Kindreds and Genomic DNA

A total of nine probands/kindred with nonsyndromic, autosomal dominantly-inherited dentin defects were available for this study. Two families segregated DD II and seven families segregated DGI II/III. Members of the Brandywine DGI cohort have been reportedly linked to chromosome 4q21 [Boughman et al., 1986] and an exon 2 mutation in a single member of this extended family (Family DGI-1) has been preliminarily reported [Hart and Hart, 2007] but the clinical information, DSPP repeat sequence, and haplotype data were not reported. One DD II family (Family DD-2) has also been clinically described and linked to chromosome 4q21 [Dean et al., 1997]. DNA was available for 22 members (15 affected and seven unaffected) from these nine kindred. Clinical photographs and radiographs on all patients were reviewed (by J.T.W. and T.C.H.). Affected individuals and family members were identified at dental clinics at the following locations: National Institutes of Health, Bethesda, MD; University of North Carolina, Chapel Hill, NC; University of Colorado School of Dental Medicine, Aurora, CO; and the University of Indiana School of Dentistry, Indianapolis, IN. Informed consent was obtained according to the Declaration of Helsinki and approved by the corresponding Institutional Review Boards. Genomic DNA was isolated from whole blood samples using standard protocols. Normal human genomic DNA was purchased from Coriell Cell Repositories (Camden, NJ): African American (HD04), Africans South of the Sahara (HD12), Amerindian Yucatan population (GM10970–GM10979), Ami tribe of Taiwan (HD25), Andes of South America (HD17), Atayal tribe of Taiwan (HD24), Caucasian (GM17208–GM17224), Chinese (HD32), Japanese (HD07), Mexican Indian (HD28), and Northern European (HD01). Genomic DNA was isolated from the HSF-6 human embryonic stem cell line (passage 69) (a gift from Dr. Pamela Gehron Robey and Dr. Sergei Kuznetsov of the NIDCR/NIH) using standard protocols. The mutation nomenclature conforms with journal guidelines (www.hgvs.org/mutnomen). All numbering assumes the A of the ATG start codon (codon 1) as nucleotide 1. Reference sequence: GenBank NM_014208.3.

DSPP Mutation Analysis in Early Exons

The first four exons and the nonrepeat 50 portion of exon 5 of DSPP were amplified using the primers listed in Supplementary Table S1 (available online at http://www.interscience.wiley.com/jpages/1059-7794/suppmat). Briefly, 100 ng of DNA was amplified using standard PCR conditions with an annealing temperature of 581C for 40 sec. Following PCR, products were either gel-purified using the GeneClean kit (Qbiogene, Santa Ana, CA) or using exonuclease (Epicentre, Madison, WI) and shrimp alkaline phosphatase (USB, Cleveland, OH) according to the manufacturers’ instructions. The purified amplicons were sequenced in both directions using Big Dye Terminator v3.1 (Applied Biosystems, Foster City, CA) chemistry and a 3730 DNA analyzer (Applied Biosystems) by the NIDCR Division of Intramural Research DNA Sequencing Core. Alignments were constructed using the basic alignment search tool (BLAST; www.ncbi.nlm.nih.gov/blast) and the DSPP reference sequence NM_014208.

Haplotype Analysis for Families with c.2525delG Mutation

Five STR markers (D4S1534, D4S2409, D4S2929, D4S2284, and D4S2460) that flank the DSPP gene were genotyped using standard methodology [Hart et al., 2003]. Two additional sets of primers were used to genotype known intragenic SNPs, using the profile listed for DSPP mutation analysis, but with an annealing temperature of 601C for 30 sec. Primer set 7 (Supplementary Table S1) includes rs2736978, rs34603924, and rs2615487. Primer set 9 (Supplementary Table S1) includes rs2627699, rs2736980, and rs2846914. PCR products were electrophoresed through 1.8% agarose gels, bands (497 bp for set 7; 564 bp for set 9) were extracted using the GeneClean kit, and sequenced as described above.

