QSOX1 Expression Is Associated with Aggressive Tumor Features and Reduced Survival in Breast Carcinomas

Modern Pathology (2016) 29, 1485–1491

QSOX1 expression is associated with aggressive tumor features and reduced survival in breast carcinomas Gøril Knutsvik1,2, Karin Collett2, Jarle Arnes1,2, Lars A Akslen1,2 and Ingunn M Stefansson1,2

1Centre for Cancer Biomarkers, CCBIO, Department of Clinical Medicine, Section for Pathology, University of Bergen, Bergen, Norway and 2Department of Pathology, Haukeland University Hospital, Bergen, Norway

The biological role of quiescin sulfhydryl oxidase 1 (QSOX1) in tumor development is not well known, and its relation to breast cancer progression and prognosis is controversial. Here, our aim was to study the expression pattern and prognostic impact of QSOX1 in breast cancer, in relation to molecular subgroups and tumor cell proliferation. We examined a population-based series as part of the prospective Norwegian Breast Cancer Screening Program, including all women (50–69 years) diagnosed with breast cancer in one county of Norway during 1996–2003. QSOX1 expression was assessed by immunohistochemistry on tissue microarrays (n = 458). Median follow-up time was 13 years. High expression of QSOX1 protein was associated with features of poor prognosis including high histologic grade, hormone receptor negativity, HER2 positivity, and increased tumor cell proliferation. High QSOX1 expression was further associated with reduced breast cancer-specific survival in both univariate and multivariate analysis, independent of molecular subtypes. High QSOX1 expression is a strong and independent factor of reduced survival in breast cancer, also reflected by elevated levels in more aggressive molecular subgroups. QSOX1 expression may represent a biomarker for aggressive disease and a potential treatment target. Modern Pathology (2016) 29, 1485–1491; doi:10.1038/modpathol.2016.148; published online 26 August 2016

Breast cancer is the most common type of malignant in breast and pancreatic cell lines.11,12 Secreted tumors among women, and despite advances in QSOX1 may also have an important role in diagnosis and therapy, it is a major cause of illness creating a pro-invasive environment.2 and death. Thus, it is important to explore the However, the prognostic impact of QSOX1 clinical relevance of novel biomarkers and possible expression in breast cancer remains controversial. treatment targets. One report showed that high QSOX1 mRNA Quiescin sulfhydryl oxidase (QSOX1) is an expression was associated with an adverse prognosis 11 enzyme mainly localized to the Golgi apparatus.1,2 in Luminal B breast cancer. In contrast to this, high QSOX1 mRNA expression has been correlated It catalyzes disulfide bond formation in proteins 13 during folding, with concurrent reduction of oxygen with favorable outcome. Thus, the value of QSOX1 expression as a prognostic biomarker in to hydrogen peroxide.3 QSOX1 is also secreted, and breast cancer remains undetermined. highest expression has been shown in secretory 4 On this background, our aim was to study the cells. expression of QSOX1 in a well-defined population- Although previously associated with fibroblast based series of breast cancers, with clinico- quiescence,5 it has become increasingly clear that 6 pathologic associations and prognostic impact as QSOX1 is also related to malignancy. Elevated end points. Second, we wanted to assess QSOX1 QSOX1 expression has been shown in prostate, expression in relation to tumor cell proliferation and 7–10 breast, and pancreatic cancer, and high levels molecular subgroups of breast cancer. have been associated with an invasive phenotype

Correspondence: IM Stefansson, MD, PhD, Centre for Cancer Materials and methods Biomarkers, CCBIO, Department of Clinical Medicine, Section for Pathology, University of Bergen, Bergen N-5021, Norway. Patient Series E-mail: [email protected] Received 23 March 2016; revised 8 July 2016; accepted 8 July A population-based series of 534 breast carcinomas 2016; published online 26 August 2016 was studied.14 In brief, women aged 50–69 years www.modernpathology.org QSOX1 in breast cancer 1486 G Knutsvik et al

