Dopamine and cAMP-regulated phosphoprotein 32 kDa (DARPP-32) and survival in breast cancer: a retrospective analysis of protein and mRNA expression

Dopamine and cAMP regulated phosphoprotein 32 kDa (DARPP-32) also known as phosphoprotein phosphatase-1 regulatory subunit 1B and encoded by the PPP1R1B gene is an inhibitor of protein phosphatase-1 and protein kinase A. DARPP-32 is expressed in a wide range of epithelial cells and some solid tumours; however, its role in breast cancer is only partially defined. DARPP-32 expression was determined using immunohistochemistry in two independent cohorts of early stage invasive breast cancer patients (discovery n = 1352; validation n = 1655), and 112 HER2 positive breast cancer patients treated with trastuzumab and adjuvant chemotherapy. PPP1R1B mRNA expression was assessed in the METABRIC cohort (n = 1980), using artificial neural network analysis to identify associated genes. In the discovery cohort, low nuclear expression of DARPP-32 was significantly associated with shorter survival (P = 0.041), which was independent of other prognostic variables (P = 0.019). In the validation cohort, low cytoplasmic and nuclear expression was significantly associated with shorter survival (both P = 0.002), with cytoplasmic expression independent of other prognostic variables (P = 0.023). Stronger associations with survival in oestrogen receptor (ER) positive disease were observed. In patients treated with trastuzumab, low nuclear expression was significantly associated with adverse progression-free survival (P = 0.031). In the METABRIC cohort, low PPP1R1B expression was associated with shortened survival of ER positive patients. Expression of CDC42 and GRB7, amongst others, were associated with PPP1R1B expression. This data suggests a role for DARPP-32 as a prognostic marker with clinical utility in breast cancer.


Methods patient cohorts.
This study is reported according to reporting recommendations for tumour marker prognostic studies (REMARK) criteria 26 . For protein expression three well-characterised patient cohorts were used; the discovery cohort, functioned as a discovery set; the validation cohort, functioned as a validation set and the HER2 cohort was used to assess DARPP-32 expression in HER2 positive patients treated with trastuzumab. Breast cancer specific survival was calculated as the time interval between primary surgery and death resultant from breast cancer. Progression-free survival was defined as the date of surgery to relapse (including local and regional relapse). Discovery cohort. 1352 early stage invasive breast cancer patients were available for assessment in the discovery cohort, with all patients treated at Nottingham University Hospitals between 1987 and 1998. All patients were managed in a standard manner, where all patients underwent a mastectomy or wide local excision, as decided by disease characteristics or patient choice, followed by radiotherapy if indicated. Patients received systemic adjuvant treatment on the basis of Nottingham Prognostic index (NPI), oestrogen receptor (ER), and menopausal status. Patients with an NPI score less than 3.4 did not receive adjuvant treatment and patients with an NPI score of 3.4 and above were candidates for CMF combination chemotherapy (cyclophosphamide, methotrexate and 5-fluorouracil) if they were ER negative or premenopausal; and hormonal therapy if they were ER positive. Median follow-up was 205 months determined using the reverse Kaplan-Meier method and clinicopathological information for this cohort is available in Table 1.
Validation cohort. 1655 early stage invasive breast cancer patients were available for assessment in the validation cohort, with all patients treated at Nottingham University Hospitals between 1998 and 2006. All patients were managed in a standard manner, as described for the discovery cohort. Median follow-up was 148 months determined using the reverse Kaplan-Meier method and clinicopathological information for this cohort is available in Table 1. HER2 positive cohort. 112 HER2 positive breast cancer patients were available for assessment in the HER2 positive cohort, with all patients treated at Nottingham University Hospitals between 2004 and 2012. Patients were treated according to local guidelines, with adjuvant therapy and trastuzumab following surgery. Adjuvant hormone therapy was received by 47% of patients (40/75), with 74% of patients receiving adjuvant radiotherapy (59/80). Trastuzumab was given on a 3-weekly regimen for 52 weeks, with patients receiving trastuzumab following six cycles of 3-weekly FEC chemotherapy (fluorouracil, epirubicin and cyclophosphamide) or patients receiving three cycles of FEC, followed by three cycles of taxane (FEC-T), to which, trastuzumab was frequently added from the second cycle of taxane onwards. Median follow-up was 50 months determined using the reverse Kaplan-Meier method and clinicopathological information for this cohort is available in Table 2.
MetABRic series. Details of the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) data set (n = 1980) data set have been published elsewhere 27 . Tumours were collected by five centres in the UK and Canada between 1977-2005 and almost all ER negative and lymph node positive patients received adjuvant chemotherapy, whereas ER negative and/or lymph node positive patient did not. No patients with HER2 overexpression received trastuzumab. Median follow-up was 141 months determined using the reverse Kaplan-Meier method. DNA and RNA were isolated from samples and hybridised to the Affymetrix SNP 6.0 and Illumina HT-12 v3 platforms for genomic and transcriptional profiling as described by Curtis et al. (2012) 27 . This cohort was used to assess the prognostic significance of DARPP-32 at the mRNA level and determine associations with other genes using artificial neural network analysis. www.nature.com/scientificreports www.nature.com/scientificreports/ immunohistochemistry. Immunohistochemistry was performed on tissue microarrays for the discovery cohort, validation cohort and the HER2 cohort, and were comprised of single 0.6 mm cores taken from a representative tumour area as assessed on Haematoxylin and Eosin stained sections by a specialist breast cancer histopathologist. Immunohostochemical staining was achieved using a Novolink Polymer Detection kit (Leica) according to the manufacturers' instructions, the use of which has been described previously [28][29][30] . In brief, xylene was used to deparaffinise tissue, followed by rehydration in ethanol then water. Antigen retrieval was achieved in 0.01molL −1 sodium citrate buffer (pH6.0), heated in a microwave for 10 minutes at 750 W followed by 10 minutes at 450 W. Tissue was treated with Novolink Peroxidase Block, washed with Tris-buffered saline (TBS), and then treated with Novolink Protein Block solution. Rabbit polyclonal anti-DARPP-32 (Abcam ab40801) diluted 1:500 was used as primary antibody and was incubated on tissue for one hour at room temperature. Antibody specificity was confirmed by Western blotting on breast cancer cell lysates prior to use. Tissue was washed with TBS prior to the application of Novolink Post Primary solution, which was subsequently washed with TBS and then Novolink Polymer solution was applied. Immunohistochemical reactions were developed using 3,3′ diaminobenzidine as the chromogenic substrate and tissue was counterstained with haematoxylin. Tissue was dehydrated in ethanol and fixed in xylene prior to mounting using DPX. Positive and negative controls were included with each staining run and were comprised of breast tumour composite sections comprising grade 1 and 2 early stage invasive tumour; negative controls had primary antibody omitted from each staining run (Supplementary Information).

