Diversity and distribution of alpha satellite DNA in the genome of an Old World monkey: Cercopithecus solatus

Retrieval of alpha satellite sequences from the Cercopithecus solatus genome

Work conducted in the early 1980s had shown that enzymatic digestion of genomic DNA from Old World monkeys with several restriction enzymes resulted in a migration profile that was characteristic for alpha satellite DNA, i.e. with bands corresponding to one and multiple repeat units of about n x 170 bp in length [8, 39]. In silico analysis of several sequences isolated from Chlorocebus aethiops led us to select the XmnI restriction endonuclease as a candidate that should cleave a majority of monomers. Experimental digestion of Cercopithecus solatus genomic DNA with this enzyme revealed the expected banding pattern (Additional file 1: Figure S1). We therefore decided to extract DNA from two bands corresponding to monomers and dimers of alpha satellites from an agarose gel and implemented high throughput sequencing on an Ion Torrent sequencing platform providing reads up to 400 nucleotide in length (see Methods).

204,990 and 353,683 raw sequences were obtained for the monomer and dimer samples, respectively. Four in silico filters were applied successively to both datasets: a quality filter keeping sequences with a Phred quality score superior to 25; an extremity filter keeping sequences with the XmnI restriction site at both ends; a length filter keeping sequences within the range 162–182 bp for monomers and 324–364 bp for dimers, and an alpha satellite filter keeping sequences similar to an alpha satellite reference sequence (see Methods). The number of sequences that remained after each filter is reported on Additional file 2: Table S1. A total of 100,713 sequences fitting with all the criteria was obtained from the monomer sample and represents what we call from now on the monomer dataset. For the dimer sample, only 3,568 were obtained, they represent the dimer dataset. The drastic reduction observed within the dimer dataset was mostly the consequence of the length filter and may reflect an intrinsic limitation of the sequencing technology, unable to obtain long reads when template sequences are made of two successive highly identical sequences. These sequences were nevertheless included for further analysis as they provided an additional source of information (see below).

Characterization of alpha satellite diversity in the monomer dataset

A principal component analysis (PCA) using the 5-mer nucleotide composition of DNA sequences was applied to the monomer dataset in order to compare these sequences and identify putative groups without direct alignment. Visualization of sequences into the plane formed by the two first components of the PCA revealed two main groups of alpha satellite monomers, as shown by the distribution of points on Fig. 1a. Monomers were classified into each group by using a hierarchical clustering analysis (HCA) based on a subset of sequences followed by a linear discriminant analysis (LDA) to extend the classification to all the sequences (see colors on Fig. 1a). The most important group, called C1 and shown in purple, contained 82% of the sequences and the other group, called C2 and shown in pastel green, contained the remaining 18%. To address the quality of this classification, 500 sequences were then randomly selected within the complete monomer dataset and were used to generate a phylogenetic tree where branches were colored according to the monomer classification (Fig. 1b). The disposition of sequences from the C1 and C2 groups on this tree provided a further support to our classification into two groups. Moreover, this tree showed a higher degree of divergence between C2 sequences compared to C1 sequences. Actually, the comparison of a subset of 500 randomly selected sequences within each group showed that the average sequence identity inside C1 was 95%, whereas the average sequence identity inside C2 was only 85%. The consensus sequences of C1 and C2 were 172 bp in length, and differed from each other by a total of 9 positions (Fig. 2). Finally, monomers were searched for the presence of CENP-B and pJalpha boxes [40]. A pJalpha box was present in the consensus of C1 and C2 and was found in 95% of C1 sequences and 85% of C2 sequences, whereas a CENP-B box was only found in 0.05% and 0.04% of these sequences, respectively.

