inferred that more aggressive tumors induce an affinity-matured (IgG) response, which may or may not be protective.

Nodal B-cell response to cell surface antigens

Many node-derived antibodies are reactive with glycoproteins (Table 2). For example, we developed an antibody from a breast tumor–draining lymph node that bound a tumor-associated glycoprotein epitope [137]. Lymphocytes from a tumor-draining lymph node were immortalized via fusion with the mouse x human heteromyeloma line (SPAZ-4). Monoclonal antibody (mAb) 15.2.3 is an IgM/κ isotype antibody that had reactivity with breast, colorectal, stomach, pancreatic, and hepatoma cancer cell lines, but not with normal diploid fibroblasts, a lactating adenoma, or noncarcinoma malignancies. Both cytoplasmic and cell membrane–binding of mAb 15.2.3 was observed. On Western analysis of proteins from a variety of carcinoma cell lines with mAb 15.2.3, it was discovered that the antibody bound four individual protein bands that were differentially represented in the tumor cell lines. Complete deglycosylation of reactive cells increased mAb 15.2.3, indicating demasking of the reactive epitope(s). The epitope may be present on multiple glycoproteins, or alternatively, the glycoprotein may exist in at least four distinct posttranslationally modified forms that can be distinguished by molecular weight. Administration of the mAb 15.2.3 to tumor-bearing mice decreased tumor size and weight, suggesting that this antibody might have antitumor activity. The percentage of tumor cells expressing the mAb 15.2.3 reactive protein could be increased by treatment with IFN-γ or IFN-α. Up-regulating tumor antigen through application of cytokines in vivo or ex vivo has been employed in a number of clinical trials (for example, see [138]).

Table 2.

Antibodies from locoregional lymph nodes

Antibody or hybridoma	Specificity	Isotype	Antigen	Subcellular location	Reference
F86	Breast tumor cell lines, MM cell lines	IgG3	Unknown	Cell surface, cytoplasm	[153]
B2, B6, B8, B10, B11	MMTV gp52, gp36, p28, p21, p14, not SSV, AMV	IgG-κ	MMTV proteins	Unknown	[243]
BMMK-33G	Breast, pancreatic and gastric cancers, low binding of normal tissue counterparts.	IgM	Sulfatide	Cell surface	[152]
H15A3	MCF7	IgM	40-kDa antigen	Cell surface	[244]
YBB190	Normal and malignant breast epithelium	IgM	Cytokeratin or associated protein	Cytoplasmic	[245]
15.2.3	Breast, colorectal, stomach, pancreatic, and hepatoma cancer cell lines, not normal diploid fibroblasts, lactating adenoma, non-carcinoma malignancies	IgM/kappa	Multiple glycoproteins	Cytoplasmic and cell membrane	[137]
Six hybridomas	Not determined	IgM and IgG	MUC160-mer peptide repeat	Cell surface	[139]
MBE6	Primary breast tumors, benign breast lesions all metastatic breast lesions. Light staining of normal or lactating breast tissues	IgM	Unknown	Cytoplasmic	[147]
Ri37	Some PBMC, melanoma, not cancer cell lines from breast, ovary, pancreas, brain, or kidney	IgG	Heat, protease stable antigen	Cell surface	[145]
Gr169	Epithelial and neuroectodermal tumors, but not normal epithelium or other normal tissues	IgM	Unknown	Cell surface	[144]
Sp909	All cell types tested	IgM	Unknown	Cell surface	[144]
Kr73	Most malignant and healthy tissues	IgM	Unknown	Cell surface	[144]
CA27(25)	Breast, colon, kidney, stomach tumor, lung, normal kidney and sebaceous gland epithelium, not erythrocytes, lymphocytes, vascular endothelium, stroma	IgG	Unknown	Cytoplasmic	[151]
JD39(22)	Breast, colon, kidney, stomach tumor, normal kidney and sebaceous gland epithelium, not erythrocytes, lymphocytes, vascular endothelium, stroma	IgG	Unknown	Cytoplasmic	[151]
CF29(34)	Breast, colon, kidney, lung, stomach tumor, normal kidney and sebaceous gland epithelium, not erythrocytes, lymphocytes, vascular endothelium, stroma	IgG	Unknown	Cytoplasmic	[151]
HMA-29	Breast carcinomas, normal breast, colon and kidney tissue, not lactating breast	IgG1	29-kDa phosphoprotein	Cytoplasmic	[150]
HMA-31	Breast carcinomas, normal breast, colon and kidney tissue not lactating breast	IgG2	31- and 34-kDa proteins	Cytoplasmic	[150]
A4–33	Breast, pancreatic and colon cancer cells, normal mammary gland, secretory epithelial cells, not cervical cancer, gastric cancer, bladder cancer, normal fibroblasts	IgM	Carbohydrate antigen	Cytoplasmic	[140]
MAC40/30	Malignant and normal epithelium	IgM	47-kDa glycoprotein	Cytoplasmic	[141]