Statistical analyses.
Slides were scanned using a Nanozoomer Digital Pathology Scanner (Hamamatsu Photonics) and staining was assessed at 200x magnification. Staining in the cytoplasm was assessed using a semi-quantitative immunohistochemical H score, where staining intensity within tumour cells was assessed as none (0), weak (1), medium (2) or strong (3) over the percentage area of each staining intensity. Staining in the nucleus was examined in a semi-quantitative manner, where the percentage of tumour cells that demonstrated any staining intensity was assessed. Greater than 30% of cores for each TMA were double assessed, with both assessors blinded to clinical outcome and each other's scores. Single measure intraclass correlation coefficients were above 0.7, indicating good concordance between scorers.
Statistical analysis was performed using IBM SPSS Statistics (version 24). Cases were stratified based on breast cancer specific survival for both the discovery and validation cohorts and the METABRIC cohort, and relapse-free survival for the HER2 positive cohort, using X-Tile software 31 . All differences were deemed statistically significant at the level of P ≤ 0.05. The Pearson χ 2 test of association was used to determine the relationship between categorised protein expression and clinicopathological variables. Survival curves were plotted according to the Kaplan-Meier method with significance determined using the log-rank test.
Identification of genes associated with DARPP-32 expression. A supervised artificial neural network was used to further understand the molecular function of PPP1R1B in breast cancer in the METABRIC www.nature.com/scientificreports www.nature.com/scientificreports/ series. PPP1B1R expression was used as the supervising variable as described by Abdel-Fatah et al. 32 . The artificial neural network was conducted with a constrained multi-layer perceptron architecture and sigmoidal transfer function, where weights were updated by a feed forward back propagation algorithm. Probes from the METABRIC data were ranked based on their root mean squared (RMS) error for predication of DARPP-32 expression as a continuous variable.

Results
DARPP-32 protein staining location and frequency. DARPP-32 protein expression was assessed in two large independent cohorts of early invasive breast cancer. Nuclear and cytoplasmic DARPP-32 expression was observed in all cohorts and varied from weak to intense, with heterogeneity observed between adjacent tumour cells. Representative photomicrographs are shown in Fig. 1. In the discovery cohort, 1352 patients were assessed, cytoplasmic DARPP-32 expression had a median H-score of 40, and ranged from 0 to 300. In the validation cohort, 1655 patients were assessed, cytoplasmic DARPP-32 expression had a median H score of 20 and ranged from 0 to 300. Nuclear DARPP-32 expression in the discovery cohort had a median H score of 0 and ranged www.nature.com/scientificreports www.nature.com/scientificreports/ from 0 to 100; in the validation cohort, the median DARPP-32 H-score was 5 and ranged from 0 to 100. X-tile was used to generate cut points for use in both cohorts based on breast cancer specific survival. In the discovery cohort, cytoplasmic DARPP-32 expression had a cut point of 35, with 49.6% of cases (670/1352) demonstrating low expression; in the validation cohort a cut point of 10 was used, with 43.6% of cases (722/1655) demonstrating low expression. In the discovery cohort, nuclear DARPP-32 expression had a cut point of 20, with 65.6% of cases (887/1352) demonstrating low expression; in the validation cohort, a cut point of 10 was used, with 51.5% of cases (853/1655) demonstrating low expression.