Characterization of alpha satellite diversity in the dimer dataset

Among the 3,568 sequences recovered from the dimer dataset, 1,095 contained an intact XmnI restriction site approximately located in the middle, suggesting that the enzymatic digestion was not complete. The remaining 2,473 sequences, which did not possess the XmnI restriction site, were split using an alignment-based process (see Methods) and the resulting monomers were submitted to a length filter, giving a total of 2,408 associated left and right monomers. We focused first on these sequences and submitted them to the same process as described above. PCA showed the existence of two groups for both the left and right monomers, which could be discriminated using HCA (Fig. 3a and b). A comparison of the consensus sequences of each group revealed that the most abundant sequence set for both the left and right monomers belonged to the C2 group, while the two smaller sequence sets had consensus sequences that differed from each other and from the consensus of the C1 and C2 groups (Fig. 2). These two new groups of sequences, which represented 5% of the left or right monomers, will be from now on called C3 and C4 and shown in dark and light pink, respectively. We decided to build a phylogenetic tree with left and right monomers mixed together, using the described color code (Fig. 3c). This tree confirmed the existence of C3 and C4 as separate groups. Their respective average sequence identities were measured to be 86% and 83%. We also checked that left and right monomers belonging to the C2 group could not be distinguished from each other on a phylogenetic tree (Additional file 1: Figure S2), or from the C2 sequences present in the monomer dataset. The comparison of the consensus sequences showed that the C3 and C4 groups differed much more from each other and from the C1 and C2 groups than C1 and C2 differed from each other. Interestingly, C4 was the only group with a consensus length of 171 bp instead of 172 bp. A search for CENP-B and pJalpha boxes showed that most sequences within the C4 group contained a pJalpha box (75%) while the CENP-B box was absent, like observed for the C1 and C2 groups. By contrast, neither the pJalpha box nor the CENP-B box was found in the sequences from the C3 group (Fig. 2).

The dimer dataset was also used to infer information regarding how monomers belonging to different groups associated with each other. All left and right monomers were assigned to one of the C1 to C4 groups (see Methods). Additional file 2: Table S2 reports the results of these assignments as well as associations between left and right monomers, distinguishing dimers that contained the XmnI site (X dataset) and those where the XmnI site was absent (noX dataset). We noticed that sequences from the C1 group were absent from the noX dataset and were poorly represented in the X dataset. This result may appear unexpected as 82% of the sequences from the monomer dataset belonged to the C1 group. Two hypotheses may explain this observation: the high sequence identity within the C1 group may reduce both the likelihood of the inactivation of the XmnI digestion site through mutations and the sequencing efficiency of dimers (see above). A statistical analysis of the X dataset showed that left monomers from the C1 and C2 groups were preferentially associated to right monomers from the same group (Additional file 2: Table S2), which suggests that sequences from the C1 and the C2 groups are tandemly repeated in the Cercopithecus solatus genome. C2-C2 associations were also found to predominate within the noX dataset. Interestingly, left monomers from the C3 group were preferentially associated to right monomers from the C4 group, suggesting the existence of a higher order organization with repeats containing at least two monomers belonging to different groups.

Genomic distribution of alpha satellite families on Cercopithecus solatus chromosomes

We were next interested in studying the genomic distribution of the four groups of sequences identified above. Short oligonucleotide probes have been shown to be more efficient at distinguishing alpha satellite sequences that differ by very few nucleotides compared with classical probes obtained by random priming or nick translation [42, 43]. We chose to use synthetic 18-mer oligonucleotides carrying locked nucleic acid (LNA) modifications at one out of two positions and capable of forming at least 7 GC base pairs, as previous work had demonstrated their interest for the detection of alpha satellite sequences [44]. An in silico probe selection process was implemented in order to identify among the most common 18-mer sequences within a group (found in more than 20% of the monomers) those that were specific for this group (found in less than 3% of the monomers of other groups). As we expected that oligonucleotide probes may still hybridize in the presence of one mismatch, we calculated the expected binding frequencies when one mismatch was present and applied the same selection criteria once again. Additional file 1: Figure S3 reports the sequences that best fitted with our requirements, albeit not completely. Due to the high sequence similarity between sequences within the C1 and C2 groups, probes had to distinguish sequences that differ mainly by only two nucleotides or even a single one (Additional file 1: Figure S3). The two sets of probes selected to target the C1 and C2 groups were therefore designed so that they would compete with each other if used simultaneously. The detection systems (fluorophores or haptens) were chosen in order to allow various combinations of probes to be tested together.