Antibodies reactive with a peptide moiety of the breast tumor glycoprotein MUC1 were isolated from tumor-draining lymph node B cells [139]. In this report, B cells from tumor-draining nodes were sorted on the basis of binding to a 60-mer peptide representing three tandem repeats of the MUC1 core protein. Both IgG and IgM anti-MUC1 B cells were from locoregional lymph nodes from five out of six patients who had serum IgM against MUC1. Because of the known improved survival of breast cancer patients with anti-MUC1 antibodies [68], it was suggested that removal of nodes in individuals without lymph node metastases might impair antitumor immunity. While this may be true, it is impossible to predict who these node-negative individuals might be prior to surgery. The trend in current surgical technique is toward removal of the sentinel nodes only, perhaps preserving immune capacity.

Other groups have also cloned antiglycoprotein antibodies from the axillary nodes. For example, the antigen reactive with IgM antibody A4-33 was periodate sensitive, suggesting a carbohydrate antigen. A4-33 was reactive with cultured breast, pancreatic, and colon cancer cells, but those of not cervical cancer, gastric cancer, bladder cancer, or normal fibroblasts [140]. However, A4-33 was also strongly reactive with normal mammary gland and other normal secretory epithelial cells from sweat gland, pancreas, and kidney. In another example, five hybridoma clones derived from axillary node lymphocytes reactive with breast tumor cell line MCF7 were identified [141]. All were IgG or mixed IgG/IgM. The antibodies displayed reactivity with cytosolic components of both malignant and normal epithelium, but bound only malignant rather than normal mesenchymal tissue. The antigen recognized by one antibody (MAC40/30) was determined to be a 47-kDa glycoprotein.

Lloyd Old and colleagues have developed a number of clones that are reactive with tumor cell surface. Clone Ri37 was derived from breast cancer draining nodes, and produced an IgG antibody reactive with a subset of PBMC cells, 3/8 lung cancer cell lines, and 2/10 melanoma cell lines, but curiously, not with cancer cell lines derived from breast, ovary, pancreas, brain, or kidney. Ri37 did not bind breast cancer cell line BT-20, which expresses high levels of MUC1 [142], or MDA-361, which expresses high levels of HER-2 [143], eliminating these as possible antigens. The antigen was found to be stable with protease treatment, but heat labile [144]. This group also developed three other cell surface reactive antibodies from axillary lymph nodes in a subsequent study [144]. Gr169 is of the IgM class, and is reactive with both epithelial and neuroectodermal tumors, but not normal epithelium or other normal tissues. The Gr169-reactive antigen is heat stable and not sensitive to protease treatment. Sp909 is an IgM antibody, and is reactive with all cell types tested. Kr73 is an IgM antibody and is reactive with most malignant and healthy tissues.

Nodal B-cell response to cytoplasmic antigens

Most antibodies from tumor-draining nodes are reactive with cytoplasmic rather than cell surface antigens. One study analyzed a total of 235 node-derived clones from both breast and other malignancies, and determined that only one was reactive with tumor cell surface, and 11 with intracellular antigen [145]. Jeff Schlom and colleagues published a series of studies of monoclonal human antibodies derived from breast tumor–draining lymph nodes [146, 147, 148] (reviewed in [149]). One IgM antibody [147] reacted strongly with the cytoplasm of both primary and metastatic allogeneic breast tumors, but not with normal mammary epithelium. This antibody, MBE6, bound 81% (54/67) of primary breast tumors, 100% (20/20) of metastatic breast lesions, and 14% (3/22) benign breast lesions, by immunohistochemistry [146]. Little staining of normal or lactating breast tissues was observed. The MBE6 antibody also bound some but not all lung adenocarcinomas, and one of eight colon adenocarcinomas. While CEA was ruled out as a possible antigen, the reactive antigen for this antibody was not determined.

The dominance of intracellular antigens reactive with node-derived antibodies was observed in two reports by Imam et al. [150, 151], consistent with those observed in other studies. In the first study, a total of 81 hybridomas were produced from axillary lymph node lymphocytes by fusion to the mouse myeloma line M5 [151]. No IgM clones bound breast tumor with sufficient avidity to be detected by immunohistochemistry, while 15 IgG clones bound tumor to varying degrees. In addition to binding tumor, most of the antibodies also bound normal tissue. Two IgG antibodies with relatively high specificity for tumor were identified, CA27(25) and JD39(22). In a subsequent study by this group 14/52 IgG and 51/94 IgM showed some level of tumor binding by immunohistochemistry [150]. Two antibodies with the highest activity with tumor were selected for further characterization. By immunohistochemistry, both antibodies were reactive with a variety of breast carcinomas, and not normal lactating breast, but did bind normal breast, colon, and kidney tissue, suggesting reactivity with an epithelial-specific antigen(s). Despite having similar tissue-binding profiles, the two antibodies bound different antigens: HMA-29, an IgG1 antibody, bound a 29-kDa phosphoprotein, while HMA-31, an IgG2 antibody, which is the dominant isotype induced by vaccination with polysaccharide antigens (reviewed in [67]), bound proteins with molecular weights of 31 and 34 kDa. Although most antigens recognized by node-derived antibodies bind glycoproteins, one group described an antibody that was reactive with the glycolipid sulfatide (galactosylceramid-12-sulfate) [152]. The antibody was reactive with breast, pancreatic, and gastric cancers by immunohistochemistry, and less so with normal tissue counterparts.

