Formalin-fixed paraffin-embedded (FFPE) samples (i.e. preoperative CNBs and surgical tumor specimens) were retrieved from the Department of Clinical Pathology at Sahlgrenska University Hospital (Gothenburg, Sweden) for 98 consecutive patients with primary breast cancer diagnosed between January and May 2017. Two patients had bilateral breast cancer. In total, 100 matching samples were examined. The age of the FFPE samples was 18–22 months. Clinical data on adjuvant treatment based on IHC, axillary node status, tumor size, tumor grade, patient age at the time of diagnosis and co-morbidities were retrieved from electronic medical records. Patients with neoadjuvant treatment, tumor size less than 5 mm, multifocal/multicentric breast cancer, bifocal breast cancer and previous surgery for breast cancer were not included. Patients treated with neoadjuvant treatment were excluded to avoid the possible loss of available matching surgical samples for assessment due to partial or complete reduction of the tumor area following therapy response. Another exclusion criteria was FFPE specimens that were fixed in fixatives other than 10% Neutral Buffered Formalin (NBF), fixed in NBF for < 6 or > 72 h, or aged more than 4 weeks since initial preparation/fixation, but this was not relevant in this study. The clinicopathologic features of the 100 specimens are shown in Table 1.
Immunohistochemistry (IHC) and cutoffs for biomarker analysis in CNB and surgical specimen
ER, PR, HER2 and Ki-67 protein expression data (IHC) were retrospectively collected for both the CNB and surgical specimens from the Sympathy database (Sahlgrenska University Hospital, Department of Clinical Pathology) and no re-assessments were conducted. As the CNB sample is only a random sampling of the tumor, IHC assessment is routinely performed on both the CNB and the matching surgical specimen at our institution. Routine IHC staining protocols at the Department of Clinical Pathology used the following antibodies to assess ER, PR, HER2 and Ki-67 expression: rabbit anti-ERα (Agilent Dako IR084, clone EP1), mouse anti-PR (Agilent Dako IR068, clone 636), mouse anti-KI-67 (Agilent Dako IR626, clone MIB-1), and HercepTest (Agilent Dako SK001, clone poly), respectively. In brief, FFPE sections were pretreated using the Dako PTLink system (Agilent Dako) and processed on an automated Dako Autostainer platform, as previously described . The Ventana dual SISH test was also performed, if necessary to resolve final HER2 status. International guidelines were used to determine biomarker status, where ER, PR, and Ki-67 were considered to be positive with ≥1%, ≥1%, and ≥ 20% immunostaining in neoplastic cells, respectively [1, 6]. HercepTest scored 2+ and 3+ was followed by SISH testing, and considered positive with confirmed HER2 amplification. SISH testing was not performed on the surgical specimen, if the matching CNB was HER2-amplified. New SISH tests were performed on the surgical specimen, for non HER2-amplified CNBs with HercepTest scored as 2+.
STRAT4 analysis was performed in November 2018 and was limited to two operators (S.J. and S.D.L.), each blinded to the IHC status of individual biomarkers. For each patient, a single 4 μm or 5 μm FFPE section (adjacent to the section used for hematoxylin and eosin (H&E) staining, determined by a board-certified pathologist (A.K.)) was prepared for both the CNB and the matching surgical specimen. For some surgical specimens, macrodissection (tumor area outlined by the pathologist) was required to remove non-neoplastic tissue. The FFPE sections were first placed in a water bath at 40 °C, mounted onto positively charged FLEX IHC microscope slides (Agilent Dako) and dried overnight at room temperature. The FFPE section was then scraped from the slide and transferred to a 1.5 ml lysis tube. FFPE samples containing CNBs were sectioned and transferred directly to lysis tubes.
Tissue lysis was prepared by adding 1.2 ml Xpert® FFPE lysis kit reagent and 20 μl proteinase K (PK) to each sample, followed by incubation at 80 °C for 30 min. The tissue lysate was then transferred to a provided 5 ml sample vial and precipitated with 1.2 ml 95% ethanol. An aliquot containing 520 μl tissue lysate was transferred to the Xpert Breast Cancer STRAT4 cartridge, the cartridge lid was closed, and then the closed cartridge was placed into the GeneXpert instrument. Furthermore, nucleic acid purification, and RT-qPCR amplification and detection are fully automated in the GeneXpert instrument, with all the reagents required for the different stages preloaded in the cartridge. The GeneXpert instrument system simultaneously measured the cycle threshold (Ct) values in multiplex for the reference gene (CYFIP1) and target genes (ESR1, PGR, MKi67, and ERBB2), and then calculated the delta cycle threshold (dCt) values (dCt = reference gene Ct minus target gene Ct) for each marker. The CYFIP1 Ct and individual dCt values for each marker were then compared to pre-specified Ct and dCt cutoffs to classify ESR1, PGR, MKi67, and ERBB2 mRNA expression as POSITIVE, NEGATIVE, INDETERMINATE, INVALID, or ERROR. Samples classified as INDETERMINATE (only applicable for PGR and/or MKi67 results, where the dCt value is below the specified cutoff value and the CYFIP1 reference gene Ct is greater than 31) or INVALID (CYFIP1 Ct > 35) were repeated using a more concentrated lysate (i.e. 260 μl FFPE lysis reagent, 5 μl PK, and 260 μl ethanol), whereas samples classified as ERROR were repeated using the remaining sample lysate.
Clinical evaluation of surrogate subtyping
For each surgical specimen, the IHC and STRAT4 analyses were then used to determine the surrogate breast cancer subtype (i.e. Luminal A-like, Luminal B-like (HER2- or HER2+), HER2+ (non-luminal), and Triple-negative) and recommended treatment for breast cancer patients according to national guidelines . In brief, the Ki-67 cut-off was set to 20% to differentiate Luminal A-like breast cancer (Ki-67 < 20%) from Luminal B-like HER2- (Ki-67 ≥ 20%) and Luminal B-like HER2+ tumors (any Ki-67) . Patients with subsequent changes in surrogate subtyping according to the STRAT4 assay results for ESR1, PGR, ERBB2, and MKi67 gene expression were further evaluated by the medical and surgical oncologists (K.L. and S.J.), whom were blinded to the administered treatment for each patient.
Statistical analyses were performed using a 0.05 p-value cutoff for statistical significance in MedCalc Statistical Software (version 18.10.2) or R/Bioconductor (version 3.6.0). To assess the concordance between STRAT4 and the reference method(s) (IHC or SISH, as applicable), standard 2 × 2 cross-tables were utilized along with calculation of the overall percent agreement (OPA), positive percent agreement (PPA) negative percent agreement (NPA) positive predictive value (PPV), and negative predictive value (NPV). A kappa statistic with 95% two-sided confidence intervals (CI) was calculated to estimate the overall agreement between the compared methods, with k-values > 0.8, between 0.6 and 0.8, 0.4 and 0.6, < 0.4, and < 0.2 classified as very good, good, moderate, fair, and poor agreement, respectively . Scatterplots were constructed using the ggplot2 package (version 3.3.0) in R .