Cloning and Sequencing of DSPP Repeat Domain

Because of the difficulty in sequencing the DSPP repeat within exon 5, the full-length 2.4-kb repeat domain (DPP) as well as the 1.2-kb 30 “hypervariable repeat region” (DSPP HVRR) were each cloned. Plasmid DNA from at least 10 colonies for each cloning event were necessary to verify a nominal 1:1 ratio of the two alleles and to distinguish true SNPs from occasional Taq DNA polymerase-generated changes. For the PCR of the HVRR, the forward primer is based on a unique Arg codon at basepair number 2692 located in the middle of the Ser Ser Asp repeat domain. In over 250 chromosomes analyzed, only one haplotype (DSPP HVRR Hap 37 GenBank accession no. EU278676) lacked this Arg and was analyzed using unusually long sequencing reactions of DPP alone.

Genomic DNA (300 ng) was used to amplify DPP and DSPP HVRR with 1 Platinums Taq PCR buffer, two units of Platinums Taq DNA polymerase (Invitrogen), 0.1mM of each dNTP, 1.5mM MgCl2, and 0.2 mM each of forward and reverse primer (Supplementary Table S1) in a total volume of 100 ml. The PCR amplification protocol for DPP was as follows: 941C for 5 min; 35 cycles of 941C for 30 sec, 551C for 30 sec, and 3 min at 721C; and then a final 5 min at 721C. The PCR amplification protocol for DSPP HVRR was as follows: 941C for 5 min; then 40 cycles of 941C for 30 sec, 651C for 30 sec, and 1.2 min of 721C; and then a final 5 min at 721C. PCR products (2.0–2.4 kb for DPP and 1.0–1.2 kb for DSPP HVRR) were excised from 1% agarose/TBE gels, purified using the Qiagen (Valencia, CA) Gel Extraction kit, and eluted in 30 ml of the provided elution buffer (EB). A total of 2 ml of gel-purified PCR cDNA was reacted with 0.5 ml Topos TA cloning vector (pCR4-TOPO; Invitrogen) in the presence of 0.5 ml of the provided salt solution, and incubated at room temperature for 30 min. A total of 20 ml of MAX Efficiencys Stbl2TM competent cells (Invitrogen) were transformed with 1 ml of the above reaction for 30 min on ice followed by 25 sec at 421C treatment, 2 min on ice, and incubation for 90 min with gentle agitation at 301C with 180ml of the provided SOC nutritional growth medium. Cells were spread on a 60-mm Luria broth (LB)- agar/ampicillin (100 mg/ml) plate and grown at room temperature for 72 hr (until colonies were greater than 1mm in size). Selected colonies were grown overnight in 5ml LB/ampicillin (100 mg/ml) at 371C with vigorous shaking. Plasmids were isolated from the entire culture using Promega’s (Madison, WI) Wizard SV Miniprep kits and eluted with 50 ml of water. DNA sequencing was performed using M13 forward and reverse primers. DPP was also sequenced with an internal forward oligonucleotide (CAGACAGCAGCAAATCAGAG). Most sequencing reactions were done with 30 cycles; however, low amounts of plasmid required more cycles.