While all other node-derived antibodies described thus far failed to bind hematologic malignancies, one antibody, F86, also bound malignant myelomonocytic cell lines, but not normal peripheral blood cells [153]. Breast tumor–reactive antibody F86 was obtained by fusion of EBV-transformed node lymphocytes with the human fusion partner HMMA2.11TG/O. F86 was of the IgG3 isotype, which along with IgG1, is most commonly induced by protein antigens (reviewed in [67]). The antibody bound both the cytosol and cell surface of breast tumor cells by immunohistochemistry.

Immunogenicity of tumor-associated antigens in regional nodes

In an overview of antibodies cloned from breast tumor axillary lymph nodes, some common trends emerge. First, most antibodies from breast tumor axillary node B lymphocytes react with cytosolic rather than cell surface determinants. This is similar to what is observed in serum responses, and may occur for similar reasons; cytosolic proteins are more abundant, more numerous, and probably more immunogenic than highly modified cell surface proteins. Second, of those antibodies that are reactive with cell surface breast tumor antigens, most recognize glycoproteins, as might be expected since most cell surface proteins are glycoproteins. Third, most reported tumor-reactive antibodies from nodes are of the IgM isotype, although studies that produced both IgG and IgM demonstrated the presence of both IgG and IgM clones with tumor reactivity [144, 150, 151]. Fourth, many node-derived antibodies are reactive with antigens whose expression is restricted to a variety of normal secretory epithelia, including normal breast epithelium, often at reduced levels in comparison to breast tumors. The immunogenicity of these antigens may be the result of tumor overexpression. Although a given antigen may be expressed at moderate levels elsewhere in the body, local overexpression may induce a immune response in regional nodes.

Breast tumor–infiltrating B cells (TIL-B)

TIL-B histology

One study by our group found that antitumor antibodies were produced by tumor-infiltrating B cells (TIL-B) in approximately 70% of nonbreast tumors examined, suggesting that this might be a common phenomenon in solid tumors [154]. TIL-B cells were the dominant lymphocyte population in nonmalignant breast lesions, suggesting that a B-cell response may occur early in tumor development [155]. The dominant population of infiltrating lymphocytes in ductal carcinoma in situ (DCIS) lesions was also B cells [156]. TIL-B in DCIS were perivascular and associated with high endothelial venule-like vessels. DCIS TIL-B clustered in aggregates surrounded by T cells. Many breast adenocarcinomas contain lymphocytic infiltrates to varying degrees. In a study of breast cancer, inclusive of all subtypes, heavy infiltrates were observed in ~20%, and moderate infiltrates in ~50% [157]. The composition of breast tumor–infiltrating lymphocytes varied between patients and was heterogeneous, containing both CD4 and CD8 T cells, with fewer numbers of B cells, macrophages, and NK cells [158]. Approximately 20% of invasive breast tumors contain significant numbers of B cells, with B cells comprising up to 60% of the TIL (Coronella-Wood et al., unpublished data, see also [158, 159]). When present, CD20⁺ TIL-B cells occurred exclusively in follicle-like aggregates [19, 160, 161, 162], which is consistent with in situ antigen-driven expansion rather than nonspecific or inflammatory chemoattraction. In most cases, TIL-B aggregates occurred in stromal areas immediately adjoining tumor nests, and were not observed outside the tumor margins [19]. B-cell follicles were surrounded by CD3⁺ T lymphocytes, the majority of which were CD4⁺, although a component of CD8⁺ cells was also present, which is also the situation in both tonsil (lymph node) (reviewed in [67]) and RA germinal centers [163].

All TIL-B aggregates contained interdigitating CD21⁺ follicular dendritic cells (FDCs), allowing identification of the aggregates as authentic ectopic follicles (Fig. 1), previously described only in the context of autoimmune diseases such as rheumatoid arthritis (RA) [164]. CD21⁺ FDCs are specific to primary and secondary follicles. CD21⁺ FDCs were not observed outside of B-cell germinal centers, as was also the case with RA [165], suggesting that FDCs provide the primary signals for proliferation and affinity maturation of germinal center B cells, and may be necessary for ectopic germinal center formation. Because FDCs contribute to affinity maturation, their discovery in breast tumors suggested that an antitumor B-cell response might develop in situ in tumors rather than in lymph nodes alone.