DARPP-32 expression in ER positive disease.
Low expression of DARPP-32 was particularly important in patients with ER positive disease. In the discovery cohort, low DARPP-32 cytoplasmic expression was significantly associated with adverse survival of ER positive patients (P < 0.001), but not ER negative patients (P = 0.099) (Fig. 2A). The same finding was observed when nuclear DARPP-32 expression was assessed, with low expression significantly associated with adverse survival of ER positive patients (P < 0.001), but not ER negative patients (P = 0.407) (Fig. 2B). In the validation cohort, low cytoplasmic expression of DARPP-32 associated with adverse survival of ER positive patients (P < 0.001), but not ER negative patients (P = 0.291) (Fig. 2C). Similar findings were for nuclear DARPP-32 expression were observed; with low expression associated with adverse survival of ER positive patients (P < 0.001), but not ER negative patients (P = 0.679) (Fig. 2D).

DARPP-32 expression in HER2 positive patients treated with trastuzumab. DARPP-32 was
assessed in a cohort of 112 HER2 positive breast cancer patients treated with adjuvant chemotherapy and trastuzumab. A similar DARPP-32 expression pattern was observed to that in early breast cancer. In this cohort, cytoplasmic DARPP-32 had a median H score of 142.9 and ranged from 0 to 300; nuclear DARPP-32 expression had a median H score of 45.36 and ranged from 0 to 100. X-tile was used to generate a cut point for analysis based on relapse free survival; cytoplasmic DARPP-32 expression had a cut point of 20, with 35.7% of cases (40/112) demonstrating low expression, nuclear DARPP-32 expression had a cut point of 5, with 34.8% of cases (39/112) demonstrating low expression.
No associations were observed between DARPP-32 expression and clinicopathological criteria in this cohort ( Table 2). Low nuclear DARPP-32 expression was significantly associated with adverse progression-free survival of HER2 positive breast cancer patients treated with trastuzumab (P = 0.031) (Fig. 3A); cytoplasmic DARPP-32 expression was not associated with progression-free survival (data not shown). Nuclear DARPP-32 expression did not remain associated with progression-free survival in multivariate Cox regression (HR = 0.387; 95% CI = 0.095-1.570; P = 0.184); when tumour size, lymph node status and NPI category were included (with individual Kaplan-Meier log rank significance of P = 0.012, P = 0.019 and P = 0.008 respectively). (2019) 9:16987 | https://doi.org/10.1038/s41598-019-53529-z www.nature.com/scientificreports www.nature.com/scientificreports/ PPP1R1B mRNA expression and patient survival. Three PPP1R1B probes were available to assess mRNA expression in the METABRIC cohort and expression was categorised using the median gene expression value to assess for association with patient survival and to perform expression profiling. PPP1R1B probe 1 (ILMN_1690096) is located within a coding area that corresponds to the N-terminal region of DARPP-32; PPP1R1B probe 2 (ILMN_1759012) and PPP1R1B probe 3 (ILMN_2304495) were both located in untranslated regions (5′ and 3′ respectively). Probe 1 and 3 are located in areas found in the sequence for DARPP-32 (NM_032192), and probe 2 and 3 are located in areas found in the sequence for t-DARPP (NM_181505.3). No association was observed between the expression of the probes and disease specific survival. The gene expression data was analysed using an artificial neural network approach that uses a machine learning based data mining algorithm. A rank order of all the genes was produced based on the minimum average root mean squared error. The top 200 transcripts were selected for PPP1R1B probe 1, 2 and 3 and 73 common transcripts were identified ( Table 3). The transcripts that were common to all three probes included CDC42, DKK1, GRB7, PNMT, and GPER amongst others.
All three PPP1R1B probes were further assessed in ER positive patients based on the observed DARPP-32 protein findings, with a cut point generated in this subgroup of patients using X-tile. PPP1R1B probe 1 expression was not associated with breast cancer specific survival of ER positive patients; however, low expression of PPP1R1B probe 2 and 3 were both associated with adverse survival of ER positive breast cancer patients (P = 0.041 and P = 0.002 respectively) (Fig. 3B,C).

