|Home | About | Journals | Submit | Contact Us | Français|
Keratoacanthoma (KA) is a benign keratinocytic neoplasm that usually presents as a solitary nodule on sunexposed areas, develops within 6-8 weeks and spontaneously regresses after 3-6 months. KAs share features such as infiltration and cytological atypia with squamous cell carcinomas (SCCs). Furthermore, there are reports of KAs that have metastasized, invoking the question of whether or not KA is a variant of SCC. To date no reported criteria are sensitive enough to discriminate reliably between KA and SCC, and consequently there is a clinical need for discriminating markers. We screened fresh frozen material from 132 KAs and 37 SCCs for gross chromosomal aberrations by using comparative genomic hybridization (CGH). Forty-nine KAs (37.1%) and 31 SCCs (83.7%) showed genomic aberrations, indicating a higher degree of chromosomal instability in SCCs. Gains of chromosomal material from 1p, 14q, 16q, 20q, and losses from 4p were seen significantly more frequently in SCCs compared with KAs (P-values 0.0033, 0.0198, 0.0301, 0.0017, and 0.0070), whereas loss from 9p was seen significantly more frequently in KAs (P-value 0.0434). The patterns of recurrent aberrations were also different in the two types of neoplasms, pointing to different genetic mechanisms involved in their developments.
Keratoacanthoma (KA) is a common, benign keratinocytic neoplasm that usually presents as a solitary, dome-shaped, pink or flesh-colored nodule developing a central keratin-filled crater. The lesion is usually located to sun-exposed areas of the skin of elderly persons and is mostly developing within 6-8 weeks. Since it spontaneously regresses after 3-6 months, KA has by some been considered to be an abortive or self-healing malignancy (Schwartz, 1994). KA may appear as multiple lesions, (Grzybowski, 1950; Muir et al., 1967; Ferguson-Smith et al., 1971), and shows increased incidence among immunosuppressed patients (Sullivan and Colditz, 1979). The histopathologic diagnosis of KA is based on the architecture as well as cytologic features, and when a typical clinical history is known, most KAs can be distinguished from squamous cell carcinomas (SCCs) (Weedon, 2003; Elder et al., 2005). However, particularly in the early proliferative phase, KA shares features with SCC such as infiltration and cytological atypia. Furthermore, there are reports of KAs that have metastasized, raising the question of whether KA is a variant of SCC (Hodak et al., 1993). There are so far no criteria that are sensitive enough to discriminate reliably between a KA and SCC, neither histopathology, biomarkers, biochemical, nor genetic studies (Kerschmann et al., 1994; Quinn et al., 1994; Cribier et al., 1999; Jensen et al., 1999; Putti et al., 2004). There is thus a clinical need for markers that may distinguish between KA and SCC. Previously, Clausen et al. (2002), found by means of comparative genomic hybridization (CGH) that about one-third of 70 KAs harbored genomic aberrations, and that some of these were not shared by cutaneous SCCs, although some aberrations were common for both types of lesions (Clausen et al., 2002). The numbers of analyzed cutaneous SCCs were, however, limited.
The aim of the present study was to characterize possible differences in gross chromosomal aberrations between KAs and SCCs by screening with CGH in a sufficiently large number of lesions. Such differences may point to different genetic mechanisms responsible for the development of the two neoplasms.
In the present study, we have analyzed a series of 62 KAs and 19 SCCs with CGH. To increase the statistical power, we have also included our previous data on 70 KAs and 10 SCCs (Clausen et al., 2002), and included data on eight SCCs from another institution (see Materials and Methods). A total of 132 KAs and 37 SCCs were thus screened for gross aberrations in the tumor genome using CGH. Genomic aberrations were found in 49 of 132 KAs (37.1%), and in 31 of 37 SCCs (83.8%), respectively, and the results are reported in Tables Tables11 and and2,2, and for a summary of all aberrations with comparison of those in KAs and SCCs, see Figure 1. The average number of aberrations per lesion detected by CGH was 3.08 and 6.45, respectively, for KAs and SCCs (Figure 2). In KAs recurrent gains of chromosomal material from 9q (18.4%), and losses from 3p (18.4%) and 9p (20.4%) were seen at about the same frequency in the extended material as in the previously reported cases (Clausen et al., 2002), whereas gains of 1p (12.2.0%) and 8q (12.2%) were seen less frequently. Gains of chromosomal material from 3q, 4p, and 19p were also seen in 12.2% of KAs with genomic aberrations. In SCCs, recurrent gains from 1p (41.9%), 3q (25.8%), 8q (25.8%), 20q (25.8%), 19p (19.4%), and 19q (22.6%) were seen, along with losses from 3p (19.4%) and 4p (16.1%). Although with a lower frequency, gains from 1q (19.4%), 5p (16.1%), 9q (12.9%), and 14q (12.9%), as well as losses from 4q (12.9%), 11p (12.9%), 11q (12.9%), 13q (12.9%), and 18q (12.9%) were also seen. Gains of material from chromosome arms 1p, 14q, 16q, and 20q as well as losses from 4p were seen significantly more frequently in SCCs compared to KAs (P-values 0.0033, 0.0198, 0.0301, 0.0017, and 0.0070), whereas losses from 9p were detected significantly more frequently in KAs (P-value 0.0434). Figure 3a illustrates that KAs and SCCs cluster differently based on their respective numbers of aberrations, whereas Figure 3b illustrates differences in the cluster patterns of individual lesion from KAs and SCCs, emphasizing differences in patterns of recurrent lesions.