Fig. 1.

Intratumoral follicle, patient 16

Immature CD1a⁺ Langerhans-type dendritic cells (DCs) were also a component of infiltrates, and were found in tumor nests of all breast cancer samples examined [166]. DCs induce differentiation of germinal center B cells into plasma cells through secretion of IL-10 (reviewed in [167]). Tumor nests contained immature CD1a⁺ DCs, while stroma contained mature CD1a^- DCs [166, 168]. In some cases, T cells clustered around mature (CD1a^-) DCs in peritumoral areas, which is characteristic of an ongoing immune reaction [166]. In contrast, normal breast tissue did not contain DCs, and it was suggested that secretion of the DC chemoattractant macrophage inflammatory factor 3α by tumor was responsible for the presence of tumor DCs [166].

Tumor-reactive antibodies from TIL-B

The first report of tumor-reactive breast TIL-B appeared in 1994 [169] (see Table 3). In this intriguing but puzzling study, an EBV-transformed TIL-B line (hairy-BM) produced a tumor-reactive antibody. Although hairy-BM cells were CD20⁺, IgG⁺, sIg kappa⁺, the cell line displayed properties not normally associated with B cells. Hairy-BM cells bound and lysed tumor cells. While the authors attributed this to a unique cytotoxic effect, it seems possible that the hairy-BM antibodies might have induced tumor cell killing, or that the hairy-BM culture was contaminated with cytotoxic cells.

Table 3.

Antibodies from breast TIL-B

Tissue specificity	Antibody	Reference
Actin, apoptotic, typical medullary carcinoma cell surface	Anti-actin	[9, 10]
Breast cancer cell surface	14.6.11	[19]
Breast cancer cell surface	14.6.19	[19]
Breast cancer cell surface	14.6.20	[19]
Breast tumor	Hairy-BM	[169]

To determine if breast TIL-B immunoglobulins were reactive with tumor, our group-cloned phage-displayed Fab (heavy chain variable region plus CH1 and light chain) libraries were cloned from breast TIL-B utilizing the pCOMBX phage display vector [170]. The libraries were panned on cultured breast cancer cells in order to enrich for Fabs that bind breast cancer cell surface antigens [19]. Enrichment for tumor reactivity was observed after only one round of panning in some cases, indicating a high fraction of tumor-binding Fabs in the library [171]. Individual Fab clones were selected from the enriched libraries and assessed for tumor cell surface binding by flow cytometry. A number of Fabs with apparent specificity for breast tumor cells have been isolated [171]. For example, Fab 14.6.11 bound all breast cancer cell lines tested thus far but not nonmalignant healthy breast epithelium, primary fibroblasts, or the leukemia cell line HL60. Fabs 14.6.19 and 14.6.20 were also highly specific for breast cancer, but less universal binding of breast cancer cell lines was observed. The three Fabs contained somatic mutations in the antigen-binding CDR regions, suggesting affinity maturation, and in FACS, generated cell-binding signals equivalent to those produced by monoclonal anti-EPCAM antibodies, suggesting high affinity and high antigen copy number. The Fab clones utilized unique germline immunoglobulin genes, and had nonidentical patterns of cell line–binding, consistent with binding of unique antigens. However, other than elimination of some common breast cancer antigens such as HER-2/neu and MUC1, the reactive antigens have not been identified.

Oligoclonal expansion of TIL-B

To determine if the aggregates of B cells observed in tumors were the result of random recruitment from the periphery or of the proliferation of tumor-infiltrating lymphocytes, IgG heavy-chain libraries were generated by RTPCR, and random IgG sequences were analyzed from a total of six breast tumors, a tumor-draining lymph node, and the peripheral blood of a healthy donor (Coronella-Wood et al., unpublished data, see also [19, 162]). IgG heavy chains were sequenced and analyzed for clonality, or descent from a shared B-cell progenitor cell. Clonality was determined by shared VDJ germline gene usage and junctional mutations, as these events occur in the marrow prior to entry into the circulating peripheral population, and are very specific to individual B cells. B-cell clones occur very infrequently (<1/20,000) in the peripheral repertoire [172, 173, 174, 175], and are not detectable by sampling and sequencing methods such as those employed in this study. Accordingly, no clones were detected in the peripheral blood sample. Lymph nodes contain a more limited repertoire, and clonal sequences were detected for 7% of tumor-draining lymph node sequences. In contrast, between 18% and 68% of IgG heavy-chain sequences from TIL-B belonged to clonal groups, demonstrating oligoclonal expansion of an extremely restricted B-cell repertoire. As occurs in lymph node germinal centers, TIL-B lineages accumulated unique somatic mutations during proliferation, allowing the derivation of genealogical trees and calculation of cell doubling numbers. Based on a somatic hypermutation level of 1 base pair per 10² to 10³ bases per generation for immunoglobulins (reviewed in [67]), between 7 and 70 cell divisions would be required to produce some observed lineages. However, few TIL-B expressed the proliferation marker Ki-67 by immunohistochemistry, indicating indolent or previous proliferation of TIL-B [19].