Bisulfite Treatment for Identification of 5-Methyl-Cytosines in DSPP Repeat

A nonmethylated control DPP PCR product (Coriell African American genomic DNA, GM17031) was added to aliquots of Ssp-I digested genomic DNA (see below) from human embryonic stem (ES) cells and a somatic cell source (Coriell Northern European, GM17007); both sets were treated with bisulfite to determine the presence of 5-methyl-cytosines. Due to a homozygotic indel in the nonmethylated PCR-product control DNA, which was absent in the experimental genomic DNA samples, the sequencing results could be distinguished as control or genomic DNA. Standard bisulfite conditions described in the literature were insufficient to completely convert all of the PCR-product control DNA cytosines to thymidines. The final bisulfite method used was a variation on previously described methods [Paulin et al., 1998; Warnecke et al., 2002]. The repeat domain of DSPP within exon 5 is flanked by multiple Ssp-I restriction sites within the adjacent introns. Like the naturally occurring deamination events in vivo [Lindahl and Nyberg, 1974], bisulfite reactions are ineffective unless the DNA is single stranded during the process. Therefore, the experimental genomic DNA was digested with Ssp-I to render the target DSPP DNA smaller in size and therefore less likely to reform the double helical DNA after denaturation and also to be more similar in size to the control PCR template DNA. Approximately equimolar amounts of experimental genomic DNA and control PCR DNA were combined and treated under final conditions of: 1.72M sodium metabisulfite, 5.36M urea, and 0.5mM hydroquinone (all from Sigma, St. Louis, MO) in a thinwalled 200-ml PCR tube. The samples were heated to 941C for 2 min and then treated for 20 cycles of 941C for 30 sec and 551C for 60 min. The product was purified using the Qiagen PCR purification kit and eluted in 80 ml of EB. A total of 8 ml of 3M NaOH was added to 72 ml of the eluted DNA and incubated for 15 min at 371C, neutralized by 56 ml of 5M ammonium acetate, purified, and eluted with 30 ml EB. A NanoDrop spectrometer (Thermo Scientific, Waltham, MA) was used to quantify the final DNA concentration and 200 ng DNA was used as template for PCR. These samples were amplified as above using bisulfite forward and reverse primers (Supplementary Table S1) in a final volume of 50 ml. Cycle parameters: 5 min at 941C, 35 cycles of 30 sec at 941C, annealing for 30 sec at 501C, 20 sec 721C extension, with a final 5 min at 721C. Primer oligonucleotides were designed to amplify the last few repeats in the sense strand after the bisulfite reaction. The 200-bp amplicons were cloned into TOPO TA cloning vector and sequenced as described above.

Sequence Analysis

Construction of the DPP sequence for each allele was completed by compiling data from both the DSPP HVRR and DPP sequencing results. Sequences were assembled using SequencherTM 4.7 (Gene Codes Corporation, Ann Arbor, MI) software and the online version of the CAP3 Sequence Assembly Program (http://deepc2.psi.iastate.edu/aat/cap/cap.html) [Huang and Madan, 1999]. Final results were verified by visual analysis. The alignment of the 37 human DSPP HVRR haplotypes from the Coriell samples was completed by hand and entered into Networkr 4.2.0.1 (Fluxus Technology, Ltd., Suffolk, UK) to compose the phylogenetic network using the reduced median network calculation option. Hydrophobicity was determined by generating a Kyte-Doolittle plot (www.vivo.colostate.edu/molkit/hydropathy/index.html) [Kyte and Doolittle, 1982]. The plots were constructed with serines instead of phosphoserines; the latter would have resulted in even stronger hydrophilic scores for the nonmutated portions of the DSPP repeat domain.

Patient Description and Mutational Analysis

As shown in Figure 1A and Table 1, a cohort of nine patients/families diagnosed with either DGI or DD were sequenced for DSPP mutations. Two probands, Patients DGI-7 and DGI-1, had mutations identified in the nonrepeat portion of DSPP (Fig. 1B). Patient DGI-7 was found to be heterozygous for a c.13511G4T alteration that affects the splice donor site of intron 3. Interestingly, a different mutation of this same nucleotide, c.13511G4A, has been reported in a family with DGI-II [Xiao et al., 2001]. Proband DGI-7 (of the long-studied Brandywine cohort) was found to be heterozygous for a c.49C4T transition that is predicted to cause a p.P17S substitution. The identical mutation was recently reported in a Chinese DGI-II family [Zhang et al., 2007] and a c.49C4A transition (p.P17 T) was previously reported in a different Chinese DGI-II family [Xiao et al., 2001]. One affected member of the same Brandywine cohort was reported to have had a 36-bp in-frame deletion followed by an 18-bp inframe insertion within the 30 end of the repeat domain of DSPP [Dong et al., 2005]. We confirmed this pair of indels in Proband DGI-1 (DSPP HVRR Hap 38, Genbank accession no. EU278677) but contend this is likely a normal variant and that the c.49C4T alteration is likely the causative event (see below).