Discussion
Low nuclear DARPP-32 expression was significantly associated with adverse survival of patients in two independent cohorts of patients treated at Nottingham University Hospitals (discovery cohort n = 1352 and validation cohort n = 1655). Furthermore, low cytoplasmic expression of DARPP-32 was associated with patient survival in the validation cohort. The epitope for the antibody used for immunohistochemistry is located within amino acids 0-30, meaning that only DARPP-32, and not t-DARPP expression was assessed in this study. There are limited reports of DARPP-32 expression and its association with patient survival in cancer. In oesophageal squamous cell carcinoma DARPP-32 is expressed after a phase of dysplasia, and low levels of DARPP-32 are associated with tumours that progress rapidly 33 . In colorectal cancer, lower expression of DARPP-32 is associated with improved overall survival and disease free survival 12 , and in non-small cell lung cancer high relative t-DARPP www.nature.com/scientificreports www.nature.com/scientificreports/ (in comparison to DARPP-32) is associated with tumour stage 34 . The expression of DARPP-32 in 230 breast cancer patients has been examined previously using N-terminal and C-terminal specific antibodies, to show that high N-terminal DARPP-32 expression is associated with adverse patient survival; this differs from the results presented here 21 . It is unclear why opposing results were observed; however, patient demographics, including clinicopathological variables, and patient treatment are not available in the previously published study, so cannot be directly compared with the current findings.
Interestingly, low DARPP-32 protein expression was particularly important in ER positive patients, in both cohorts of patients; there was also a strong association between DARPP-32 expression and ER receptor status. In addition to ER status, strong consistent associations were observed between DARPP-32 expression and PgR status and triple receptor negative disease in both patient cohorts.
There is accumulating evidence that DARPP-32, in particular t-DARPP, plays a role in response to trastuzumab; this study shows that low nuclear expression of DARPP-32 is significantly associated with adverse progression free survival in 112 HER2 positive patients treated with trastuzumab and adjuvant chemotherapy. This is agreement with published in vitro data, that indicates expression of DARPP-32 and t-DARPP expression in HER2 positive breast cancer is involved in resistance to trastuzumab 22 . Expression of t-DARPP has been shown to activate IGF-1R signalling in trastuzumab resistant breast cancer cells through increased glycolytic capacity 35 . In addition, DARPP-32 mRNA and protein levels have also been shown to fall in HER2 targeted agent, lapatinib, resistant breast cancer cell lines, with a t-DARPP mediated survival advantage observed 24 .
In models of murine tumourigenesis DARPP-32 expression is expressed in normal mouse tissue and some breast tumours, with t-DARPP expressed only within tumours 18 . It is interesting to hypothesise that the association between patient survival and low expression of DARPP-32 is observed with a corresponding shift to increased t-DARPP expression; this will be determined in future studies.
PPP1R1B expression was assessed in the METABRIC cohort, where three probes were available for assessment, probe 1 and 3 located in DARPP-32 and probe 2 and 3 located in t-DARPP. Artificial neural network analysis identified a number of genes associated with PPP1R1B expression. Artificial neural network analysis was performed using all three PPP1R1B probes and commonalities within the top 200 genes for each probe identified. Validation of these associations will be performed as part of future studies. Interestingly, CDC42 was identified from this analysis; cdc42 plays a role in filopodia formation and breast cancer cells expressing DARPP-32 have, in a study looking at Wnt-5A activation of DARPP-32, been shown to have lower cdc-42 activity 17 . In addition to www.nature.com/scientificreports www.nature.com/scientificreports/ CDC42, Dickkopf-1 (DKK1) was also identified and functions a wnt-5A pathway inhibitor. Furthermore, artificial neural network analysis identified an association between PPP1R1B with both GRB7 and PNMT, a link between PPP1R1B and these genes has already been described in upper gastrointestinal adenocarcinomas where DNA amplification at 17q is often detected (containing PPP1R1B, STARD3, TCAP, PNMT, PERLD1, ERBB2, C17orf37, and GRB7) 36 .

conclusion
This study demonstrates that low DARPP-32 protein expression is associated with shorter survival in two large, independent, early stage invasive breast cancer patient cohorts, with a stronger association observed in ER positive disease. This finding was also observed at the mRNA level, with low PPP1R1B expression significantly associated with shorter survival of ER positive patients in the METABRIC cohort. Furthermore, low DARPP-32 expression was associated with shorter progression-free survival of HER2 positive patients treated with  Table 3. Genes associated with two or more PPP1R1B probes in the METABRIC dataset identified by artificial neural network analysis. Those highlighted in bold were associated with expression of all three PPP1R1B probes.