A first series of FISH experiments on Cercopithecus solatus metaphase spreads was performed using probes C1a and C2a or C1b and C2b simultaneously. Probes targeting the C1 group produced intense signals at the centromere (primary constriction) of all chromosomes except a single one (Fig. 4a and c), while probes targeting the C2 group provided signals that are located in the pericentromeric regions (around the primary constriction) of several chromosomes pairs with different labeling patterns (Fig. 4b and c and Additional file 1: Figure S4a). Some chromosomes were extensively labeled by C2 probes on both sides of the centromere, others seem to be labeled at only one side, and others seemed to display no signal (see arrows on Fig. 4d). Stronger signals were observed on the acrocentric chromosome short arms (Additional file 1: Figure S4a). When these probes were used alone, each C1a or C2a probe produced a labeling pattern similar to what was observed in the presence of the other. On the contrary, each C1b or C2b probe used alone labeled regions that are larger than in the absence of the other. These experiments suggest that our probes may hybridize to sequences that differ by a single nucleotide (i.e. C1b binds to sequences from the C2 group and C2b binds to sequences from the C1 group) but that this binding is inhibited in the presence of an adequate competitor probe. In addition, when target sequences differ by at least two nucleotides, a specific detection is achieved in the absence of competitor.

Additional experiments showed that in presence of competitors, the signal produced by C1a overlapped with the signal produced by C1b and the signal produced by C2a almost perfectly overlapped with the one produced by C2b (Additional file 1: Figure S5). This observation supports the idea that the labeling patterns observed with the chosen oligonucleotide probes reflect the distribution of the sequence groups identified by sequence analysis. Moreover, the absence of overlap between signals provided by probes targeting sequences from the C1 and C2 groups suggests that monomers within each group are clustered together and do not mix with each other. Combined with the arguments described above that are in favor of a tandem organization of monomers for both the C1 and C2 groups, these features support the fact that the C1 and C2 groups of sequences represent distinct families of alpha satellite DNA that display a monomeric organization in the genome of Cercopithecus solatus.

Further experiments were performed with probes targeting the C3 and C4 groups. All C3 and C4 probes provided identical labeling patterns, with a strong signal located on a single chromosome, as well as very weak pericentromeric signals on some other chromosomes (Fig. 5 and Additional file 1: Figure S5). The chromosome labeled by the C3 and C4 probes, which is in fact the chromosome that was not labeled by the probes targeting C1, was also identified to be the Y chromosome by cytogenetic experiments (Additional file 1: Figure S4b). The colocalization of probes targeting the C3 and C4 groups and the absence of overlap with probes targeting the C1 or C2 group is consistent with the sequence analysis described above. These results taken together suggest thus that sequences belonging to the C3 and C4 groups represent additional families of alpha satellite DNA that display a higher order organization within the genome of Cercopithecus solatus. As a further control of the consistency between the results from FISH experiments and sequence analysis, we showed that a 13-mer LNA probe that was designed to target the four C1 to C4 groups of sequences (called Cx) provided signals that overlapped with the combined signals of probes targeting each group, i.e. was able to label all chromosomes within the centromeric and pericentromeric regions (Additional file 1: Figure S6).

We were also interested in studying the chromosomal distribution of some of the repeated sequences found in high copy number in the monomer dataset. The results of our previously described FISH experiments suggest that the specific detection of single nucleotide variations may be difficult to achieve using individual probes but that using several probes in competition may provide the possibility to achieve the required level of specificity. Therefore, we designed new oligonucleotide probes targeting a common region, aiming at distinguishing three different highly repeated sequences with single nucleotide variations (Fig. 6, see Methods). When all probes were used in combination, probes targeting sequence 4 (T39G variation) and sequence 10 (C42G variation) seemed to label all chromosomes, albeit with non-overlapping patterns (see for example insets in Fig. 6f), while probe targeting sequence 5 (A40C variation) was clearly shown to produce a signal on only 8 chromosomes.

Comparison of Cercopithecus solatus alpha satellite families with known primate families