The remaining seven new probands were found to have frameshift mutations in the 2-kb repeat portion of DSPP exon 5. All seven showed a net −1 frameshift (i.e., a loss of 1 or 4 bp resulting in the same, new, −1 open reading frame) in the first half of the repeat (Fig. 2). The locations of these four mutations are reported with respect to the current DSPP reference sequence (NM_014208.3), numbering from the start codon (A is nucleotide 1). Because of the high degree of sequence variability among the normal DSPP haplotypes (see below), the values are not the exact numerical location of the mutation within the specific patient’s own DNA. The exact sequence for each unique frameshift is available in GenBank (accession no. EU284750–EU284753) and an alignment is available in Supplementary Table S2). Two different frameshift events were found in the two DD families. One DD family (Family DD-2) had a −4-bp (loss of TCAG) frameshift (c.1870_1873delTCAG; GenBank accession no. EU284750) while the other DD family (Family DD-3) had a similar 4-bp deletion 50 bp 30 (c.1918_1921delTCAG; GenBank accession no. EU284751). The two frameshifts were verified to be separate events when aligned with each other and to a normal control haplotype, due to the presence of identifiable sequences between the two mutation sites (Fig. 2; Supplementary Table S2). Interestingly, the TCAG lost in both DD families are two of the three times this precise sequence is found in the entire DSPP repeat, and these four bases may represent currently unexplained mutational hotspots in humans.

Among five DGI patients/families, two different −1-bp deletions were found 30 to the DD frameshifts (Fig. 2). In a single DGI patient (Patient DGI-2), the loss of an A was identified (c.2272delA; GenBank accession no. EU284752) while in the other four families (Families DGI-3 to DGI-6), a G was deleted (c.2525delG; GenBank accession no. EU284753). There were no apparent unique qualities to the DNA sequence surrounding the deleted G to explain a mutational hotspot. All of the c.2525delG patients had the identical DSPP repeat haplotype (DPP Hap 2A; GenBank accession no. EU278627). Furthermore, detailed analysis of an 4-Mb portion of chromosome 4 surrounding the DSPP gene for all of the patients carrying the c.2525delG mutation identified in an identical haplotype (Supplementary Table S3). (A control sample with the same DPP Hap 2A haplotype but lacking the c.2525delG mutation was analyzed for comparison.) These data suggest the mutation identified in these four families was inherited identical by descent and is likely an example of a founder-effect mutation that has been propagated within the U.S. population. For every case in which DNA from family members was available, one normal and one frameshift mutation allele was found in each affected family member, and unaffected relatives had two normal alleles. In addition, no frameshifts were identified within the full DSPP repeat domain in 100 chromosomes from normal individuals of diverse geographic locations. The 2-kb repeat portion of DSPP in humans is the result of 220 tandem repeats of the nominal sequence, AGC AGC GAC. Each repeat encodes a Ser Ser Asp (SSD) tripeptide, although the exact number of the tandem repeats and the specific sequence of each underlying repeat have drifted to some degree since an apparent ancient expansion event(s). All of the −1 and −4 frameshifts would change the long, hydrophilic (phosphorylated) SSD repeats into a polypeptide rich in the hydrophobic amino acids valine (Val), alanine (Ala), and isoleucine (Ile). The purely hydrophilic and soluble DSPP protein therefore abruptly changes into a protein with an essentially hydrophobic carboxyterminal domain starting at the frameshift location and continuing past the normal stop codon until an in-frame stop codon is reached 12 codons later (Fig. 1C). The most 50 frameshift mutation (c.1870_1873delTCAG) in Family DD-2 would have 4600 hydrophobic amino acids at the carboxy-terminus. The slightly more 30 mutation (c.1918_1921delTCAG) in Family DD-3 would have a similar number. The DGI patients in our cohort had frameshifts that would translate into 500 (c.2272delA) and 400 (c.2525delG) hydrophobic amino acids.

