To the Editor: It was with great surprise that we read the article entitled ‘Identification of c-kit gene mutations in primary adenoid cystic carcinoma of the salivary gland’ by Vila et al,1 who purported to show a high rate (7/8, 88%) of activating c-kit gene mutations mainly in exon 11. In the study published in 1999 that first identified KIT as an up-regulated gene product in adenoid cystic carcinoma (ACC), exons 11 and 17 of c-kit were sequenced in 28 cases from the University of Virginia and no activating mutations were found.2 This study used sequencing of total PCR products to help obviate the detection of PCR artifacts, and the samples were macrodissected and histologically verified to ensure that tumor cells comprised the majority of the sample. As Vila et al. also correctly noted, there are three other groups who have also published the results of negative screens for c-kit mutations in ACC.3, 4, 5, 6 In addition, we have recent unpublished data from 25 cases of ACC (frozen tissue) obtained from the University of Texas MD Anderson Cancer Center that were subjected to a complete exon sequencing of c-kit, in a study performed at the Wellcome Trust Sanger Institute, in which no mutations were found.

The technology utilized by Vila et al. included 38 PCR amplification cycles of DNA obtained from formalin-fixed, paraffin-embedded (FFPE) tissues, with the subsequent selection of a few individual colonies (2–5) of cloned PCR products for sequencing. The authors did not report the percentage of clones in which sequence changes were detected. We contend that this approach is inherently susceptible to artifact caused by DNA damage that occurs during FFPE processing of tissue, leading to base substitutions during PCR template replication, especially when limited material is used. This problem of mutation detection in FFPE tissue has been clearly documented.7 The authors' findings would therefore have been more cogent if standard controls for such analyses had been used and reported, including the sequencing of products of independent PCR amplifications for confirmation and sequencing matched non-neoplastic tissues.7 Another concern regarding the results in this paper is the fact that several of the tumors had multiple mutations identified in the c-kit gene, including silent mutations. More than one activating mutation in a single sample is very rare in c-kit in the published literature, with only a handful of instances in over 2000 mutations reported, with the majority of these occurring during the development of resistance to imatinib (COSMIC).8 The high incidence of multiple point mutations reported by Vila et al thus further points to the possibility of technical artifact. The authors described their technique as ‘more sensitive’ than those used in other studies, but, in contrast, we believe the previous studies were designed and controlled to prevent artifact and were perfectly adequate to detect mutations that were present in the majority of tumor cells in tissue samples.

The authors characterized the data in the literature regarding the activity of KIT antagonists in treating patients with ACC as ‘conflicting’. We, however, argue that, except for two anecdotal studies that they cite,9 the published data from prospective clinical studies examining the role of anti-KIT targeted therapy (imatinib) in ACC show a very low rate of objective response (1 in 42 subjects, see the table). These clinical data, together with the preponderance of sequencing data, suggest that KIT as a driver mutation in ACC is, at best, rare.

In conclusion, studies from FFPE samples are prone to artifacts and, as such, should be validated in fresh frozen material where possible. In this instance, there is a marked lack of corroborative support for a high frequency of activating KIT mutations in ACC in the data generated by several independent studies in such samples. We would therefore suggest that, on balance, there is compelling evidence to support only a very minor role, if any, for activating KIT mutations in ACC, and that caution should be exercised while drawing conclusions for therapeutic application of the data presented by Vila et al.