The study population comprises women affected by invasive breast carcinoma, treated by adjuvant or neoadjuvant chemotherapy, and with archived tissue specimens of the primary tumor. Tissue samples (standard archived, formalin-fixed, paraffin-embedded tissue) and clinical data were retrospectively collected at the Institute of Pathology and at the Breast Unit of our center, respectively.
In this study, all breast cancer cases treated by adjuvant or neoadjuvant chemotherapy between January 2002 and December 2016 were included in order to have at least 12 months of follow-up. All cases were included where sufficient tumor tissue was stored in paraffin blocks at the time of surgery. Consequently, all cases of complete pathological response were excluded, as the quantity of tissue from preoperative biopsy did not provide enough material for the current study. Also, the cases of poorly conserved or quantitatively insufficient tissue were excluded. Furthermore, all of the following cases were excluded from the study: those without a documented follow-up, breast cancer patients who did not undergo adjuvant or neoadjuvant treatments, and male patients.
This retrospective observational study reviews pathological archives and medical records to identify breast cancer cases treated by adjuvant or neoadjuvant chemotherapy. For the purpose of the study, two different types of sample selection from the original cohort were performed: case-cohort study; and case-control study.
The case-cohort study design model is a sampling methodology to randomly select a sample (called subcohort) from an assembled epidemiologic cohort and to use this subcohort as a comparison group for the selected cases that occur in the cohort (in this particular case, the group of breast cancer patients that relapsed within 12 months of follow up) . This kind of study is particularly suitable for very numerous cohorts, where it is cost-prohibitive to provide a complete follow-up either for disease outcomes or for specific information on the whole cohort [28,29,30].
Of the 4504 women in the cohort treated for breast pathology, 894 were eligible for the present study because of chemotherapy treatment and fulfilling the required inclusion criteria. The study sample was selected from the latter 894 women. It was composed of a random group of 163 of the 894 eligible women (hereafter called the subcohort) together with all eligible women diagnosed with a breast cancer recurrence within 12 months of treatment initiation. We considered all recurrences (loco-regional recurrences or distant metastases) and adverse events (breast cancer-related death) occurring between the baseline attendance and April 30, 2017.
The final sample included 203 women: 163 belonging to the subcohort and 42 breast cancer patients who experienced recurrence within a 12-month follow-up. We factored in that about 35% of samples could potentially be unavailable for the assessment in the TMA. Power and sample size considerations were based on Cai and Zeng’s approach, and the study had > 80% power to detect HRs of 3.0, assuming alfa = 0.05 .
In the case-control sample selection, all breast cancer patients included in the tissue micro-array were considered. Cases and controls were randomly selected from the patients mentioned above.
The number of the analyzed samples was established from the maximum available resources to test PITX2 methylation. Therefore, the total number of cases was 13, and the total number of controls was similarly 13.
Information collected included some patient characteristics, such as the age at diagnosis, the body mass index (BMI), any familial history for breast or ovarian carcinoma, the current fertility status (pre-or post-menopausal), any use of estro-progestin therapies (with contraceptive intent in the pre-menopause or as hormone replacement therapy in the post-menopause). Tumor characteristics considered were the histotype, grading, expression of estrogen receptor (ER), progesterone receptor (PR), HER2/neu and Mib1/Ki-67, as well as any presence of multifocality/multicentricity, perivascular invasion (PVI), peritumoral lymphocytic infiltration, nodal extracapsular invasion or bunched axillary nodes ). Surgical and non-surgical treatments were also taken into account for data elaboration.
Pathological specimens were routinely assessed following the European guidelines [33, 34]. In particular, samples sized 30 mm or less were wholly sliced and evaluated, whereas specimens sized over 30 mm underwent sampling based on the European guidelines [33, 34].