Identification of Normal DSPP Haplotypes and Their Possible Use in Geo-Ethnic Studies

The entire DSPP repeat (DPP) was sequenced from 50 unaffected control individuals (100 chromosomes) of geographically diverse origins. Although, no frameshifts were observed in any control DSPP genes, an extraordinary amount of variation was identified within the 2.2-kb repeat domain, more similar to what is seen in microsatellite repeat polymorphisms. For example, there was a net difference of 432 bp (144 amino acids) between the longest (DSPP HVRR Hap 24; GenBank accession no. EU278663) and shortest (DSPP HVRR Hap 33; GenBank accession no. EU278672) haplotypes due to indels. It was also evident that within the entire repeat domain of DSPP, the majority of the differences occurred in the 30 half of the repeat. This 1.2- kb HVRR has somewhat arbitrarily been denoted to begin at a unique Arg (2,692 bp downstream from the start codon; GenBank accession no. NM_014208.3) and continues until only a few basepairs before the stop codon. Because the majority of differences in the DSPP gene in humans occurred in the HVRR portion, this 1.2-kb domain was analyzed further to define haplotypes within a population of control individuals. A comparison of 188 chromosomes from 94 individuals from 10 geographically distinct regions identified 37 unique haplotypes of DSPP HVRR (DSPP HVRR Hap 1–37; GenBank accession no. EU278640–EU278676, respectively). The 37 SNPs and 20 indels, which in specific combinations defined the individual haplotypes, are summarized in Supplementary Table S4A and B, respectively A complete alignment of all of the DSPP HVRR haplotypes is available (Supplementary Table S5).

Similar to the findings in human mitochondrial DNA and Y chromosome genes, common haplotypes of autosomal DSPP HVRR were present in all the studied ethnic groups from Coriell Cell Repositories [Maca-Meyer et al., 2001; Tishkoff and Verrelli, 2003; Underhill et al., 2001]. Other haplotypes appeared to be more limited to peoples sharing hypothesized human migration routes often associated with out-of-Africa theories (Fig. 3A). For example, haplotype 21 was very frequent in the Asian-descended groups (21% of the 100 chromosomes obtained from Chinese, Japanese, Taiwanese, and their migrant relatives, the Mexican and Andes groups) but was absent in the 88 chromosomes of African (African South Sahara and African American) and European (Caucasian and Northern European) descent (Fig. 3A). The cluster including haplotypes 27, 30, 31, 32, and 36 was only observed in our African populations. While analysis of the HVRR is the most simple and data-rich approach, additional information can be obtained by analyzing the entire DSPP repeat. Many of the most common, shared DSPP HVRR haplotypes could be further separated into one or more related but distinct haplotypes based on indels and SNPs that occurred 50 to the HVRR (DPP; GenBank accession no. EU278619–EU278639). For example, with the addition of the 50 SNPs/indels, two of the most frequent and otherwise shared HVRR haplotypes (1 and 17) separated along ethnic group lines: the Chinese and Japanese-associating alleles separating from the Caucasian, Northern European, and Amerindian groups for both of these haplotypes (GenBank accession no. EU278619–EU278624 for Hap 1 and EU278630–EU278631 for Hap 17).

