Hans-Peter Adams

Transcriptional census of 36 microdissected colorectal cancers yields a gene signature to distinguish UICC II and III

UICC stage II and III colorectal cancers (CRC) differ fundamentally in prognosis and therapeutic concepts. To analyze differential gene expression between both stages and to establish a relationship between molecular background and clinical presentation, tumor material from 36 unselected consecutive patients presenting with sporadic CRC, 18 UICC stage II and 18 UICC stage III, were laser microdissected to separate epithelial tumor cells. Gene expression levels were measured using U133A Affymetrix gene arrays. Twelve CRC associated signal transduction pathways as well as all 22,000 probe sets were screened for differential gene expression. We identified a signature consisting of 45 probe sets that allowed discrimination between UICC stage II and stage III with a rate of correct classification of about 80%. The most distinctive elements in this signature were the gene GSTP‐binding elongation factor (GSPT2) and the transcription factor HOXA9. Differential expression of these genes was confirmed by quantitative real‐time polymerase chain reaction (p(HOXA9) = 0.04, p(GSTP2) = 0.02). Despite the reliability of the presented data, there was no substantial differential expression of genes in cancer‐related pathways. However, the comparison with recently published data corroborates the 45 gene signature showing structural agreement in the direction of fold changes of gene expression levels for our set of genes chosen to discriminate between both stages.

Stratification of patients with colorectal cancer of UICC stage II using prognostic mutations signatures obtained by deep amplicon sequencing of cancer genes.

377 Background: Assignment of patients (pts) with UICC-stage II colorectal cancer (CRC) to adjuvant therapy remains controversial. The clinical utility of histo-pathological parameters (pT4, L+, V+) as well as tumor RNA expression signatures like Oncotype Dx Colon Cancer Assay, ColoPrint, or Predictor C is low given their positive predictive value (PPV) of 0.13, 0.22, 0.19, and 0.33, respectively, when adjusted to an incidence of 10% for occurrence of metastatic disease (mets) within 3 years after diagnosis. Methods: We applied deep amplicon sequencing of 48 well-known cancer genes to DNA samples of primary tumors from 173 pts with UICC-stage II CRC using the Illumina MiSeq technology. Patients were selected from a prospective, multicenter clinical diagnostic study named MSKK. More than 6,500 patients with CRC have been recruited into this study conducted by 39 hospitals in Germany. 79 of the 173 pts had progression of disease events within 3 years after R0 resection including 40 pts with mets, 12 pts wit...

A RNA signature with high sensitivity and specificity discriminating between responder and nonresponder to cetuximab monotherapy in colorectal cancer.

e14049 Background: Patients (pts) with advanced, metastatic colorectal cancer (CRC) and wild-type KRAS mutation status benefit from anti EGFR antibodies cetuximab (CE) and panitumumab. Pts with CRC of UICC stage III do not benefit from CE plus FOLFOX irrespectively of the KRAS mutation status. There is a high medical need for new biomarkers that can differentiate between responder (R) and non-responder (NR) to CE therapy in the adjuvant setting. METHODS We established a panel of 148 stable, passagable CRC xenografts from 240 primary CRC tumors of all four stages (Becker et al. EORTC-NCI-AACR meeting, Berlin Nov. 2010). All xenografts were characterized by histological and molecular methods including Affymetrix array profiling and KRAS, BRAF, PIK3CA mutation analysis. Pharmacological experiments were carried out with 68 models (8 stage I, 22 stage II, 29 stage III, 9 stage IV tumors) testing therapy response to CE. In the CE group 18/68 animals responded to treatment of 50mg/kg/d given over 14 days. T/C <20 responders, T/C>20 non-responder. We isolated RNA from all 18 R and 18 NR control tumors. Labeled and amplified cDNAs (Ambion) was hybridized onto Affymetrix U133 Plus 2.0 arrays. Raw cel files were condensed (FARMS). Feature selection and classification was performed using the random forest algorithm and SVM in a nested bootstrap approach. RESULTS We identified a panel of signatures containing between 20 to 300 RNA markers discriminating between R and NR. The best signature containing 300 probesets showed a prospective sensitivity (S+) of 82.78% and a prospective specificity (S-) of 84.0%. A signature of 21 probesets has a S+ of 79% and a S- of 77.78%. We also assigned the response data of the mouse models to the original human tumors and developed corresponding RNA signatures that discriminated successfully the two human tumor groups and proving our model. CONCLUSIONS We have developed highly accurate RNA signatures predicting response to CE in xenograft models carrying human CRC tumors of all four stages. This signature, if successfully validated in a larger cohort, could be applied to pts with CRC stage II and III for predicting response to CE with a PPV of 64% and a NPV of 93%.

Independent validation of a prognostic classifier (Predictor-C) in a set of 292 patients with colorectal cancer of UICC stage II.