diagnosed with primary breast carcinoma during Inc., Rosemont, IL, USA), corresponding to pre- 1996–2003 as part of the prospective Norwegian viously published protocols aside from a small Breast Cancer Screening Program were included. adjustment of concentration.10–13 Detection was The patients received treatment according to done by Dako EnVision rabbit for 30 min. standard protocols at the time. Follow-up information was provided by the Norwegian Cause CK5/6. Staining procedures were performed using of Death Registry, and can be considered accurate the K4061/Envision Dual Link System (rabbit and complete. Last date of follow-up was 31 +mouse). Slides were incubated for 30 min with a December 2011. Outcome data include survival mouse monoclonal antibody (Dako M7237, clone status, survival time, and cause of death. During D5/16 B4) at a 1:200 dilution.16 the follow-up period, 79 patients (15%) died from breast carcinoma, and 62 (12%) died from other causes. The median follow-up was 13 years P-cadherin. P-cadherin staining was obtained by calculated by the reverse Kaplan–Meier method. incubating sections for 60 min with a mouse This study was approved by the Western Regional monoclonal antibody (clone 56) from BD Biosciences (San Jose, CA, USA) at a dilution of 1: Committee for Medical and Health Research Ethics, 15,17 REC West (REK 2012/1704). 400. Detection was done by Dako Envision mouse for 30 min. Sections were developed in DAB chromogen (Dako REAL DAB+chromogen) and Clinico-Pathological Variables counterstained with Dako REAL hematoxylin. For QSOX1, sections from a breast carcinoma The following variables were recorded: age at known to yield positive staining were used as diagnosis, tumor diameter, histologic type, histologic controls. For CK5/6 and P-cadherin; tonsillar tissue grade, lymph node status, and hormonal receptor was used as positive control. Negative controls were status. Mitotic count was done separately and obtained by replacing the primary antibody with reported as mitoses per mm2. Furthermore, selected Tris-buffered saline. results of previously investigated biomarkers (HER2, Ki67) were included.14 Of 534 patients in the complete cohort, tissue microarrays were available for 458 patients (86%). Evaluation of Staining Tissue microarrays were examined and scored Tissue Microarrays by one pathologist (GK) blinded to patient characteristics and outcome. Staining was recorded H&E-stained slides were used for tumor verification, 14 by use of a semi-quantitative and subjective grading as previously described. In brief, 1.0 mm cores in system as previously described.18,19 In brief, a triplicate were punched and mounted into a staining index (values 0–9) was calculated as recipient paraffin block using a semiautomated a product of staining intensity (0–3) and proportion precision instrument (Minicore 3, Tissue Arrayer, of tumor cells showing a positive reaction (0 = no Alphelys, France). Areas of high tumor purity, staining, 1o10%, 2 = 10–50% and 3450%). In the tumor periphery and the highest histologic grade subsequent statistical analyses, cutoff points were were selected. 190 cases had previously been mainly based upon median or quartile values, also processed;15 from these cases three tissue cores with considering the frequency distribution for each a diameter of 0.6 mm were obtained by a different marker. Categories with similar survival estimates instrument (Beecher Instruments, Silver Spring, were merged. MD, USA). QSOX1 staining was cytoplasmic with a granular, dot-like appearance. In some sections a weak, cytoplasmic staining of normal breast epithelium, Immunohistochemistry endothelium, stromal cells, and lymphocytes was Staining was performed on thin tissue microarray noted. There are no validated QSOX1 cutoff points sections (5 μm) from formalin-fixed and paraffin- based on immunohistochemical staining. In this embedded tumor tissue. The sections were de-waxed study, QSOX1 low expression encompassed with xylene/ethanol before target retrieval in a carcinomas showing weak to moderate staining microwave oven (6th Sense) in Target Retrieval and/or more focal staining (staining index 0–6). Solution (pH = 6) for 20 min. Endogenous peroxidase QSOX1 high expression was denoted in cases activity was blocked. Staining procedures were demonstrating strong cytoplasmic staining intensity performed using a DAKO autostainer. in 50% or more of the tumor cells (staining index 9). CK5/6 positivity was cytoplasmic. P-cadherin QSOX1. Sections were incubated for 60 min at staining was mainly membranous, although some room temperature (22 °C) with a rabbit polyclonal cytoplasmic staining was observed in cases showing QSOX1 antibody diluted 1:100 (Proteintech Group, strong overall staining.