Origin of the DSPP Repeat Expansion and Evidence for Subsequent SNP Formation

Due to the repetitive nature of DSPP’s exon 5 and the large number of variants among humans (even more so when compared to our closest living relative, the chimpanzee; data not shown), the use of a standard outgroup was not a practical method to directly decipher the common ancestral sequences of the entire DSPP repeat. The current human population is generally agreed to have descended from a relatively small group of genetically narrow individuals that survived one or more severe population bottlenecks within the last few hundred thousand years or less [Liu et al., 2006]. Therefore, a significant number of the differences in the HVRR observed among the 94 individuals studied can reasonably be hypothesized to be predominantly the result of relatively recent changes in the DNA. When the SNPs of the HVRR were tabulated (Supplementary Table S4A), 75% of the changes could logically be concluded to be the result of C-to-T transitions (or G-to-A in the antisense strand). These nucleotide changes were usually at a CpG or a CpApG, suggesting that at least the repeat portion of the gene may be methylated and therefore susceptible to the relatively rapid deamination transition of 5-methyl-cytosine to thymine.

To measure the frequency of transitions within the entire DSPP HVRR and not just the SNPs observable between modern haplotypes, we quantified the changes in each consecutive repeat unit compared to that of a proposed primordial repeat sequence of AGC AGC GAC. This hypothesized primordial repeat is based on the following observations for a representative human DSPP sequence (NM_014208 from basepair 1765 to 3897): 1) Most of the serines in the repeat are either AGC or AGT codons, with 1% of the TCN type. There were only three TCN serine codons (TCA) in the entire human 2.2-kb DSPP repeat and these are all found in the most 50 portion of the sequence. TCN-type codons are not rare-usage codons for human proteins; indeed, they are not even rare in the nonrepeat coding portions of DSPP. This suggests that the majority of the DSPP repeat (and all of the HVRR) was likely due to a series of expansions involving a short repeat that lacked any TCN-type serine codons. 2) C-to-T transitions can be considered predominantly unidirectional due to deamination events (often due to 5-methyl-cytosine); therefore, T-to-C transitions are rare compared to C-to-Tevents. Thus AGTserines within the DSPP repeat are likely to have derived from AGC codons. A similar argument can be made for the silent change in the aspartic acid codon (GAC to GAT transitions). The GAC was probably the primordial codon and the GATs present are a result of later C-to-T transitions. 3) The nonomer repeats observed are GC-rich (i.e., AGC AGC GAC and not including AGT or GAT codons) and are more likely to be remnants of a primordial sequence expansion than to have been assembled from rare T-to-C type changes. Therefore, we propose that AGC AGC GAC is the primordial nonomer and that almost all of the current variations on this repeat in humans (particularly within the HVRR) are the result of C-to-T transitions (or G-to-A from the antisense strand). Using this primordial repeat as a simple outgroup, the relative changes in specific positions within the consecutive nonomers throughout the entire repeat can be estimated. Only complete nonomers within the repeat were used in these analyses. Because the processes of expansion and contraction of the repeats by slipreplication can occur using any set of two or more neighboring, usually identical, nonomers, the absolute number of changes in any one position over the entire repeat may not be particularly insightful. Finally, although the HapMap (www.hapmap.org) shows the DSPP repeat domain to be a recombination coldspot and we saw no direct evidence of recombination events (i.e., new haplotypes that logically are the recombination of common haplotypes), such processes cannot be ruled out as having occurred in the past. Considering all of the above, we determined the percent of change at each base position of the 9-bp primordial repeat compared to all other positions in the repeat unit (Table 2). Within DSPP HVRR, approximately two-thirds of second position AGC serine codons were found to be changed to the alternative serine codon, AGT. This high frequency of change fits well with CpG methylation and subsequent higher rates of deamination of the 5-methyl-cytosine in AGC(pG). Bisulfite sequencing analyses of somatic cell genomic DNA did confirm the presence of methylated cytosines in the repeat’s second serine CpG site just 50 to the stop codon (Fig. 3B-1). Bisulfite sequencing of nonmethylated (PCR) control DSPP DNA added to the same reaction (but recognizable as control due to the presence of an indel in the original template DNA) demonstrated that the bisulfite treatment conditions were sufficiently stringent to result in 100% loss of cytosines in all control reactions (Fig. 3B-2). (Paradoxically, the methylation of a cytosine that makes it more likely to change into a thymidine in normal biology, protects it from deamination into a thymidine in a bisulfite reaction.) The complementary CpG in the antisense strand results in a GAC-AAC (Asp to Asn) codon change in the sense strand. This change occurred in 16% of the repeat units found in DSPP HVRR, significantly lower than the CpG on the sense strand (Table 2). Asparagine amino acids appear to be tolerated periodically along the coding repeat, but such substitutions may be limited in spacing or in total number in order for the DSPP protein to function properly. Methylation-mediated deamination of CpG dinucleotides to TpG in vivo is significantly affected by the identity of the adjacent 30 base. Krawczak et al. [1998] have shown that CGA-to-TGA transitions in the noncoding strand (as in the second position Ser-to-Ser transition described above) occur about twice as frequently as the CGC-to-TGC transitions (as in the antisense transition resulting in Asp-to-Asn changes) in humans. Therefore, the lower percent of the antisense CpG transition (16% vs. 64%) may be due to a combination of lower rates of CpGpC methylation/deamination events and a moderate selection pressure against the occurrence of too many Asn amino acids within the DSPP repeat.