3558 Background: Adjuvant chemotherapy in stage II colorectal cancer (CRC) is generally not recommended. Patients (pts) with T4 tumors, performations, < than 12 lymph nodes assessed, or positive vessel / lymphatic invasion may receive adjuvant chemotherapy. Following these guidelines 15-20% of the true high-risk pts remain undetected, undertreated and will suffer from progression of disease. We previously reported on the validation of a prognostic 32-gene signature (Predictor-C) in a cohort of 164 pts with CRC of stage II and III (J Clin Oncol 28:15s, 2010 (suppl; abstr 3612). METHODS For this study 292 pts with stage II disease underwent curative resection (R0) between 1990-2000 in medical centers of Denmark, Finland, Germany, Austria, Netherlands and Canada. From all pts frozen CRC tissue, clinical data and a follow-up data of five years were available. Gene expression data were obtained using U133 Plus 2.0 arrays from Affymetrix. RESULTS In this cohort 49/292 pts suffered from a progression of disease, while 243/292 pts had no progression. Applying Predictor-C to this set of 292 UICC stage II pts we observed the following performance chararacteristics: sensitivity of 0.55 (95% CI: 0.402-0.693), specificity of 0.76 (95% CI: 0.706-0.816), positive predictive value (PPV) of 0.32 (95% CI: 0.224 -0.432), negative predictive value (NPV) of 0.89 (95% CI: 0.844-0.932). The point estimate for progression-free survival at 5 years after surgery for pts classified as low-risk is 0.91 (95% CI: 0.772-0.962), and for pts classified as high-risk this estimate is 0.73 (95% CI: 0.630-0.825). The hazard ratio is 3.289 (95% CI: 1.839-5.881), log-rank test and Wilcoxon test of survival times result in p-values < 0.001. The baseline characteristics of these 292 pts were representative for the whole population of CRC patients diagnosed with UICC stage II, in almost all aspects. CONCLUSIONS Predictor-C was successfully evaluated in an independent set of 292 pts with CRC of UICC stage II. Our genomic risk classifier improves the management of pts with stage II after surgery. Low-risk pts identified by Predictor-C may not need adjuvant chemotherapy while high-risk pts may benefit from additional chemotherapy.

Abstract LB-317: BRAF and PIK3CA mutations in addition to KRAS, improve prediction of resistance to cetuximab in a large panel of colon carcinoma xenografts

Proceedings: AACR 102nd Annual Meeting 2011‐‐ Apr 2‐6, 2011; Orlando, FL Accurate prediction of response to therapy is a prerequisite for individualized approaches to colorectal cancer treatment. Testing of various cancer drugs on the individual patient tumor can support selection of effective therapies. To this end, we have established 149 stably passageable NOD/SCID mouse tumor xenografts from patients of all four Dukes stages within the framework of the prospective multicenter clinical study. All patients gave written informed consent prior to surgery. In a pharmacological study, 75 xenograft tumor models underwent single-agent treatment with oxaliplatin (OX), cetuximab (CE), or bevacizumab (BE). Each of the 75 tumors was transplanted onto 20 mice (5 controls and 5 for each drug). Models with a treated-to-control tumor growth ratio of 20% or lower were defined as responders. This resulted in response rates of 7% for OX, 24% for CE, and 4% for BE. Mutation analysis was conducted on 71 xenografts by allele-specific RT-PCR for the following hotspots: KRAS: 34G>A, 34G>T, 34G>C, 35G>A, 35G>T, 35G>C, 38 G>A and 436 G>A; BRAF: 1799 T>A, and PIK3CA: 1624G>A, 1633G>A, 3140A>G. KRAS mutations were found in 46%, BRAF in 10%, and PIK3CA in 15% of the xenografts. 59% of the xenografts carry at least one mutation. KRAS mutations were never observed together with BRAF-mutations while it combined with PIK3CA in 8 cases (11%). The corresponding primary tumors showed the same mutation profile. We examined mutation status with response to therapy to CE and OX. 3/5 OX responders were KRAS wildtype (wt) (Sensitivity (S+) 66%) while 31/66 OX non-responders were KRAS mutated (Specificity (S-) 46%). If also mutations in BRAF and PIK3CA are considered 40/60 OX non-responders were identified (S-66%). Of the 18 CE responders, all are wt for BRAF and PIK3CA. Three show mutations of KRAS, two of them in codon 13, which seems to confirm a recently published clinical observation by De Roock et al. (JAMA 2010, 304:1812) that this mutation may be related to an improved response under CE treatment. Wt status in all three genes was observed in 83% of the responders. KRAS mutations were seen in 57% of the non-responders. If mutations in BRAF and PIK3CA are also considered, the fraction of recognized non-responders increases to 73%. We also investigated how well response can be predicted from the mutation data. A machine learning approach was used to estimate unbiased prospective prediction rates for response to CE based on the mutation profile. In each of a large number of bootstrap loops, a new predictor was trained on a new randomly selected subset of the tumors and applied to the remaining tumors. KRAS, BRAF and PIK3CAmutations achieve a S+ of 84% (95% CI: 59% and 96%) and a S- of 68% (95% CI: 54% and 80%). KRAS, BRAF, and PIK3CA mutation status combined has potential clinical utility higher than KRAS alone to individualize CE treatment, also in the adjuvant setting, for patients with Dukes B and Dukes C. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr LB-317. doi:10.1158/1538-7445.AM2011-LB-317

Detector-C: A blood-based IVD with high sensitivity and specificity for early detection of colorectal cancer.

3580 Background: Colonoscopy is the gold standard for screening of colorectal cancer (CRC). However, only 20%-30% of the screening population aged 55-75 years use colonoscopy due to its inconvenien...