Modern Pathology (2016) 29, 1485–1491 QSOX1 in breast cancer G Knutsvik et al 1487

For CK5/6, a positive staining was defined as assess the agreement of categorical scores. Ordinal staining index40, whereas for P-cadherin, staining variables were compared between two or more index43 was used, as previously described.15,20 groups using Mann–Whitney U or Kruskal–Wallis One observer (GK) assessed the series twice, with H-tests. good intra-observer reproducibility, Kappa-value For survival analyses, the end point was breast 0.89, Po0.001. For QSOX1, a subset (n = 66) was cancer-specific survival, defined as the time in scored by two observers (GK, IS) with substantial months from diagnosis until death from breast concordance (kappa-value 0.83, Po0.001). For cancer. Patients with missing data were excluded P-cadherin and CK5/6, good inter-observer from analyses. Univariate survival analyses were agreement has previously been shown.15 performed using the product-limit procedure (Kaplan–Meier method), and differences between categories were estimated by the log-rank test. Surrogate Definitions of Molecular Subclasses of Patients who died from other causes were censored. Breast Cancer Univariate and multivariate survival analyses were conducted using Cox’ proportional hazards methods Molecular classes were defined as proposed by 21 (likelihood ratio test). Multivariate analyses were St Gallen 2013, with minor modifications. Luminal adjusted for standard prognostic factors including A-like: ER and/or PR positive, HER2 negative, tumor size, histologic grade, nodal status and age. o Ki67 20%; Luminal B-like (HER2 negative): ER Covariates were examined by log–log plot to deter- ≥ positive, HER2 negative, Ki67 20% and/or mine their ability to be incorporated in multivariate PR negative; Luminal B-like (HER2 positive): ER models, and tests for interactions were performed. positive and HER2 overexpressed or amplified, any For continuous variables, the proportional hazard Ki67, any PR; HER2 positive (non-luminal): HER2 assumptions were also assessed by studying the overexpressed or amplified, ER and PR absent; graphs of Schoenfelds residuals. Triple negative: ER and PR absent, HER2 negative. Our modifications include the use of an ER and PR cutoff point of 10% according to national guidelines Results at that time. A small subgroup (11 cases) in our series showed ER negativity, PR positivity and HER2 In total, 55 (12%) of the breast cancer cases showed negativity; these were not possible to classify by strong QSOX1 staining, whereas 403 cases showed the 2013 criteria. Based on the St Gallen 201122 we low expression. Images of tumors with high and low decided to categorize these in the luminal subgroup. QSOX1 expression are displayed in Figure 1.

Statistical Methods Association of QSOX1 Expression with Clinico- Pathologic and Immunohistochemical Features Analyses were performed using IBM SPSS Statistics for Windows, Version 21.0. (Armonk, NY, USA: IBM High QSOX1 expression was associated with high Corp). Statistical significance was assessed at histologic grade, ER and PR negativity, HER2 the two-sided 5% level. Associations between positivity, and P-cadherin positivity, as summarized categorical variables were evaluated by Pearson’s in Table 1. No association was found between χ2-test. The Cohen’s kappa measure was used to QSOX1 expression and axillary lymph node status.

Figure 1 Immunohistochemical staining showing (a) strong and (b) weak expression of QSOX1 in breast cancer (magnification × 400).

Modern Pathology (2016) 29, 1485–1491 QSOX1 in breast cancer 1488 G Knutsvik et al

Table 1 Associations between QSOX1 protein expression and clinico-pathologic variables

