Establishment of patient-derived non-small cell lung cancer xenografts as models for the identification of predictive biomarkers. and PIK3CA-E545G/K). Our data on an independent cohort support the recent clinical observation, but against the current practiced patient stratification in the cetuximab CRC treatment. Meanwhile, our data seem to suggest that a set of the six-oncogenic alleles may be of better predictive value than the current practiced stratification, justifying a new prospective clinical investigation on an independent cohort for confirmation. = 0.67 if only considering codon-12/13; = 1.0 if considering all KRAS mutations), suggesting lesser roles of KRAS mutation in determining response than originally believed. In particular, there are 4/6 G13D falling into responders, while none for the non-G13D KRAS mutations (0/9), suggesting that indeed G13D patients can benefit from the treatment, while other KRAS mutation patients do not. This observation is consistent to the recent analysis of clinical observation [13, 14]. Considering that our data is completely independent of previous analysis (unrelated subjects and test methods), the observation is more likely to be true. It has been known that not all KRAS mutations are equal with regard to their activity and oncogenicity [14], which is strongly supported by our data. Open in a separate window RI-2 Figure 1 Waterfall plot of T/C% values of CRC-PDXsA. Per KRAS codons 12/13 mutation rule wild type vs. mutations. B. Per the set of oncogenic allele rule Cwild-type/KRAS-G13D vs. at least 1 activating alleles on KRAS-G12G12C/D/V, -Q61X, -A146T, NRAS-Q61X, AKT1-L52R, PIK3CA-E545K/-Q546L and BRAF-V600E. Certain oncogenic alleles better predictive of cetuximab response in CRC The 5/5 G12C/D/V are all non-responders, 2/2 A146T (CR0010 and CR0245) and 2/2 Q61H (CR1515 and CR1530) are all non-responders. 1/1 NRAS Q61R is a nonresponder (CR1574), mutually exclusive to KRAS mutation and BRAF mutation. Both BRAF-V600E containing models, CR0004 and CR0029, are non-responders (2/2) (Table ?(Table11 and Supplementary Figure 2), consistent to the observation that BRAF-V600E causes resistance to cetuximab [24] and is mutually exclusive to KRAS mutation. CR1744 with AKT1-L52R mutation is a nonresponder (1/1), mutually exclusive to KRAS, NRAS, and BRAF mutation. 5/5 PIK3CA-E545K/Q546L mutants (exon 9) are all non-/partial responders, not mutually exclusive to other oncogene alleles, suggesting a possibly role of PIK3KCA mutations in cetuximab resistance [25], although not statistically significant (= 0.28, Fisher’s exact test). In summary, 16/19 non-/partial responders have at least one of the activating alleles of KRAS-G12G12C/D/V (5/19), -Q61X (2/19), -A146T (2/19), NRAS-Q61X (1/19), AKT1-L52R (1/19), PIK3CA-E545K/-Q546L (5/19) and BRAF-V600E (2/19) (Table ?(Table1).1). This is RI-2 in contrast to that 0/8 tested responders are wild-type for all these alleles (Fisher’s exact test = 7.43 10?5). Apparently, there are 5 models (5/19), CR-0560, ?0205, ?1795, ?0012, ?2226, where cetuximab resistant alleles are still yet to be identified [26]. This suggests that the composite oncogenic alleles profile could be more predictive. We should point out that the validity of this set of oncogenic alleles for predicting cetuximab resistance need to be further validated by testing in an independent cohort using a prospective RI-2 design. DISCUSSION Although KRAS-G13D has been suggested not used as predictor for poor response to cetuximab per recent retrospective review of past clinical data [13], it is still insufficient to change cetuximab label to recommend these patients for cetuximab treatment. Usually, only a confirmation in a prospective study using independent cohort of similar disease can potentially be used to change the label. Such a study is still to be conducted. PDXs have very similar histopathology and molecular pathology as patient tumors, and are thus considered closest surrogate experimental models for human tumors [15C19]. A cohort of diverse CRC-PDXs can be particularly useful to add a confirmation in a similar clinical trial as in human. Our prospective mouse clinical trial (MCT), using an independent cohort of test subjects, confirmed that G13D indeed cannot predict the poor response to Rabbit Polyclonal to ELOVL4 cetuximab, in agreement with results from retrospective analysis of human data. This result further supports the notion that a human prospective trial should be conducted to confirm this and to change the label thus many G13D patients can also benefit from the treatment. Our trial, using a random enrolled subjects, seems to discover a new set of oncogenic alleles, with only one of them being.