Supplemental Materials
Phase II neoadjuvant clinical trial of carboplatin and eribulin in women with triple negative early stage breast cancer (NCT01372579)
Dose modifications
Dose delays and reductions were performed for any non-hematologic toxicity of ≥ grade 3 (or toxicities that were ≥ grade 2 that were medically concerning or unacceptable to the patient) on the planned day of therapy. If the toxicity was felt to be related to carboplatin, the dose was held until recovery to grade 1 and then resumed at an AUC of 5. If a second occurrence of ≥ grade 3 toxicity occurred, the dose was held until recovery to grade 1 and then resumed at an AUC of 4. If the toxicity was felt to be related to eribulin, the dose was held until recovery to grade 1 and then resumed at 1.1 mg/m2. If a second occurrence of ≥ grade 3 toxicity occurred, the dose was held until recovery to grade 1 and then resumed at 0.8 mg/m2. A third dose reduction of the same drug was not allowed.
RCB assessment
Briefly, Symmans and his group base calculations of the residual tumor burden on a number of pathology variables that are derived from the gross description of the resection specimen and the routine hematoxylin and eosin (HE) slides generated from this resection specimen after neoadjuvant therapy. These variables are as follows: 1) the cross sectional dimensions of the residual tumor bed (d1 and d2), 2) the overall residual invasive tumor cellularity (as a percentage of the area of the tumor bed), 3) the percentage of the residual cancer that is in situ disease, 4) the number of positive lymph nodes and 5) the diameter of the largest nodal metastasis. The residual cancer burden (RCB) [1] is then calculated and the patient is assigned into one of four RCB classes (pCR, RCB-I, RCB II and RCB III).
HRD assay
A 5 micron H&E stained slide was reviewed by a pathologist to facilitate enrichment of tumor derived DNA. Regions of highest tumor density were scraped from consecutive 10 micron sections. A Maxwell 16 FFPE Plus LEV DNA purification kit (Promega, Madison, WI) was used to isolate DNA from the macrodissected tissue. 50–200 ng of genomic DNA was sheared on a Covaris E220, and the resulting DNA was end-repaired and blunt end ligation was use to ligate adapters containing molecular barcodes (indices). Individual adaptor-ligated libraries were then amplified and pooled to create library pools derived from 16 samples. Each library pool was hybridized for 22–24 hours at 65oC with a custom SureSelect XT2 Target Enrichment panel (Agilent, Santa Clara, CA) that included probes targeting 54,091 SNPs distributed across the genome, and an additional 685 probes targeting the coding regions of the BRCA1 and BRCA2 genes. Following hybridization the target enriched library was amplified, quantitated, diluted and then sequenced on an Illumina HiSeq2500 running in Rapid mode.
Sequence reads were trimmed for quality and aligned to expected wild type sequence from BRCA1/2 and 400 bp sequence windows flanking each targeted SNP. For non-artifactual BRCA1/2 variant detection, any difference between the sequence and wild type reference represent variants. Variant classification was performed based on previously described criteria [2].
Sequences covering SNP positions were used to generate allelic imbalance profiles. A hidden Markov model (HMM) was used to define regions and the corresponding break points within allelic imbalance profiles. TAI score is defined as the number of regions with allelic imbalance that extend to one of the subtelomeres but do not cross the centromere [3]. LST score is defined as the number of break points between regions longer than 10 Mb after filtering out regions shorter than 3 Mb [4]. Allele specific copy number (ASCN) for each of the regions was determined by an algorithm similar to that described in Popova et al [5]. ASCN was used to calculate LOH score defined as the number of subchromosomal LOH regions longer than 15 Mb [6]. Finally, the HRD score is the sum of the LOH, TAI and LST scores.
An additional 100ng genomic DNA was subjected to bisulfite conversion using the EpiTect Bisulfite kit (Qiagen). PCR primers specific for bisulfite converted DNA were used to amplify the promoter region of exon 1A of the 5’ untranslated region of the BRCA1 gene. The resulting PCR products were sequenced using a MiSeq (Illumina, San Diego, CA) sequencer. Sequence reads were aligned to the BRCA1 promoter amplicon and the called base at ten CpG sites was inspected. The number of CpG site bases with a “C” nucleotide was counted; this represents the number of methylated CpG sites. A methylation score for the sample was computed as the proportion of methylated reads relative to the total number of reads that were either methylated or not methylated. Samples with methylation score >10% were classified as methylated.