QSOX1 QSOX1 Variables low high OR 95% CI P-valuea

Tumor diameter ≤ 2 cm 305 35 1 42 cm 98 20 1.78 0.98–3.22 0.055

Histologic grade 1–2 341 33 1 3 62 22 3.67 2.01–6.71 o0.001

Nodal statusb Negative 289 37 1 Positive 110 18 1.28 0.70–2.34 NS

ER Positive 346 33 1 Negative 57 22 4.05 2.20–7.40 o0.001

PR Positive 293 24 1 Negative 110 31 3.44 1.93–6.13 o0.001

HER2 Negative 360 34 1 Positive 43 21 5.17 2.76–9.70 o0.001

Mitotic countc ≤ 5.5 314 19 1 45.5 89 36 6.69 3.66–12.22 o0.001

Ki67 countc ≤ 31.5 314 22 1 431.5 89 33 5.29 2.94–9.53 o0.001

CK5/6 Negative, 351 43 1 score = 0 Figure 2 Tumor cell proliferation as estimated by median Ki67 (a) Positive, 52 12 1.88 0.93–3.81 0.074 4 and median mitotic count (b) according to QSOX1 expression score 0 levels by staining index (0–9). P-cadherin Low, score ≤ 3 328 32 1 High, 75 23 3.14 1.74–5.68 o0.001 HER2+ subgroup, 8 out of 23 cases (35%) showed score43 high QSOX1 level. A subset of the lumB/HER2+ cases (13 cases/32%) also showed high QSOX1 Abbreviations: CI, confidence interval; ER, estrogen receptor; NS, not significant; OR, odds ratio; PR, progesterone receptor. expression. aχ2-test. bFour cases with missing information on nodal status. cCutoff value by upper quartile. QSOX1 Expression in Relation to Patient Outcome

QSOX1 Expression in Relation to Proliferation Within the QSOX1 low group, no differences in survival were demonstrated for cases showing High QSOX1 expression showed significant negative staining (staining index 0–1) compared associations with increased tumor cell proliferation with cases with weak to moderate staining (staining assessed by Ki67 and mitotic count. (Table 1, see also index 2–6). These subgroups were therefore merged Figure 2). (QSOX1 low). High QSOX1 expression was significantly associated with reduced breast cancer- specific survival in univariate survival analysis o QSOX1 Expression in Molecular Subgroups (Figure 3, log-rank test, P 0.001) with 10 year survival of 67% for cases showing high QSOX1 There were significant differences in QSOX1 expression compared to 89% for cases with expression according to molecular subgroups low expression (Table 3). Univariate Cox survival (Kruskal–Wallis H-test, Po0.001). In the triple- analysis showed poor outcome for cases with high negative subgroup, 14 out of 46 cases (30%) showed QSOX1 expression with a hazard ratio of 3.07 high QSOX1 expression, (Table 2), whereas in the (confidence interval 1.83–5.16, Po0.001).

Modern Pathology (2016) 29, 1485–1491 QSOX1 in breast cancer G Knutsvik et al 1489

Table 2 QSOX1 expression according to molecular subgroups of breast cancer, based upon St Gallen guidelines 2013

Molecular subgroup QSOX1 low SI 0–6 QSOX1 high SI 9 Total

N (%) N (%)

Lum A 171 (98) 4 (2) 175 Lum B/HER2 − 157 (91) 16 (9) 173 Lum B/HER2+ 28 (68) 13 (32) 41 HER2+ 15 (65) 8 (35) 23 Triple negative 32 (70) 14 (30) 46 Total 403 (88) 55 (12) 458

Kruskal–Wallis test. Po0.001.