Affinity maturation of TIL-B

There is some evidence that the somatic hypermutation of TIL-B is part of selective affinity maturation. TIL-B IgG heavy chains contained somatic mutations that clustered in the antigen-contacting CDR regions [19], as previously observed in affinity-matured antibodies (reviewed in [67, 175]), and as was also seen in tumor-draining lymph node but not peripheral blood IgG [19]. As calculated by the polynomial algorithm of Lossos et al. [176], replacement and silent mutations occurred nonrandomly in 36–84% of TIL-B–derived IgG heavy chains. Low levels of TIL-B IgG heavy chain nonsense mutation and a modest bias in germline gene usage also suggested clonal selection. Indirect measurements from flow cytometry experiments also suggested that individual Fabs from TIL-B had high affinity for tumor antigen (Coronella-Wood, unpublished data).

Comparison of ectopic germinal centers in autoimmunity and breast cancer

B cells can proliferate and undergo affinity maturation in ectopic germinal centers, as has previously been described only in autoimmunity. Extranodal B cell proliferation has been described in rheumatoid arthritis [20], multiple sclerosis [21], Sjogren's syndrome [22], and Grave's disease [23], and has led to a pathogenic disease state. In the case of rheumatoid arthritis, "rheumatoid factor" B cells form germinal centers in the synovium, then undergo somatic mutation, affinity maturation by selection [177], receptor revision [178], clonal expansion [179], and differentiation to plasma cells [20]. This suggests that mechanisms to delete autoreactive B cells may be deficient in ectopic germinal centers. Like the germinal centers we have described in breast tumors, synovial germinal centers in RA do not have clearly delineated light and dark zones [19, 180]. Despite this, antibodies produced by germinal centers in both cases are affinity matured. It is thought that this lack of structural organization in RA might be due to insufficient production of CXCL13 by follicular dendritic cells or its receptor on B cells, CXCR5 [180]. An unusual population of CD8⁺, CD40L⁺, IFN⁺, perforin^- T cells were also absolutely required for germinal center formation in RA [163]. These cells were not observed in tonsil germinal centers. While we have observed CD8⁺ T cells in the germinal centers in breast tumors, we have no information regarding expression of these markers.

Tumor-infiltrating lymphocytes and prognosis

Breast tumor–infiltrating lymphocytes have been extensively studied in the context of prognosis. While a conclusion of improved survival with lymphocytic infiltration has been controversial, a review of the literature indicates a clear link between lymphocytic infiltration and prognosis—in some subsets of breast cancer patients. In a definitive study by Menard et al., a multivariate analysis of breast tumor histology in 676 patients found that breast carcinomas can be divided into two phenotypic groups based on a number of factors including expression of HER-2/neu, tumor size, nodal status, lymphocytic infiltration, and patient age. One group had more aggressive histologic features, and involved younger women, who have poorer survival [181]. The presence of infiltrating lymphocytes in these tumors associated strongly with improved survival in both the short and long term. There was also a positive correlation between degree of lymphocytic infiltration and prognosis. This association was strongest in patients under the age of 40 years (p=0.0002) [182]. In patients under the age of 40, who have the most aggressive tumors, there was an average 70% survival after 14 years in those with lymphoid infiltration as compared to 30–40% survival without with lymphoid infiltration. However, there was little or no association between infiltration and improved prognosis in patients aged 40–49, and no association in patients aged 50 or over. A second study came to a similar conclusion by studying breast tumor infiltration and HER-2/neu expression [183]. In high-grade or HER-2/neu-positive cases, lymphocytic infiltration was associated with improved overall survival. This study suggested that "macrophages predominated in tumors for which (lymphocytic infiltration) is indicative of good prognosis.…whereas T cells were more frequent in other subgroups for which (lymphocytic infiltration) had no prognostic significance" [183]. This might suggest a role for ADCC in naturally occurring immunity. However, Barbera-Guillem and colleagues have demonstrated that breast tumor–infiltrating macrophages (TIM) produce VEGF, and that increased VEGF production was associated with the presence of endocytosed sialyl-Tn-IgG complexes in vivo [184]. In vitro, binding of secreted/shed tumor glycoproteins and antibodies against these glycoproteins also increased the secretion of IL-1β, IL-6, and VEGF by stromal cells [56]. These immune complexes also induced proliferation of FcγRI-expressing tumor cells [185]. As a consequence, binding of IgG to shed glycoprotein tumor-associated antigens was associated with increased tumor invasion, metastasis, and angiogenesis in vitro [56]. These studies would suggest that a B-cell response against tumor might promote tumor growth. The only direct support for a TIL-B role in tumor control comes from an animal model. The chemically induced mammary adenocarcinoma of rat, MADB106, metastasizes primarily to lung when cultured cells are injected i.v. into rats [186]. Development of TIL-B in MADB106 lung metastases was rapid, and ablation of B cells with anti–B-cell antibody infusion resulted in an increase in lung tumors. This suggests that B-cell antitumor responses and perhaps TIL-B might be important in tumor control, although the relevance to human tumors is unknown.