Protein expression measurement and evaluation
The expression of Ki67, p53, AR, Cyclin E, CDK2, Cyclin D, CDK4, Pin1 and Smad3 were assessed in breast cancer samples prior to treatment. Post-treatment expression of Cyclin E, CDK2, Cyclin D, CDK4, Pin1 and Smad3 were examined in patients with residual tumor at the time of surgery. All of the above markers were validated according to protocol standardization at NU PathCore. Appropriate known control tissue was used for positive control and primary Abs were omitted in negative controls. All samples were scored by an independent pathologist (Dr Siziopikou).
Resected breast cancer tissue was promptly fixed using 20% alcoholic formalin and then with 10% normal buffered formalin solution for a minimum of 3 hours. Representative tissue was embedded at 56°C in paraffin blocks. Consecutive sections were cut at 4 μm. First and last sections from each block were analyzed through HE stain for tumor existence. Unstained slides were stored at 4°C before IHC. Sections were placed in a 58–60°C oven overnight for tissue to adhere. The sections were de-waxed in xylene, rehydrated through graded alcohols to water, washed with PBS, pretreated with citrate buffer [0.01 mol/L citric acid (pH 6.0)] and pressure cooked for 30 seconds at 125°C and gradually reduced to 90°C for an overall time of 40 minutes and then cooled down at room temperature for 20 minutes for antigen retrieval. Antigen unmasking and endogenous peroxidase activity was quenched by incubation with 3% hydrogen peroxide in 50% methanol for 10 min at room temperature by automated stainer. After nonspecific binding sites were blocked, slides were exposed to 10% normal goat serum in PBS for 20 min at 37°C.
Ki67 and p53 were assayed according to standard protocols using the Ventana 30-9 and Ventana Bp53-11 respectively. The other markers were performed at Pathology Core Facility, Robert H. Lurie Comprehensive Cancer Center, Northwestern University, Chicago, IL (NU PathCore) using the following antibodies: PIN1, SC-46660 (Santa Cruz, CA) at 1:800 dilution; Smad-3, ab28379 (Abcam, England) at 1:200 dilution, Cyclin-E, NCL-CYCLIN E (Novacastra (Leica), IL) at 1:25 dilution; Cyclin D AC-0017 (Epitomics, CA) at 1:100 dilution; CDK2 SC-163 (Santa Cruz, CA) at 1:1000 dilution; CDK4 1279 (Cell Signaling, MA) at 1:50 dilution; AR, M3562 (Dako, CA) at 1:50 dilution.
The expression of Ki67, p53, AR, Cyclin E, CDK2, Cyclin D, CDK4, Pin1 and Smad3 was analyzed by light microscopy in invasive breast cancer regions. Specific evaluation criteria were used for respective markers, including the following (nuclear and cytoplasmic staining in invasive tumor cells): Ki67: ≤10%=low, 11-20%= intermediate, >20%= high; p53, AR, Cyclin E, CDK2, CDK4 staining score: 0-9%= negative, ≥10%= positive; Cyclin D, Pin1 and Smad3 scoring was performed as follows: negative= < 1% staining, 1-9%=low positive and 10% as positive.
Statistical methods
All analyses were conducted using R version 3.0.2 (R Foundation for Statistical Computing, Vienna, Austria, http://R-project.org). All reported P values were two-sided. P values less than 0.05 were considered to indicate statistically significant results. No adjustments were made for multiple comparisons.
Patient outcomes
Patient outcomes following the investigational regimen were summarized in terms of pathological complete response (pCR), clinical complete response (cCR) and progression-free survival (PFS). Confidence intervals for the proportion of patients who achieved pCR, and cCR, were calculated according to the exact two-sided binomial test using the method of Blaker [7]. PFS was estimated by the Kaplan-Meier method.