In the HER2+/HR − subgroup (n = 23) high QSOX1 expression was a significant prognosticator in univariate survival analysis (Supplementary Figure 3 Breast cancer-specific survival according to QSOX1 Figure 1A, log-rank test, P = 0.003). In the expression. For each category, the number of events is given triple-negative subgroup (n = 46), high QSOX1 levels followed by the number of cases. were significantly associated with reduced survival (Supplementary Figure 1B, log-rank test, P = 0.004). suggesting that QSOX1 promotes proliferation No associations were demonstrated between QSOX1 in vivo.23 In contrast, high QSOX1 expression has expression and outcome within the luminal/HER2- been correlated with decreased proliferation negative category (data not shown). in vitro24 and reduced tumor growth in vivo.13 Subsequently, we performed multivariate Cox We included the basal markers CK5/6 and survival analyses including the basic prognostic P-cadherin as a refinement of the triple-negative factors age, tumor size, histologic grade, and lymph subgroup.15,17 High QSOX1 levels were associated node status in addition to QSOX1 expression. In this with increased P-cadherin expression, whereas the model, tumor size, nodal status, and high QSOX1 association with CK5/6 did not reach statistical expression were independent predictors of prognosis significance (P = 0.074). (Table 4). By adding the luminal status to the model A previous study showed higher levels of QSOX1 (luminal versus non-luminal); tumor size, nodal mRNA in ER-negative breast cancer cell lines status, QSOX1, and luminal status retained compared to ER-positive cells.25 One study reported independent prognostic significance (data not that QSOX1 was repressed by estrogen stimulation in shown). There was no interaction between QSOX1 an ER-positive breast cancer cell line.26 Contrasting and luminal status. Finally, an extended model this, an association between high QSOX1 and ER with additional adjustment for CK5/6 and positivity was suggested by Katchman et al.,11 P-cadherin was examined. Here, independent prog- although this was not confirmed in our study. nostic impact was shown for tumor size, nodal To the best of our knowledge, this is the first report status, and QSOX1 expression (data not shown). of QSOX1 protein expression and prognosis in breast cancer. In our study, high QSOX1 level was associated with reduced survival in univariate Discussion analysis, and also showed independent prognostic impact in multivariate Cox analyses. This is in line Here, we examined the QSOX1 protein expression in with a previous report where high QSOX1 gene a large, population-based series of breast carcinomas expression was associated with poor prognosis,11 with long and complete follow-up. Our findings although we could not confirm that the prognostic indicate that high QSOX1 expression is associated impact was restricted to the Luminal B subgroup. We with an aggressive tumor phenotype characterized found no interaction between QSOX1 expression by high histologic grade, increased proliferative rate, and luminal status, and in subgroup analyses, hormone receptor negativity, HER2 positivity, and QSOX1 was prognostic within the HER2 positive reduced survival in both univariate and multivariate and triple-negative subgroups. models. Consistent with this, elevated QSOX1 In contrast, high QSOX1 mRNA expression has mRNA and protein levels in breast cancer have been associated with improved prognosis,13 and previously been related to high histologic grade.9,11 increased QSOX1 expression has been suggested In line with our results, an association between high to protect breast cancer cells against apoptosis,24 and QSOX1 protein expression and increased Ki67 has to inhibit autophagy in vitro.27 Contrasting to this, been shown.11 Moreover, knockdown of QSOX1 in Katchman et al.11 found no correlation between an animal model led to reduced tumor growth, QSOX1, apoptosis, or autophagy.

Modern Pathology (2016) 29, 1485–1491 QSOX1 in breast cancer 1490 G Knutsvik et al

Table 3 Univariate survival analysis according to QSOX1 and Table 4 Multivariate survival analysis according to Cox propor- histopathological variables tional hazards method

Variables N Events 5 years 10 years P-valuea Variables N HR 95% CI P-valuea

QSOX1 Tumor diameter Low, SI 0–6 403 50 95.2 89.2 ≤ 2 cm 340 High, SI 9 55 21 74.5 66.8 o0.001 42 cm 114 1.88 1.11–3.19 0.020

Tumor diameter Nodal status ≤ 2 cm 340 35 95.8 90.9 Negative 326 42 cm 118 36 83.8 73.8 o0.001 Positive 128 3.29 1.94–5.60 o0.001

Histologic grade QSOX1 1 178 16 96.0 92.5 Low, SI 0–6 399 2 196 34 94.8 85.9 High, SI 9 55 3.90 2.31–6.59 o0.001 3 84 21 81.0 74.8 0.001 Abbreviations: CI, confidence interval; HR, Hazard ratio. b Nodal status Final model after initial inclusion of age, tumor diameter, histologic Negative 326 29 95.6 92.4 grade, nodal status, and QSOX1. Positive 128 39 85.9 72.6 o0.001 Four cases with unknown nodal status were excluded. aLikelihood ratio test. ER Positive 379 45 96.5 90.4 Negative 79 26 74.4 67.8 o0.001 invasiveness. Moreover, it has been demonstrated PR that QSOX1 secreted from fibroblasts is required for Positive 317 35 96.8 91.2 proper laminin incorporation into the extracellular Negative 141 36 83.5 75.7 o0.001 matrix, thus supporting the migration of lung cancer 2 HER2 cells in vitro. Negative 394 55 94.4 88.1 Interestingly, QSOX1 might represent a possible Positive 64 16 82.8 76.3 0.013 therapeutic target. QSOX1 antibodies have been 2,30 c developed, and QSOX1 inhibitors have been Mitotic count assessed both in vitro and in vivo.23 QSOX1 might ≤ 5.5 333 37 96.1 90.7 45.5 125 34 83.9 75.3 o0.001 therefore represent a novel and important biomarker in breast cancer. However, it should be mentioned Ki67 countc that in the current study, outcome was assessed ≤ 31.5 336 39 96.1 91.0 4 o within a context where the treatment administered 31.5 122 32 83.6 74.1 0.001 may not have been homogeneous. The lack of CK5/6 treatment information on a case level may represent Negative, score = 0 394 53 94.9 88.7 a limitation in prognostic analyses. Positive, score40 64 18 79.7 72.9 0.001