Typical medullary carcinoma of the breast

Typical medullary carcinoma of the breast (TMC) accounts for approximately 5% of all breast cancer cases. Because TMC has a number of unusual histologic and immunologic features, it is treated separately in this review. TMC tumors are circumscribed but anaplastic, high grade, with large cells, abundant cytoplasm, numerous mitoses, and a diagnostic plasmacytic infiltrate (reviewed in [187]). Despite these unpromising prognostic features, patients with TMC have improved survival in comparison to matched cases of infiltrating ductal carcinoma of the breast. Long-term survival of individual TMC patients is directly correlated with the degree of tumor infiltration [188, 189, 190], and infiltration is correlated with decreased tumor size [191].

We demonstrated that a majority of TIL-B are plasma cells are of the IgG isotype [162], in contrast to normal breast, which typically contains low numbers of mainly IgA isotype plasma cells [192]. TIL immunoglobulin sequences showed a preponderance of clonal groups, as identified by germline gene usage and junctional mutation patterns, indicating intratumoral proliferation. Based on the presence of clonal terminally differentiated plasma cells, we also determined that TIL-B differentiate into plasma cells in situ. TIL IgG genes exhibited patterns of mutations that were consistent with antigenic selection and affinity maturation. From these data, we inferred that the diagnostic PC-infiltrate in medullary carcinoma of the breast is compatible with a tumor antigen–driven humoral immune response. This led us to speculate that TIL-PC in TMC might account for its favorable prognosis.

However, the dominant epitope for TMC TIL-B is aberrantly processed B-actin in apoptotic TMC tumor cells [9, 10]. Hansen et al. cloned phage-displayed IgG antibody libraries from TMC TIL-B, and selected tumor-binding antibodies on fresh-frozen TMC tumor sections rather than intact cultured cells, a critical choice as B-actin is not present on the cell surface of viable cultured cells. Dominant tumor-reactive IgG clones were isolated from each of two patients. All clones exhibited similar cytoplasmic staining of breast cancer cells, and were reactive with a 43-kDa protein band. Reactive protein was identified by immunoblotting of protein from cultured TMC cells and mass spectrometry of the corresponding protein band from a parallel SDS-PAGE gel. B-actin was identified as the reactive protein. Anti-actin antibodies were highly specific to B-actin, and affinity matured [9]. While actin is not normally a cell surface protein, it was displayed on the surface of apoptotic but not necrotic or viable typical medullary carcinoma cells in vivo. Western analysis determined that primary TMC tumors contained actin peptides similar to those produced by granzyme B digestion, suggesting that B-actin was proteolyzed during apoptosis, and subsequently displayed upon the surface of TMC cells. It was suggested that fragmentation of actin in conjunction with other apoptotic products might increase the immunogenicity of this ubiquitous protein. Furthermore, TMC undergoes apoptosis at an unusually high rate in comparison to other breast tumors [193]. Since TMC expresses high levels of HLA-DR [194] (unusual for a cell of epithelial origin), the actin peptides may be presented in the context of MHC, driving the affinity-matured response against this normally nonimmunogenic protein. The significance of this to other histologic types of breast cancer is not clear, as typical medullary carcinoma has a number of anomalous immunologic features in comparison to nonmedullary breast cancers [188, 193, 194, 195]. Regardless, reactivity against actin is unlikely to directly effect tumor control, although it is possible that it contributes indirectly through participating in tumor cell opsonization and subsequent phagocytosis and tumor-antigen presentation.

Discussion and hypotheses

In summary, a wide variety of tumor-associated autoantigens elicit antibody responses in breast cancer patients. However, none of the epitopes are tumor-specific. The variety of antibody-targeted antigens indicates that virtually any protein expressed by breast tumor cells including such ubiquitously expressed proteins as actin, heat shock proteins, DNA polymerase, and cytokeratin, can elicit a humoral response. What then is the basis of an immune response against these proteins? Because many tumor-associated autoantigens are not immunogenic under normal circumstances, we suggest that expression in tumor enhances the development of an antibody response. Some of the tumor-related factors contributing to immunogenicity of tumor-associated autoantigens are apoptosis-associated proteolysis and/or dendritic cell loading, heat shock protein binding, overexpression, aberrant processing, and sequestered expression. Tumor-adjuvant activity has been demonstrated in murine models [196]. In one study, ovalbumin was administered either in Freund's complete adjuvant, or in the form of ovalbumin-transfected tumor cells. The tumor-expressed ovalbumin elicited antibody titers equivalent to those achieved by vaccination with adjuvant, suggesting that factors in the tumor itself enhanced the response. In addition, some proteins are undoubtedly more immunogenic than others due to intrinsic properties.