Predictors of pCR
Pretreatment clinical factors, tumor biomarkers based on genomic profiles, and biomarkers based on protein expression profiles were tested for associations with pCR in univariate logistic regression models. Logistic regression P values were based on chi square statistics from likelihood ratio tests. For categorical variables exhibiting complete or quasi-complete separation, Firth’s penalized likelihood [8] was used in place of the conventional likelihood.
Clinical records were obtained for patient age at diagnosis, clinical stage at diagnosis, and tumor grade. Age (in years) was modeled as a quantitative variable. Clinical stage at diagnosis (I, II, III) and tumor grade (1, 2, 3) were coded as categorical variables.
Biomarkers based on genomic profiles included the HRD score, HR status and BRCA1/2 mutation status. The HRD score was tested as a numeric covariate. HR deficiency status (deficient, non-deficient), and BRCA1/2 mutation status (mutant, wild type) were analyzed as binary factors.
Nuclear and cytoplasmic protein expression levels of pretreatment tumor specimens were modeled as categorical variables. Nuclear expression variables included Ki67 (low, intermediate, high), p53 (negative, positive), and AR, Cyclin E, CDK2, Cyclin D, CDK4, Smad3 and Pin1 (each coded as negative, low positive, positive). Cytoplasmic expression variables included CDK2, Cyclin D, CDK4 and Smad3 (each modeled as negative, positive).
Pre- versus post-treatment protein expression
In the subset of patients with residual tumor burden following the investigational treatment regimen, protein expression levels of surgically resected tumor specimens were compared to those measured prior to treatment. The concordance of pre- and post-treatment expression levels was quantified in terms of percent agreement. The paired sample Wilcoxon signed rank test was used to evaluate differences in expression.
Reference List
1. Symmans WF, Peintinger F, Hatzis C, Rajan R, Kuerer H, Valero V, Assad L, Poniecka A, Hennessy B, Green M, Buzdar AU, Singletary SE, Hortobagyi GN, Pusztai L (2007) Measurement of residual breast cancer burden to predict survival after neoadjuvant chemotherapy. JClinOncol 25 (28):4414-4422
2. Beaudet AL, Tsui LC (1993) A suggested nomenclature for designating mutations. HumMutat 2 (4):245-248
3. Birkbak NJ, Wang ZC, Kim JY, Eklund AC, Li Q, Tian R, Bowman-Colin C, Li Y, Greene-Colozzi A, Iglehart JD, Tung N, Ryan PD, Garber JE, Silver DP, Szallasi Z, Richardson AL (2012) Telomeric allelic imbalance indicates defective DNA repair and sensitivity to DNA-damaging agents. Cancer Discov 2 (4):366-375
4. Popova T, Manie E, Rieunier G, Caux-Moncoutier V, Tirapo C, Dubois T, Delattre O, Sigal-Zafrani B, Bollet M, Longy M, Houdayer C, Sastre-Garau X, Vincent-Salomon A, Stoppa-Lyonnet D, Stern MH (2012) Ploidy and large-scale genomic instability consistently identify basal-like breast carcinomas with BRCA1/2 inactivation. Cancer Res 72 (21):5454-5462
5. Popova T, Manie E, Stoppa-Lyonnet D, Rigaill G, Barillot E, Stern MH (2009) Genome Alteration Print (GAP): a tool to visualize and mine complex cancer genomic profiles obtained by SNP arrays. Genome Biol 10 (11):R128
6. Abkevich V, Timms KM, Hennessy BT, Potter J, Carey MS, Meyer LA, Smith-McCune K, Broaddus R, Lu KH, Chen J, Tran TV, Williams D, Iliev D, Jammulapati S, FitzGerald LM, Krivak T, DeLoia JA, Gutin A, Mills GB, Lanchbury JS (2012) Patterns of genomic loss of heterozygosity predict homologous recombination repair defects in epithelial ovarian cancer. BrJCancer 107 (10):1776-1782
7. Blaker H (2000) Confidence curves and improved exact confidence intervals for discrete distributions. Canadian Journal of Statistics 28 (4):783-798. doi:10.2307/3315916
8. Firth D (1993) Bias reduction of maximum likelihood estimates. Biometrika 80 (1):27-38. doi:10.1093/biomet/80.1.27