The World Health Organization criteria were used to determine the histology  and nodal status (TNM classification VII ed. AJCC/UICC, 2009) . AFIP and Elston Ellis’s recommendations were considered while assessing the grading in ductal carcinoma in situ and invasive carcinoma, respectively [37, 38]. Peritumoral lymphocytic infiltration, PVI, multifocality/multicentricity, and nodal status were determined as described in previous studies [39, 40].
Estrogen Receptor (ER), Progesterone Receptor (PR), Mib1/Ki-67, and HER2/Neu expression were evaluated by immunohistochemistry. We defined positive ER or PR where positivity was ≥1% in any nuclear staining. In addition, HER2/Neu was defined as overexpressed when staining 3+ or 2+ with FISH amplification and negative if the value was 0, 1+ or 2+ without FISH amplification. Through the combination of ER, PR, HER2/Neu, and Mib1/Ki-67, all invasive breast cancers were classified in the following molecular subtypes as previously described: luminal A, luminal B, luminal HER2, HER2-enriched, and basal-like .
Immunohistochemistry and molecular biology analyses
In this study, the following analyses were performed in both the case-cohort and the case-control subject selection. We analyzed the presence and the quantity of mRNA and the relative protein synthesis in the selected breast cancer tumor samples. In addition (solely for subjects selected for the case-control study), the Therascreen PITX2 RGQ PCR was used to test the PITX2 methylation.
Real-time PCR (RT-PCR) to quantify mRNA of TFEB, CARM1, and SIRT1
To perform real-time quantitative PCR, the primers were prepared for TFEB, CARM1, and SIRT1 (Supplemental Table 1). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a housekeeping protein (Supplemental Table 1). The mRNA was extracted from formalin-fixed paraffin-embedded tumor tissue samples by manually micro-dissecting the tumor area that had been histologically marked by a pathologist and using an RNeasy kit (Qiagen®, Venlo, Netherlands) that was used following the manufacturer’s instructions. RNA quantity and purity were measured using the Qubit 2.0 spectrophotometer (Invitrogen®, Carlsbad, CA), and RNA integrity was quantified using the RIN (RNA integrity number) assessed by Agilent 2100 Bioanalyzer (Agilent Technologies®, Santa Clara, CA). From this m-RNA, employing retrotranscription (SuperScript®, III REV transcript; Life Technologies), cDNA was obtained and quantified. Quantitative real-time PCR analysis was performed in triplicate by three independent experiments, using the LightCycler® 480 (Roche®) and LC SYBR Green I Master (Roche®), according to the manufacturers’ instructions. Quantitative data were collected as cycle threshold (CT) values considering the means of the triplicate runs. Finally, quantitative ΔCT expression values (2(−ΔCT)) were calculated.
Immunohistochemical analysis (IHC)
Along with the RT-PCR analysis, TFEB, CARM1, and SIRT1 protein expressions for all breast cancer cases in our samples were investigated by immunohistochemistry. Beclin-1 was investigated by immunohistochemistry. Cases were evaluated with respect to both staining percentage and intensity.
Tissue Micro Array (TMA)
The preparation and analysis of the TMA were carried out as previously described [41,42,43,44,45,46]. Having identified formalin-fixed, paraffin-embedded tumor tissue corresponding with our sample of patients, the hematoxylin-eosin-colored sections were analyzed. Then the tissue core samplings for the TMA were performed, taking care to include neoplastic tissue (two core biopsies per primary tumor). The receiver blocks were assembled. From these, 4-μm cross-sections were obtained, which were stained with hematoxylin-eosin. At a later stage, additional 4-μm cross-sections were obtained to prepare slides for immunohistochemical staining and subsequent analysis.