Apoptosis

Apoptosis can increase tumor immunogenicity by a number of mechanisms. Apoptosis induced by neoadjuvant paclitaxel chemotherapy leads to increased lymphocytic infiltration of breast tumors [197]. Increased proteolytic cleavage of tumor-associated autoantigens with apoptosis may increase immunogenicity. For example, TMC tumors contained actin peptides similar to those produced by granzyme B digestion, and these fragments are present on the cell surface of apoptotic tumor cells [9, 106]. Although the role of MHC was not determined in these studies, it is known that TMC expresses high levels of HLA-DR [194], which is unusual for a cell of epithelial origin. It is possible that the actin peptides are presented in the context of MHC, driving the affinity-matured response against this normally nonimmunogenic protein. Poly(ADP-ribose) polymerase is a ubiquitously expressed nuclear protein against which a serum antibody response was detected [34]. When MDA-MB-468 breast tumor cells were induced to undergo apoptosis by ionizing radiation, caspase-mediated proteolysis of the nuclear enzyme poly(ADP-ribose) polymerase produces multiple peptide fragments [198]. This also may reveal cryptic peptides for MHC I or MHC II display by antigen-presenting cells that engulf apoptotic tumor cells. Spectrins are also degraded during apoptosis [199], and are known to elicit an antibody response in breast cancer patients [113].

Increased apoptosis also leads to loading of local dendritic cells with tumor-associated antigens, dendritic cell maturation, and induction of antitumor immunity [200, 201]. The maturation state of breast tumor–infiltrating DCs correlated with improved survival [166], suggesting that this process might be central to the development of effective antitumor immunity in vivo. When CD83⁺ (mature) TIL-DCs were present, relapse-free survival at 6 years was almost 90%, but only ~65% when CD83⁺ TIL-DC were absent. DCs induce differentiation of germinal center B cells into plasma cells. While lymph node follicles contain a population of CD83⁺ DCs [202], it is unknown what if any role these cells might have on B-cell responses. The maturation state of DCs also influences the humoral response indirectly through stimulation of antigen-specific CD4⁺ T cells.

Heat shock protein binding

Binding of tumor-associated antigens to heat shock proteins (HSPs) may also increase protein immunogenicity. For example, p53 only elicits an antibody response when bound to HSP70 in the breast tumor [86]. In contrast, tumors that did not elicit an antibody response did not contain HSP70-p53 complexes. The native immunogenicity of HSPs has been demonstrated in many studies, and a variety of antitumor vaccination strategies have been proposed that utilize HSP proteins as adjuvant (reviewed in [87]). Furthermore, HSP70 and other HSPs are up-regulated in apoptotic cells [199].

Aberrant processing

In the case of MUC1, overexpression and aberrant posttranslational processing contribute to immunogenicity. While MUC1 is expressed by most glandular epithelial cells and many hematologic cell types (reviewed in [59]), but is overexpressed by breast tumors. On breast tumors, MUC1 glycans are truncated and sialylated [64, 65], exposing the underlying protein epitopes. This aberrant posttranslational processing contributes to MUC1 immunogenicity by increasing both MHC class I and class II presentation [66], although serum antibodies are against both MUC1 glycans and peptide motifs.

Sequestered expression

Sequestered expression of antigens can contribute to immunogenicity. For example, HER-2/neu is expressed primarily in cardiac tissue [203] in addition to breast tumors. Naturally occurring antibody responses against cancer-testis antigens are also common in breast and other cancers [34]. Examples include thyroid hormone–binding protein [32], cathepsin L2 [33], NY-BR-1 [34], NY-ESO-1 [34], and SSX2 [34]. Oncofetal antigens are also among the antigens identified by antibody responses, including IGF-II mRNA-binding protein I [204], P62 [204], and Koc [204]. Another group of proteins with enhanced immunogenicity due to sequestered expression are what we have termed the cancer-CNS proteins. Unfortunately, antibody responses against these proteins are associated with paraneoplastic neurologic syndromes. One example is amphiphysin, which is expressed in breast cancer, and at low levels in a variety of normal tissues, but is primarily expressed in the brain [109]. Other cancer-CNS antigens with identified antibody responses in breast cancer include Yo [114], nova [115], and pentraxin [33]. Although the initial B-cell repertoire is selected against autoantigen reactivity in the bone marrow, it is unlikely that lymph node cells are exposed to every possible autoantigen throughout the life of an individual, which may lead to "repertoire drift" or the survival of B and T cells that are reactive with autoantigens. Lymph node dendritic and follicular dendritic cells are exposed to antigens expressed in tissues draining to that node, but not to antigens absent or expressed at low levels in local tissue. As a result, an antibody response may be initiated against antigens that are expressed only at distant or immunologically privileged sites.