It is well known that KAs as well as SCCs develop preferentially on sun-exposed areas of the skin, and that actinic keratosis as a result of sun damage is a precursor lesion for SCC (Elder et al., 2005). The reported frequencies of malignant transformation within actinic keratosis are debated, but recent data indicate a significant frequency of transformation, possibly higher than 10% (Mittelbronn et al., 1998). Genetic studies have demonstrated a high frequency of loss of heterozygosity in actinic keratosis (Rehman et al., 1994). Contrary to SCCs, no such precursor lesion has been recognized for KAs, which undergo complete regression in almost all cases (Weedon, 2003). There are thus distinct differences in biological characteristics of these two types of lesions that may be reflecting abrogations of different signaling pathways.
There are only few reports on KAs that have been subjected to CGH analysis (Ashton et al., 2005), including those from our previous study on 70 lesions, of which 36.4% showed genomic aberrations (Clausen et al., 2002). We therefore wanted to get a more comprehensive overview of the genomic alterations aquired by KAs and SCCs, analyzing and comparing a total of 132 KAs and 37 SCCs. In KAs the previously reported recurrent gain of genomic material from 9q (18.4%) and losses from 3p (18.4%) and from 9p (20.4%) were still seen in the extended material with high frequency, whereas previously reported gains from 1p and 8q in this study were reduced to 12.2% each. Our data from CGH analysis of 37 cutaneous SCCs is to our knowledge the largest data set to date. The recurrent aberrations in SCCs are mainly different from those in KAs, with gains from 1p (41.9%), 3q(25.8%), 8q(25.8%), 20q(25.8%), 19q(22.6%), and 19p(19.4%), and losses from 3p(19.4%) and 4p(16.1%) being the most common. Of recurrent aberrations only loss of 3p is seen to about the same extent in KAs and SCCs, whereas gains of 3q, 8q, and 19p to a lesser extent also are seen in KAs.
In addition to sporadic reports about limited numbers of cutaneous SCCs analyzed by CGH, one research group has analyzed 15 SCCs and 12 actinic keratosis by CGH (Ashton et al., 2003). Their conclusion is that numerous genetic imbalances are shared by actinic keratosis and SCC, consistent with the former lesion being a precursor for SCC. The most frequently occurring imbalances of SCCs in our material, that is, gains of 3q, 8q, and 20q, and loss of 3p, are in accordance with results reported previously (Ashton et al., 2003). Gain of material from 1p was the most recurrent aberration in our study. Scoring of imbalances in this region, however, may be challenged by the presence of artifacts, and it is known that many authors do not score imbalances here (Ashton et al., 2005). This may suggest that imbalances on 1p in fact are more frequent than reported so far. To overcome the possibility of false results, we have used reverse hybridization experiments to confirm our data, that is, reversed fluorochrome labeling of test and control DNA (Kallioniemi et al., 1994).
Many of the aberrations in SCCs found by CGH, that is, +3q, +8q, -3p, are consistent with reports from cytogenetic studies (Ashton et al., 2005). The present analyses confirm and extend our previous data (Clausen et al., 2002), showing that more than one-third (36.5%) of KAs harbor genetic imbalances, whereas more than twice as many SCCs (83.8%) showed such imbalances. We now also show that SCCs harbor on average about twice as many aberrations per lesion as KAs with detectable imbalances (6.45 versus 3.08, respectively), and that in addition to generally increased numbers, cutaneous SCCs also harbor significantly higher numbers of specific genetic aberrations than KAs (Figure 3a). The most significant difference is gain of chromosomal material from 20q, which was found in eight SCCs and in only one KA (P-value 0.0017). Gains from 1p, 14q, and 16q, as well as losses from 4p were also observed significantly more frequently in SCCs than in KAs, whereas losses from 9p were seen significantly more frequently in KAs (Tables (Tables11 and and2,2, Figure 1). Furthermore, the patterns of recurrent aberrations were also different for the two types of lesions, as illustrated by the dendrogram in Figure 3b.