Immunohistochemical staining was performed according to standard protocol and manufacturer instructions. For antigen retrieval, the slides were heated, after deparaffinization, for 20 min at 98 °C in Target Retrieval Solution (low pH; Dako K8005, Glostrup, DK) with PT-link (Dako), and endogenous peroxidase activity was blocked with H2O2 for 10 min at environmental temperature. Slides were rinsed in PBS and then incubated with the following primary antibodies for 1 h at room temperature: TFEB (OriGene Technologies Inc., diluted 1:100, Rockville, MD, USA); CARM1 (OriGene Technologies Inc., diluted 1:100, Rockville, MD, USA); SIRT1 (OriGene Technologies Inc., diluted 1:200, Rockville, MD, USA); and Beclin-1 (Abcam plc., diluted 1:100, Cambridge, UK). A Dako REAL™ EnVision™ Dako Rabbit/Mouse (Dako, K5007, Glostrup, DK) was used as a second antibody. HRP activity was identified utilizing Dako REAL™ DAB+Chromogen (Dako, K5007, Glostrup, DK) as the substrate for 3 min in accordance with the manufacturer’s instructions. Tissue sections were then counterstained with hematoxylin. Negative controls included sections incubated with non-immune rabbit serum instead of the primary antibody. The immunohistochemical staining was evaluated independently by two pathologists in terms of H-score (the product of the actual percentage of positive-stained cells and intensity score, evaluated as strong 3, moderate 2, and weak 1, giving a possible range of 0–300).
PITX2 DNA methylation assay
The PITX2 promoter methylation was evaluated for DNA extracted from formalin-fixed paraffin-embedded sections (QIAmp DNA mini kit-Qiagen). Specifically, DNA was extracted from each primary tumor tissue sample kept at room temperature until the extraction time. Therascreen® PITX2 RGQ PCR kit is a methylation-specific PCR (MSP) based on real-time PCR, intended for the quantitative assessment of percent methylation ratio (PMR) in the promoter 2 (P2) of the PITX2 gene and it was validated in primary formalin-fixed paraffin-embedded breast cancer tissue [23, 47, 48]. PMR was calculated according to manufacturer instructions.
Genomic DNA was extracted from the samples using the QIAmp DNA FFPE Tissue Kit (QIAGEN Inc.). DNA was quantified using a Qubit dsDNA analysis kit for QBIT 2.0 fluorimeter (Thermo Fisher Scientific). The bisulfite conversion of the DNA was performed using the EpiTect Plus DNA Bisulfite Kit (QIAGEN Inc.), and the methylated DNA was then purified using the purification module reagents provided by the same kit and quantified at QBIT 2.0. After bisulfite conversion, the PMR of 3 CpG motifs of the PITX2 gene P2 was quantified by MSP using the Therascreen® PITX2 RGQ PCR kit, this containing a quantitative RT-PCR reaction mix, a primer, probes, and positive and negative controls .
The quantitative real-time PCR reaction was performed using the Rotor-Gene Q MDx real-time PCR platform (QIAGEN, Inc.) and evaluated by QIAGEN Rotor-Gene AssayManager® (Version 2.1.0) software with Therascreen PITX2 FFPE (C) analysis plugin for analysis and quality control .
During all tissue sample analyses, the operators were blinded to the clinical data of the patients. All primary formalin-fixed paraffin-embedded breast cancer tissue samples were coded to ensure blinding of the operator while conducting the PITX2 DNA methylation assay.
Data were analyzed through R (version 3.6.1; R Core Team (2019); R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria; URL https://www.R-project.org/) and considering p < 0.05 as significant. The normality of the distribution was tested with the Kolmogorov-Smirnov test. Univariate analysis was performed by the Wilcoxon or t-test in cases of continuous variables, or the Fisher exact test or chi-square test in cases of categorical variables. Univariate and multivariate survival analyses were also performed by Kaplan-Meier curves, Log-rank test, and Cox proportional hazards regression models. OS was considered to be the main outcome. Besides, in the multivariate model, all selected factors and their interactions were accommodated in a single analysis, except when the interaction term was non-significant (in which case we analyzed the non-interaction model). Correlations were tested by Spearman Rho and the relative p-value.