The second most frequent change in the primordial repeat was a GAC(AG)-GAT(AG) transition. More than one-third of GAC codons had undergone a translationally silent change to GAT (Asp-to-Asp) throughout DSPP HVRR (Table 2). This is a much higher percent of change than seen in the other repeat’s non-CpG cytosines. This SNP could be explained by CpNpG methylation/ deamination events known to be common in plants and more recently reported in mouse ES cells [Dodge et al., 2002]. Sequencing data from bisulfite-treated human ES cell and somatic cell genomic DNA showed the occasional presence of CpNpG methyl-cytosines in this GAC codon near the end of the repeat (Fig. 3B-3). No GAC methyl-cytosines were detected in the PCR control samples (Fig. 3B-4). Other CpNpG sequences exist in the nonomer repeat unit, but these were not observed to change in the repeats as frequently nor were cytosines observed at these positions after bisulfite treatments in either ES cell or somatic cell genomic sense-strand DNA, suggesting that these sites are rarely, if ever, methylated. This low number of changes supports the previous observation that CpT methylation occurred much less often than CpA methylation in the mouse ES cells [Meissner et al., 2005]. Finally, the three adenosines within the repeat were the least likely to change (0–1%, Table 2).

Indel Formation Within the DSPP Repeat Domain

The nominal 9-bp DSPP repeat unit was prone to the formation of indels, most likely due to slip-replication events. Indels identified in the normal samples within the HVRR occurred in multiples of nine nucleotides (9, 18, 27, …) up to 189 nucleotides in the largest apparently single indel event. (Some indels within DPP [50 to the HVRR] are multiples of 3 bp and not the typical 9 bp seen in the HVRR.) The indels always resulted in an in-frame final product in the normal population, although the precise location of any one event (to within a single base) and/or the exact sequence of the bases physically inserted or deleted could not always be unambiguously assigned. In all cases, however, a narrow region for the slip-replication event was clearly indicated by the simplest alignment. In our analysis, the number and location of the indels were set to minimize the total number of possible haplotypes. Evidence of the requirement to often have two identical, neighboring repeats for a slip-replication event to occur is demonstrated when comparing all human DSPP HVRR haplotypes at nucleotide position 2861–2877 (Fig. 3C). Each of the identified haplotypes had one of the three sequences. In Figure 3C-1, the two nonomers are not identical; however, upon SNP formation (Fig. 3C-2) they became identical and perfectly situated for an in-frame slip replication (Fig. 3C-3). Most of the 20 indels identified in the 37 HVRR haplotypes had a clear pattern of slip-replication similar to this example.