Common Genetic Variants Influence Whorls in Fingerprint Patterns.

To the Editor

Early work on dermatoglyphics identified three major categories of fingerprint patterns: arches, whorls, and loops, differentiated according to landmark structures formed by dermal ridges: the triradii and core (Holt, 1968). These pattern formations are determined by the ratio of volar pad height to width in utero (Mulvihill and Smith, 1969) influenced by gene interaction with intra-uterine environment (Penrose, 1968). Mathematical models suggested for dermatoglyphic development include heterogeneous genetic factors influencing development versus between-digit differences, with a pattern of covariation between digits suggestive of a morphogenetic field effect(Martin ). Multivariate linkage analyses revealed a pattern of factor loadings for ridge count which supported this argument, and also found linkage to 5q14.1 driven by index, middle and ring fingers (Medland ). A very high heritability (h2 = .65-.96) has been reported for up to 12 dermatoglyphic characteristics (Machado ), suggesting a genetic basis for pattern type. Building on previous findings, the present study sought to identify genetic variants associated with fingerprint patterns on each digit. Two samples of twin and families were recruited from the Queensland Institute of Medical Research (QIMR), and from the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort study(Boyd ). Participants and participant’s parents provided written informed consent and ethical approval was obtained from the ALSPAC Ethics and Law committee and the Local Research Ethics Committees.

Adult and adolescent samples from QIMR were analyzed as one sample of 3301 participants from 1764 families. Fingerprints were collected for the adult sample using rolled ink prints on paper and an electronic archiving system (Medland ) was developed for the adolescent sample. Pattern intensity (the number of triradii) and ridge count (for whorls, the greater of two counts was used) were then manually coded (by SEM and DZL). Within the ALSPAC cohort, 5339 individuals who had GWAS and finger pattern information were used in this paper (please note the study website contains details of data available through a fully searchable data dictionary, http://www.bris.ac.uk/alspac/researchers/data-access/data-dictionary/). Pattern type for each digit was scored and coded (by SEM) from photocopies of the palmar surface of the hands (Medland ). Any digit where the fingerprint pattern was not clearly visible was coded as missing information. As the full patterns of the thumbs were not clearly visible we excluded this digit from analyses. Intensity and ridge count data were then re-coded in terms of presence or absence of whorls and arches on each digit, with loops as the reference category as they are the most common pattern type, and arches were not analyzed due to low pattern frequency. For reference, the thumb on each hand is coded 1 and the little fingers 5, and right/left hand designated using the prefix L or R. After quality control, 10 variables were included in the study: presence of whorls across all digits (L1-5, R1-5), except L1 and R1 for the ALSPAC cohort and L4, L5, R4, and R5 in the QIMR adult sample.

Both QIMR and ALSPAC samples were imputed using the Hapmap2 r22.36 CEU reference. SNPs that had a minor allele frequency (MAF) of >.01 and could be imputed with confidence (R2>0.3) were used in these analyses. Only genotyped SNPs were used for chromosome X.

Heritability estimates were conducted in OpenMX (Boker ), using binary coded data from the QIMR dataset and with sex as a covariate (Table S1). Principal components analysis (PCA) with Varimax rotation was performed to investigate latent factors within phenotypes after orthogonal transformation of correlations. Results showed 3 underlying components of pattern type: whorls on the middle three fingers (digits 2, 3, and 4) on both hands, whorls on the thumbs (digit 1), and whorls on the little fingers (digit 5) (Table S2).

GWAS were conducted using merlin-offline for each digit and each cohort, and combined using Stouffer’s Z score method in METAL to calculate meta-analytic p-values(Willer ). There was no evidence of systematic inflation in the QIMR (λ= 1.004–1.027) or ALSPAC (λ =1.007–1.034) results (Figures S3-S5) and several genome-wide significant SNPs (p < 5×10−8) were found. As Table S6 shows, univariate GWAS for the QIMR sample yielded genome-wide significant p-values for the SNP rs1523452 (p = 7.12×10−9) and adjacent SNPs rs 2244503 (p = 1.52×10−8), rs796973 (p = 5.47×10−8), and rs17071864 (p = 5.42×10−8) for the WL5 phenotype. These were independently replicated in the ALSPAC sample (rs1523452, p = 1.60×10−20; rs2244503, p = 3.76×10−19; rs796973, p = 2.15×10−20; rs17071864, p = 5.68×10−20) and are encompassed by a single gene ADAMTS9-AS2 (Figure S7). As Figure 1 shows, these signals were strengthened in meta-analysis. SNPs within a 2 kbp intergenic region downstream of TBX3 and upstream of MED13L within chr12 were also genome-wide significant for WL4 and WR4, peaking at rs1863718 (p = 8.04×10−9 and 1.36×10−8 respectively), and a variant within the OLA1 gene region was significant for WL2 (rs10201863, p = 3.46×10−8). To further explore variants at a gene level, gene-based tests were conducted using GATES procedure on KGG2.5 (Li ) (Table S8).

