A shift from membranous and stromal syndecan‐1 (CD138) expression to cytoplasmic CD138 expression is associated with poor prognosis in breast cancer

Syndecan‐1 (CD138) is a transmembrane proteoglycan expressed in normal and malignant tissues. It is of interest because of a possible prognostic effect in tumors and as a target for Indatuximab, a monoclonal antibody coupled to a cytotoxic agent. To assess the prognostic role of CD138 expression in breast cancer (BCa), a tissue microarray containing 1535 BCa specimens was analyzed by immunohistochemistry. Cytoplasmic, membranous, and stromal CD138 staining was separately analyzed. In normal breast tissue, CD138 staining was limited to epithelial cell membranes. In cancers, membranous staining tended to become weaker or even disappeared (38.3% of cancers with absence of membranous staining) but cytoplasmic and stromal staining newly appeared in 29.7% and 58.1% of cancers. Loss of membranous epithelial CD138 staining as well as presence of cytoplasmic and stromal CD138 positivity were—to a variable degree—associated with high pT, high grade, nodal metastasis, estrogen receptor‐negative, progesterone receptor‐negative, human epidermal growth factor receptor 2+, and poor overall patient survival. A combined analysis of epithelial and stromal CD138 expression revealed a link to overall patient survival (P < .0001) with best prognosis for patients with stromal positivity and absence of cytoplasmic staining, the worst prognosis for cancers with cytoplasmic staining and stromal negativity and intermediate prognosis for patients having either cytoplasmic staining or stromal negativity. In multivariate analyses, CD138 was not independent of established prognostic features. In summary, these data reveal a compartment depending prognostic effect of CD138 expression in BCa with cytoplasmic positivity being linked to aggressive cancer and stromal CD138 being linked to a more favorable prognosis.

effect of CD138 expression in BCa with cytoplasmic positivity being linked to aggressive cancer and stromal CD138 being linked to a more favorable prognosis. Currently used prognostic features for assessing a cancers aggressiveness mainly include tumor size, histological grade, lymph node metastasis, as well as immunohistochemical assessment of estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), and tumor cell proliferation (Ki67 labeling index). 2,3 As these parameters are statistically powerful but not sufficient to safely predict cancer properties in every patient, additional molecular parameters are analyzed in an increasing number of cases. [4][5][6] Molecular classifiers that are commercially available are based on the analysis of RNAs of 21 to 70 genes. [7][8][9] It is a major disadvantage of these tests that isolated RNAs from cancer tissues always include a variable quantity of noncancer cells.
Syndecan-1 (CD138) is the first member of the four-member syndecan family encoded by the SDC1 gene. [10][11][12] It is built out of three structural domains from which one is an extracellular domain capable to bind heparin sulfates and chondroitin sulfates. 13 As a cell surface protein, it plays a central role in cell-cell and cell-matrix interactions, functions as a coreceptor for chemokines and growth factors and is involved in cell proliferation, migration, and the organization of the cytoskeleton. 13 In normal tissues, CD138 is expressed on epithelial cells of various tissues and on plasma cells.
Altered CD138 expression has previously been described in various malignant tumors. 12 Overexpression of CD138 has been found in breast, gallbladder, urinary bladder, pancreatic, ovarian, endometrial, and prostate cancer. 13 Reduced CD138 expression as compared to normal tissues has been reported for lung, head/neck, gastric, renal, and colorectal cancer. 13 In several of these tumor types either reduced or increased CD138 expression was linked to unfavorable tumor phenotype and poor patient prognosis. [14][15][16][17] Thirteen different studies have analyzed the impact of CD138 expression on BCa prognosis analyzing 37 to 254 cancers by immunohistochemistry and described complex staining patterns including membranous, cytoplasmic, and stromal staining. [18][19][20][21][22] The conclusions from these studies were highly controversial, however, the authors described both increased, 17,23,24 and decreased 25 CD138 expression in BCa as compared to normal breast epithelium and reported associations of high CD138 expression with possibly favorable 25 and unfavorable, 17,20,23 prognosis.
To learn more on the prognostic role of CD138 expression in BCa and the potential impact of different staining patterns, a tissue microarray (TMA) was analyzed containing more than 1500 tumors with clinical follow-up data.

| Immunohistochemistry
Freshly cut TMA sections were stained in one experiment at 1 day.
Slides were deparaffinized and exposed to heat-induced antigen retrieval for 5 minutes in an autoclave at 121°C in pH 9 Dako Target

| CD138 in epithelial cells
Representative images of normal breast staining and cancers with purely cytoplasmic and membranous staining are shown in Figure 1.  (Table S2). For the tumor cell membrane, reduced CD138 staining tended to be associated with unfavorable tumor features, but this was statistically not always significant ( Figure 2A). Presence of cytoplasmic staining was, however, significantly linked to all evaluated unfavorable tumor features (Table   S2, P ≤ .0032) and also to reduced overall survival ( Figure 2B, P = .0038). A combinatorial approach defining four groups according to the presence or absence of membranous and cytoplasmic staining showed that only tumors with purely cytoplasmic (without concomitant membranous) staining-behaved particularly poor ( Figure 2C, P = .0037). Accordingly, the best prognostic distinction was seen if tumors were divided into tumors with pure "cytoplasmic staining" (membrane negative and cytoplasma positive) vs all "others" (membrane positive/cytoplasma positive; Figure 2D, P = .0021).