Our results clearly show that chromosomal instability is significantly more frequent in cutaneous SCCs than in KAs. The presence of different imbalances in the genome of the two types of lesions may be correlated to their different biological behavior, and suggests different pathogenetic pathways for their developments. This is also consistent with a previous study reporting that loss of heterozygosity was found more frequently in SCCs than in KAs, with fractional allelic losses of 32.0 and 1.3%, respectively (Waring et al., 1996).
The most significant difference between genetic imbalances in KAs and SCCs in this study was the high degree of gains of 20q in SCCs. 20q is a region which is frequently amplified in breast as well as colorectal cancer (Tanner et al., 1994; Ried et al., 1996). Amplification of the consensus region 20q13 is seen in about 70% of colorectal cancers and is associated with increased S-phase fraction as well as bad prognosis (De Angelis et al., 1999, 2001). Candidate oncogenes localized to 20q13 are ZNF217, which is associated with immortality in mammary epithelial cells (Nonet et al., 2001), and STK6/STK15 which codes for Aurora A, a kinase important for segregation of chromosomes during mitosis and which is also linked to tumorigenesis and development of aneuploidy (Goepfert et al., 2002). Detailed genetic analysis on these genes will clarify their possible involvement in SCCs. Recurrent losses from 4p as well as from 4q are also seen in SCCs in the present study, and allelic losses from chromosome 4 arms have been shown to correlate with tumor aggressivness in colorectal cancer (Arribas et al., 1999). Imbalances of the above-mentioned chromosomal regions are hardly ever seen in KAs, suggesting that such aberrations may affect genes inducing infiltrative and metastatic potential of SCCs. Interestingly, the most frequently occurring recurrent aberration in SCCs is gain on chromosome arm 1p, which also, although to a lesser extent, occurs in KAs. The protooncogene JUN is located at 1p32-31 and encodes for a transcription factor that has been reported to be overexpressed in human cancers. JUN regulates cell cycle progression as well as apoptosis, and it is therefore possible that proliferation of SCC as well as of KA may be stimulated through 1p amplification and subsequent JUN overexpression. It is also possible that JUN amplification in KAs may be related to apoptosis as demonstrated during regression. In contrast to the recurrent aberrations discussed above, loss of 9p was seen significantly more frequently in KAs than in SCCs. It is known that the tumor suppressor gene CDKN2A (p16INK4A) which is inhibiting cell cycle traverse, is located to 9p21. Many reports mention 9p deletion as an early aberration that is seen in histomorphologically normal tissues, including nasopharyngeal mucosa, and premalignant tissues (Chan et al., 2002; Stoehr et al., 2005). Thus, 9p deletion may be associated with growth advantage in areas with a retained architectural tissue hierarchy like in KAs, whereas other aberrations may give selective growth to tissues with disrupted architecture. An obvious trait of KAs is the maintenance of symmetric architecture with well-differentiated epithelial strands, which may have fundamental implication for the surveillance of the neoplastic cells.
Other overlapping recurrent imbalances of KAs and SCCs are gains from 3q, 8q, 9q, and 19p, and loss at 3p (Table 2). Many different genes from these regions may be altered and stimulate uncontrolled proliferation, one possibility is the protooncogene C-myc localized to 8q24.
CGH is a screening method with the advantage that it allows for simultaneous detection and mapping of amplified, duplicated, or deleted chromosomal regions of the total genome of tumors in one hybridization procedure (Kallioniemi et al., 1992). Limitations are that balanced aberrations, such as translocations and inversions, will not be detected, and that the hybridization target (the normal human metaphase chromosomes) limits the resolution of the method. Deleted regions must be as large as 5-10 Mb to be detected, whereas amplified regions may be detected down to 1-2 Mb if the degree of amplification is sufficiently large (Forozan et al., 1997). This implies that the percentages of lesions with genetic aberrations, as well as the number of aberrations per lesion reported here, are minimum values. Methods with higher resolution, such as comparative hybridization to genomic arrays (array CGH) and positional hybridization, will be able to reveal more specific genetic aberrations that may serve to further differentiate between KA and SCC.