P-cadherin Low, score ≤ 3 360 49 94.4 88.2 Conclusion High, score43 98 22 86.7 80.3 0.024 Taken together, our findings indicate that high aLog-rank test. QSOX1 protein expression reflects adverse tumor bFour cases with unknown nodal status were excluded. features and reduced breast cancer survival. The c Cutoff value by upper quartile. function of QSOX1 in these tumors is presently not clear and further studies are needed. Still, QSOX1 High QSOX1 expression in breast, pancreatic and expression may represent both a biomarker for renal cancer cell lines has been related to increased aggressive disease and a future treatment target. invasiveness.11,12,23 Also, an association between high QSOX1 and post-translational activation of MMP-2 and MMP-9 has been suggested.12 In breast Acknowledgments cancer, MMP-2 and MMP-9 are upregulated in We thank Gerd Lillian Hallseth, Randi Hope Lavik response to HIF-1 expression, and they facilitate and Bendik Nordanger for excellent technical invasion through degradation of collagen.28 assistance. This work was partly supported by the Intriguingly, a recent report indicated that QSOX1 Research Council of Norway through its Centres of expression in pancreatic cancer cells were induced Excellence funding scheme, project number 223250. by HIF-1 in vitro, with concurrent increased GK is supported by grant No 911637 from the invasiveness.29 Two binding sites for HIF response Helse Vest Research Fund. LAA is supported by elements were identified in the promoter region of other grants from the Norwegian Cancer Society, the the QSOX1 gene, implicating that QSOX1 is a target Research Council of Norway and Helse Vest gene for HIF-1. QSOX1 silencing led to decreased Research Fund.

Modern Pathology (2016) 29, 1485–1491 QSOX1 in breast cancer G Knutsvik et al 1491