Figure 1

Manhattan plots for meta-analysis results

rs1523452 within the ADAMTS9-AS2 gene region presented the strongest signal for phenotypesWL5 (p = 9.74×10−27) and WR5 (p = 7.62×10−15), accounting for 1.61% and 0.93% of the variance. rs1523452 also influences WL4 (p = 2.16×10−21), WR4 (p = 1.33×10−17), and WR2 (p = 3.08×10−10), attenuating at WL2 (p = 9.24×10−6) and further reduced for WL1 (p = 2.66×10−5) and WR1 (p = 0.0001).

Amongst these hits, ADAMTS9-AS2 and OLA1 are documented oncogenes, affecting underexpression of glioma with high WHO grade (III/IV) tumors (Yao ) and inhibition of in vitro cell migration for breast cancer cell lines (Zhang ) respectively, suggesting genetic factors regulating dermatoglyphic morphogenesis may also be present in subtypes of cancer. Furthermore chr12 hits between 113904923 – 113903069bp are located in an intergenic region close to TBX3, which is known to cause ulnar-mammary syndrome. This concurs with previous literature that limb development in utero is influential on subsequent fingerprint patterns that emerge (Mulvihill and Smith, 1969).

ADAMTS9-AS2 is an antisense RNA located at 3p14.1, which may be an mRNA inhibitor for the adjacent gene ADAMTS9. Although there is no direct explanation of the role of ADAMTS9-AS2 in development of whorls on the little fingers, RNAseq analyses show high expression in reproductive organs as well as in the colon and lungs, suggesting it may be influential in early organ development. ADAMTS9 and OLA1 are also expressed in low to medium levels in the skin. Interestingly, variants in ADAMTS9-AS2 also appear to influence whorls on all digits to differing levels of significance and variances (Figure 2A, 2B). Allele frequencies at this variant show that the G allele was associated with higher incidence of whorls in digit 5 (Figure 2C).

Figure 2

Histograms of a: meta-analyses -log10(p-values) and b: % variation explained by rs1523452 (within the ADAMTS9-AS2 gene region) across digits, obtained by Z2 /(N-2 + Z2); c: Trait frequency as a function of allelic variation - Frequency of whorls on the left little finger (WL5; blue bars) and right little finger (WR5; red bars) as a function of the genotype rs1523452 in a sample of unrelated individuals from the QIMR1 cohort (n[AA] = 708, n[AG] = 335, n[GG] = 49). With more G alleles, the proportion of whorls increases. Vertical bars correspond to the 95% confidence intervals on prevalence.

In conclusion, although this study did not find direct evidence for the effects of single genetic variants on specific fingerprint pattern phenotypes, variants within ADAMTS9-AS2 show a gradient influence on whorls across all digits.

Table S1. Frequency and Heritability of presence of whorls on all 10 digits

Table S2. Principal Components Analysis calculated using data from the QIMR adolescent sample (N= 2296). Results show three latent factors underlying the whorl pattern type across digits: factor 1 influencing the 2nd to 4th finger of both hands, factor 2 influencing the little fingers, and factor 3 influencing the thumbs. This concurs with morphogenetic field theory of limb growth, which proposes chemical growth factors that decide differentiation of cells.

Figure S3. QIMR univariate quantile-quantile (Q-Q) plots. Figures illustrate absence of inflation and support the lack of technical artifacts affecting the results.

Figure S4. ALSPAC univariate quantile-quantile (Q-Q) plots. Figures illustrate absence of inflation and support the lack of technical artifacts affecting the results.

Figure S5. Meta-analytic quantile-quantile (Q-Q) plots

Table S6. Results of univariate and meta-analyses GWAS for cohorts QIMR and ALSPAC on each digit (p < 5×10-8)

Figure S7. Locuszoom plots of meta-analysis results

A: top SNPs for WL2 are encompassed by a single gene OLA1; B: hits for WR2 are within and adjacent to MECOM; C: signals for WL4 is between TBX3 and MED13L; D: WL5 and WR5 top SNPs are within ADAMTS9-AS2. In summary, the top SNPs in the meta-analysis results are all within oncogenes with the exception of TBX3 and MED13L, and are associated with inhibition of breast cancer cells, poor prognosis in leukemia, and high versus low grade glioma tumors.

Table S8. Results of meta-analysis for whorls on each digit (p < 5 × 10-8)