If we assume that KA and SCC are biologically separate lesions with different pathogenetic background, the presence of overlapping genetic imbalances detected by CGH must be considered. First of all, many of the aberrations detected are probably without influence on tumor progression and cell death. Secondly, it is well known that sun exposure through UV irradiation is important for development of KAs as well as for SCCs. Thus, many of the overlapping aberrations may be due to DNA damage and selection mechanisms caused by the common influence of UV. The observation that genetic aberrations are seen more frequently and are more numerous in SCC than in KA, may result in abrogation of signaling pathways in SCC that are intact in KA, the latter possibly facilitating apoptosis and spontaneous regression of KA.
A total number of 132 KAs and 37 SCC were analyzed. One hundred and twenty-six KAs and 18 SCCs were collected from patients at the Departments of Dermatology, Surgery, and Plastic Surgery at Rikshospitalet, University of Oslo, during the years 1995-2004. A total of 101 KAs and 23 SCCs developed in organ transplant recipients who received immunosuppressive treatment, whereas the remaining lesions developed in patients who were not transplanted or immunosuppressed. Eight of the SCCs lesions were excised and analyzed in the period of 2000-2001 at the Department of Cancer Genetics, The Norwegian Radium Hospital, Oslo (Micci et al., 2003). Center for Cutaneous Research, Barts and the London Queen Mary’s School of Medicine and Dentistry, University of London, UK, provided us with DNA extracted from six KAs and 11 SCCs collected in the period of 2000-2004. The studies have been performed according to requirements from Institutional Review Board and the Declaration of Helsinki Principles has been followed.
Most of the lesions were divided into two halves, one of which was fixed in formalin, embedded in paraffin, and processed for routine histopathologic evaluation; the other half was immediately stored at -80°C for further processing. From all lesions a cross-section through the central part encompassing the whole lesion was used for CGH analysis, tumor tissue thus representing more than 75% of the analyzed material; microdissection was therefore deemed unnecessary.
Sections were cut at 5 μm thickness from the formalin-fixed material and stained with hematoxylin-eosin for routine histopathologic diagnosis. Criteria used for the diagnosis of KAs and SCCs were according to Elder et al. (2005). Briefly, KAs were characterized by symmetrical appearance with a centrally located, keratin-filled crater with overhanging epithelial lips or shoulders. Epithelial strands, often with ground glass appearance, were penetrating dermis from the center of the lesions. According to generally accepted criteria, and based on previous findings where no significant genomic differences by CGH analysis could be demonstrated in bulk tumor DNA between KAs with and without infiltrating growth and atypia (Clausen et al., 2002), we have included all such lesions, irrespective of the degree of cellular atypia and infiltration. Only SCCs that were not crateriform with possible confusion with KAs were included. The diagnoses were given by one of the authors (O.P.F.C.), who is an experienced dermatopathologist.
CGH was performed on fresh frozen tissue according to a previously described protocol (De Angelis et al., 1999, 2001). Normal reference DNA was Texas-Red or Spectrum Red labeled, whereas tumor DNA was labeled with either FITC or Spectrum Green, respectively. A positive control using the cell line MPE 600 and a negative control were usually processed in order to control for artifacts. All samples that had possible artifacts were either subjected to rehybridization with reversed fluorochrome labeling of test and control DNA (dye swap) to confirm true aberrations, or - in a few cases where no more DNA was available - areas with possible artifacts were excluded from the data set. Amplifications and deletions were scored if the red/green ratio was above 1.15 or below 0.85, respectively.
The open-source environment and programming language R, (R Development CoreTeam, 2004), was used for statistical analysis of chromosomal aberrations. Differences between lesions were analyzed as contingency tables for each aberration using the Fisher’s exact test with two-sided P-values calculated as defined by Agresti (1992), where P<0.05 was considered significant. Also a binary matrix was made to explore possible correlations between patterns of chromosome arm aberrations of individual tumors. Gains or losses of whole chromosomes were scored as two events related to each chromosome arm. Two separate aberrations (both gains or both losses) on the same arm were scored as one change, whereas one gain and loss on the same arm were recorded as two separate events. This binary matrix was used to build a frequency matrix of aberrations, where the sum of one aberration in a tumor type was divided by the total number of samples of the particular tumor type analyzed in this study. This weighted matrix was used as input for hierarchical clustering in R (complete hierarchical clustering with euclidian distance metric).
This work was supported by the Norwegian Cancer Society and Medinnova AS.
CONFLICT OF INTEREST
The authors state no conflict of interest.