Intrinsic immunogenicity

Some tumor-associated autoantigens have more innate immunogenicity than others. Many of the breast tumor–associated proteins against which there are antibody responses also elicit an antibody response in autoimmune disease, suggesting that these proteins have native immunogenicity. For example, UIsnRNP elicits antibody responses in both breast cancer [34] and in systemis lupus erythematosus (SLE) [205]. Of the 74 antigens listed in Table1, ~20% are also the target of antibodies in a variety of autoimmune diseases. Antibodies can also be detected in healthy individuals against almost 30% of the antigens in Table 1. The basis of the intrinsic immunogenicity of individual proteins may be due to the presence of high-affinity HLA-binding peptides, proteolytic sites conducive to production of HLA peptides, mimicry between self-proteins and foreign proteins, and other factors.

Future directions

While naturally occurring B-cell responses against breast tumors are common, they are not currently effective in tumor control. An effective high titer antibody response against breast tumors must initiate the response, generate high titers, and must target appropriate targets. The presence of intratumoral germinal centers may present an immunologic loophole for the initiation of antibody responses against tumor-associated autoantigens. Ectopic germinal centers are thought to be etiologic in the development of responses against autoantigens in autoimmune diseases such as rheumatoid arthritis [20]. It is possible that the normal mechanisms of autoimmune suppression are absent in ectopic germinal centers. In the future, it would be useful to determine if ectopic germinal centers are in fact more permissive than those located in lymph node follicles. If so, initiation and augmentation of intratumoral germinal centers might be a means of enhancing antitumor immunity. Studies of rheumatoid arthritis synovium demonstrated that a population of CD40L⁺CD8⁺ T cells is necessary for the formation of ectopic germinal centers [163]. Although CD4⁺ T cells are more typically viewed as the stimulators of B cells, CD40L⁺CD8⁺ T cells vigorously stimulate B-cell proliferation and antibody secretion [206]. This cell population has not been described in breast tumors, and intratumoral germinal centers are not well characterized in any cancer.

Tolerance is a significant obstacle to induction of high titer antibodies. Naturally occurring antibody responses may be initiated in tumor and node, yet fail to convert to a high titer systemic response due to systemic tolerance against the tumor-associated autoantigen. It has been demonstrated in murine models that all mice are capable of mounting low titer primary responses against autoantigens, but that only mice susceptible to autoimmunity produce high-titer secondary antibody responses [11]. A negative association between autoimmune diseases such as systemic lupus erythematosus (SLE) and rheumatoid arthritis and some cancers has been noted [207, 208], suggesting that a predisposition to autoimmunity may impart some protection against cancer. High systemic levels of interferon α have been described in SLE, as have elevated numbers of B and plasma cells [1, 49, 54]. GM-CSF has also been implicated in the differentiation of TIL-B into plasma cells [209]. Signaling molecules, chemokines and cytokines required for the organization of secondary follicles such as LIGHT [210], CXCL13 [180], lymphotoxin-α [211], and secondary lymphoid tissue chemoattractant (SLC) [212] may also prove to be fruitful areas of investigation for the induction of high-titer secondary antibody responses against tumor-associated autoantigens.

Selection of appropriate targets is also a crucial issue. Targets of antibody therapy should be cell surface, and relatively specific to tumor. Most naturally occurring antibodies are directed against cytoplasmic antigens, but cell surface targets will be required to effect tumor killing. TIL-B can potentially provide an enriched source of antibodies against cell surface targets. Although the repertoire of tumor-infiltrating B cells is extremely limited and cell surface tumor antigens are normally poorly immunogenic, we have cloned multiple cell surface reactive antibodies from TIL-B. TIL-B antibodies have also undergone affinity maturation, and may be of high affinity [19]. Selection of naturally immunogenic antigens and induction of antibody responses against tumor-associated antigens are also critical components of vaccine-based therapies. Antibody responses are a requisite component of effective antitumor immune responses in murine systems [16] and associate with clinical course in human antitumor vaccine protocols [17, 18]. Antibody responses can augment vaccine responses through direct antitumor effects including ADCC and complement activation [49], participation in epitope spread [213], potentiation of dendritic cell responses [214], and cross-priming of cytotoxic T cells [215]. Vaccination with antibody-bound antigen elicits protective immunity, while vaccination with antigen alone does not [216]. Future investigations of naturally occurring antibody responses against breast cancer may thus provide new antibody reagents for therapy, target antigens, and insights for the development of more effective vaccination strategies.