Disclosure/conflict of interest compared with screen detected tumors. Cancer Epidemiol Biomarkers Prev 2005;14:1108–1112. The authors declare no conflict of interest. 16 Nalwoga H, Arnes JB, Wabinga H, et al. Expression of aldehyde dehydrogenase 1 (ALDH1) is associated with basal-like markers and features of aggressive tumours in African breast cancer. Br J Cancer 2010;102: References 369–375. 17 Arnes JB, Brunet JS, Stefansson I, et al. Placental 1 Chakravarthi S, Jessop CE, Willer M, et al. Intracellular cadherin and the basal epithelial phenotype of BRCA1- catalysis of disulfide bond formation by the human related breast cancer. Clin Cancer Res 2005;11: sulfhydryl oxidase, QSOX1. Biochem J 2007;404: 4003–4011. 403–411. 18 Aas T, Borresen AL, Geisler S, et al. Specific P53 2 Ilani T, Alon A, Grossman I, et al. A secreted disulfide mutations are associated with de novo resistance to catalyst controls extracellular matrix composition and doxorubicin in breast cancer patients. Nat Med 1996;2: function. Science 2013;341:74–76. 811–814. 3 Thorpe C, Hoober KL, Raje S, et al. Sulfhydryl oxidases: 19 Straume O, Akslen LA. Alterations and prognostic emerging catalysts of protein disulfide bond formation significance of p16 and p53 protein expression in in eukaryotes. Arch Biochem Biophys 2002;405:1–12. subgroups of cutaneous melanoma. Int J Cancer 4 Coppock DL, Thorpe C. Multidomain flavin-dependent 1997;74:535–539. sulfhydryl oxidases. Antioxid Redox Signal 2006;8: 20 Foulkes WD, Stefansson IM, Chappuis PO, et al. 300–311. Germline BRCA1 mutations and a basal epithelial 5 Coppock DL, Kopman C, Scandalis S, et al. Preferential phenotype in breast cancer. J Natl Cancer Inst 2003;95: gene expression in quiescent human lung fibroblasts. 1482–1485. Cell Growth Differ 1993;4:483–493. 21 Goldhirsch A, Winer EP, Coates AS, et al. Personalizing 6 Lake DF, Faigel DO. The emerging role of QSOX1 the treatment of women with early breast cancer: in cancer. Antioxid Redox Signal 2014;21:485–496. highlights of the St Gallen International Expert 7 Ouyang X, DeWeese TL, Nelson WG, et al. Loss-of- Consensus on the Primary Therapy of Early Breast function of Nkx3.1 promotes increased oxidative Cancer 2013. Ann Oncol 2013 24:2206–2223. damage in prostate carcinogenesis. Cancer Res 2005;65: 22 Goldhirsch A, Wood WC, Coates AS, et al. Strategies 6773–6779. for subtypes–dealing with the diversity of breast 8 Song H, Zhang B, Watson MA, et al. Loss of Nkx3.1 cancer: highlights of the St. Gallen International Expert leads to the activation of discrete downstream target Consensus on the Primary Therapy of Early Breast genes during prostate tumorigenesis. Oncogene Cancer 2011. Ann Oncol 2011 22:1736–1747. 2009;28:3307–3319. 23 Hanavan PD, Borges CR, Katchman BA, et al. Ebselen 9 Soloviev M, Esteves MP, Amiri F, et al. Elevated inhibits QSOX1 enzymatic activity and suppresses transcription of the gene QSOX1 encoding quiescin invasion of pancreatic and renal cancer cell lines. Q6 sulfhydryl oxidase 1 in breast cancer. Oncotarget 2015;6:18418–18428. PLoS One 2013;8:e57327. 24 Morel C, Adami P, Musard JF, et al. Involvement of 10 Antwi K, Hostetter G, Demeure MJ, et al. Analysis of sulfhydryl oxidase QSOX1 in the protection of cells the plasma peptidome from pancreas cancer patients against oxidative stress-induced apoptosis. Exp Cell connects a peptide in plasma to overexpression of the Res 2007;313:3971–3982. parent protein in tumors. J Proteome Res 2009;8: 25 Coppock D, Kopman C, Gudas J, et al. Regulation of the 4722–4731. quiescence-induced genes: quiescin Q6, decorin, and 11 Katchman BA, Ocal IT, Cunliffe HE, et al. Expression of ribosomal protein S29. Biochem Biophys Res Commun quiescin sulfhydryl oxidase 1 is associated with a 2000;269:604–610. highly invasive phenotype and correlates with a poor 26 Inoue A, Yoshida N, Omoto Y, et al. Development of prognosis in Luminal B breast cancer. Breast Cancer cDNA microarray for expression profiling of Res 2013;15:R28. estrogen-responsive genes. J Mol Endocrinol 2002;29: 12 Katchman BA, Antwi K, Hostetter G, et al. Quiescin 175–192. sulfhydryl oxidase 1 promotes invasion of pancreatic 27 Poillet L, Pernodet N, Boyer-Guittaut M, et al. QSOX1 tumor cells mediated by matrix metalloproteinases. inhibits autophagic flux in breast cancer cells. PLoS Mol Cancer Res 2011;9:1621–1631. One 2014;9:e86641. 13 Pernodet N, Hermetet F, Adami P, et al. High expres- 28 Gilkes DM, Semenza GL, Wirtz D. Hypoxia and the sion of QSOX1 reduces tumorogenesis, and is asso- extracellular matrix: drivers of tumour metastasis. Nat ciated with a better outcome for breast cancer patients. Rev Cancer 2014;14:430–439. Breast Cancer Res 2012;14:R136. 29 Shi CY, Fan Y, Liu B, et al. HIF1 contributes to 14 Knutsvik G, Stefansson IM, Aziz S, et al. Evaluation hypoxia-induced pancreatic cancer cells invasion via of Ki67 expression across distinct categories of promoting QSOX1 expression. Cell Physiol Biochem breast cancer specimens: a population-based study of 2013;32:561–568. matched surgical specimens, core needle biopsies and 30 Grossman I, Alon A, Ilani T, et al. An inhibitory tissue microarrays. PLoS One 2014;9:e112121. antibody blocks the first step in the dithiol/disulfide 15 Collett K, Stefansson IM, Eide J, et al. A basal epithelial relay mechanism of the enzyme QSOX1. J Mol Biol phenotype is more frequent in interval breast cancers 2013;425:4366–4378.

Supplementary Information accompanies the paper on Modern Pathology website (http://www.nature.com/ modpathol)

Modern Pathology (2016) 29, 1485–1491