| CD138 in tumor stroma
Normal breast tissue completely lacked stromal CD138 staining.
Stromal staining was present in 58.1% of cancers (Table S3). Its intensity was unrelated to clinicopathological or outcome data (data | 2309 not shown) and was thus disregarded. Two distinct staining patterns were seen (Figure 3). In the peritumoral pattern, a distinct frame of high-intensity stromal staining of acellular material was surrounding cancer cells ( Figure 3A). In the diffuse staining pattern, CD138 staining was not topically related to cancer cells and potentially involved stromal cells such as fibroblasts or myofibroblasts in desmoplastic tumor stroma ( Figure 3B). Combinations of peritumoral and diffuse patterns did also occur ( Figure 3C). Both peritumoral and diffuse stromal staining were significantly linked to all analyzed favorable tumor features except HER2 status (Table S3) and were also associated with favorable patient outcome ( Figure 4A and 4B, P ≤ .0011 each). A combined analysis of the parameters "diffuse" and

| Combined epithelial and stromal CD138 analysis
On the basis of the results of the epithelial and stromal analysis we categorized cancers in four patterns A-D defined by presence or absence of cytoplasmic and stromal CD138 staining. Definitions of these patterns are detailed in Figure 5A. The patterns were significantly linked to all analyzed molecular and clinicopathological tumor features (P < .001 each, Table 2) and patient outcome ( Figure   5B). Patient prognosis was worst in cancers with purely cytoplasmic CD138 positivity in the absence of stromal staining (pattern D) and best in patients with stromal staining but absence of cytoplasmic cancer cell positivity (pattern A, P < .0001).

| CD138 and histological tumor subtypes
The relationship between CD138 and histologic tumor subtypes is given in Table 2. Cancers of no special type (NST) and lobular carcinomas-the two largest histopathologically defined subgroupsdiffered significantly in their stroma staining, which was particularly  Figure S1, P = .0027). This was also retained if nodal positive cancers were separately analyzed ( Figure S1, P = .0376). Significant associations between CD138 expression and tumor phenotype were also seen in the subgroup of NST cancers (Table S4).

| Multivariate analysis
Separate multivariate analyses were performed to test the independent prognostic impact of cytoplasmic CD138 staining (model 1), stromal CD138 staining (model 2), and the CD138 pattern (model 3).
For this purpose, CD138 data were compared with the classical histopathological prognosticators pT, pN, grade as well as the ER and PR status. In all three models, the pathological stage was the only histopathological parameter that predicted patient prognosis independently from the other parameters (P < .0001 each). Neither cytoplasmic CD138 (P = .3010), nor CD138 in the tumor environment (P = .3081) or the CD138 pattern (P = .5417) provided prognostic information independently from the tumor stage. All data are summarized in Table 3.  shown to induce reduced cell adhesion, increased invasive potential, and dysregulated growth of mammary epithelial cells in vitro. 28 The observed peritumoral and stromal CD138 staining could be a consequence of enzymatic cleavage of membrane-bound CD138, which is then shed as soluble SDC1 (sSDC1) into the tumor environment. 12 Our data identified a striking patient protective effect of stromal CD138 staining in BCas. Stroma staining exhibited two distinct patterns. The peritumoral pattern displayed a dense acellular rim-like CD138 positive zone demarcating the cancer cells from the stroma.

| DISCUSSION
The diffuse pattern was defined by staining of spindle-shaped cellular elements, morphologically reminiscent of fibroblasts or myofibroblasts. We cannot exclude, however, that another more specific cell type was detected by CD138 in these cases. Sharpe et al 29 recently described a new CD138 positive cell type in the stroma of high-grade prostate cancers. These cells were excluded to represent typical stromal, epithelial or immune cell types by multicolor immunefluorescence. 29 The same peritumoral and diffuse patterns of CD138 stroma staining as described in this study had earlier been reported by Stanley et al. 28 Either alone or in combination, these stroma patterns occurred in 58.1% of our cancers. This frequency is comparable to the findings of most of the earlier studies, which had described stromal staining in smaller cohorts in the range of 9%   In summary, our data suggest a complex role for CD138 in BCa-with variable impact on disease outcome. Our study also demonstrates the high importance of detailed topographical and cell type-specific information for understanding the role of biological molecules. The prognostic impact of CD138 expression was considerable but not independent of established parameters in this study. However, CD138 analysis may become important if ongoing efforts to develop anti-CD138 treatments should be successful.

ACKNOWLEDGMENTS
We are grateful to Janett Lütgens, Sünje Seekamp, and Inge Brandt for excellent technical assistance.
T A B L E 3 Cox proportional hazard ratios for raw survival of established prognostic markers and cytoplasmic CD138 (model 1), environmental CD138 (model 2) and the CD138 staining pattern (model 3) Abbreviations: CI, confidence interval; ER, estrogen receptor; PR, progesterone receptor.