The sequence families defined above may provide information regarding the evolutionary history of alpha satellite DNA in Primates. We were therefore interested in investigating phylogenetic relationships between these families and alpha satellite sequences that were previously described for other primate species. Interestingly, the first alpha satellite consensus sequence ever described, which was obtained for the cercopithecini Chlorocebus aethiops [45], was exactly the same as the consensus sequence of our C1 family, which is also the most abundant repeated sequence in our dataset. This identity suggests the conservation of the C1 family between Cercopithecini species. Although very few sequences were available, a tentative classification was previously proposed for alpha satellite DNA present in Old and New World monkeys, involving five families termed S1 to S5 [17]. We built a phylogenetic tree containing 50 sequences randomly selected within each of our C1 to C4 families and several sequences representative for S1, S2, S4 and S5 (Fig. 7a, see Methods). The S1 sequences obtained from Chlorocebus aethiops were intermingled in this tree with our C1 and C2 sequences. Other sequences classified in S1 but obtained from other species were dispersed in other parts of the graph, suggesting that the proposed S1 family was not relevant. There was also no clear proximity of each one of the C1 to C4 family with sequences belonging to the so-called S2, S4 or S5 family. The phylogenetic tree showed on the contrary that sequences from macaque (identified as S1 or S2) may form a sister group of the C4 family whereas the only available baboon sequence (identified as S1) was close to the C3 family. None of the sequences from macaque or baboon resembled those from our C1 or C2 family. All these results suggest that, contrary to S4 and S5, S1 and S2 do not correspond to alpha satellite families. We also built a phylogenetic tree involving our C1-C4 families and seven families (termed M1, R1-2, V1, and H1 to H4) that were previously identified in human pericentromeric regions, some of them being reported as similar to sequences found in other primates [19] (Fig. 7b). The tree suggests that alpha satellite families found in Cercopithecus solatus have an evolutionary history that is largely independent from that of the alpha satellite families found in human pericentromeric regions.

DNA collection and metaphase preparations

Fibroblast samples of Cercopithecus solatus (ID: 2012–028, male sample, ethic permission n° FR1207510445-I) from the Collection of cryopreserved living tissues and cells of vertebrates (RBCell collection, Muséum national d’Histoire naturelle, Paris) were used for DNA extraction and metaphase preparations. DNA was extracted using the Omega Biotek Tissue DNA Kit (Doraville, USA). Cell cultures and metaphase preparations were achieved according to [59].

Alpha satellite DNA isolation and sequencing

The Serial Cloner software (Serial Basics, serialbasics.free.fr) was used to perform in silico digestions of the Cercopithecini alpha satellite sequences registered as such in Genbank (Accession numbers: AM235889, AM235890, AM237210, AM237214, AM237213, AM237212, X04339, V00145, M26844 and AM237211), which contained both monomers and dimers. The restriction site of the XmnI restriction enzyme (GAANNNNTTC) was observed once in a great proportion of monomers and twice in almost all dimers. XmnI was then used to digest Cercopithecus solatus DNA in vitro. 10 μg of Cercopithecus solatus genomic DNA were digested for 4 h 30 min at 37°C with 60 units of XmnI activity (New England Biolabs) in a total volume of 34 μL. The enzyme was inactivated for 20 min at 65°C. The sample was loaded on a 1% agarose gel after addition of 6.8 μL loading buffer (50% glycerol) and electrophoresis was performed in 0.5X Tris-borate-EDTA buffer, at room temperature for 2 h 45 min at 100 V. The gel was briefly stained with ethidium bromide and then imaged by UV transillumination. Bands corresponding to alpha satellite monomers (~170 bp) and dimers (~340 bp) were cut and DNA was extracted from the gel with the Omega Biotek Gel extraction kit and resuspended in 100 μl of elution buffer. About 220 ng and 110 ng were obtained for the 170 bp and 340 bp samples, respectively.

Sequencing was performed on a PGM sequencing platform (Ion Torrent technology) using the 400 bp sequencing kit. Two libraries were generated using 50 ng of both blunt digest pools and the Ion Plus Fragment Library Kit (4471252, Life Technologies) and tagged with Ion Xpress barcode adapters (4471250, Life Technologies). After purification (1.8X) with Ampure XP Beads (A63880, Agencourt Bioscience, Beverly, USA), the libraries were quantitated using a SsoAdvanced Sybr Green qPCR assay (Biorad, Hercules, USA) based on a custom E. coli reference library. After a dilution of each library down to 26 pM, 0.22 fmol for the 170 bp library and 0.44 fmol for the 340 bp library were pooled as templates for the clonal amplification on Ion Sphere particles during the emulsion PCR, performed on a One Touch2 emPCR robot according to the Ion PGM Template OT2 400 Kit user guide (4479878, Life Technologies). The amplification products were loaded onto an Ion 316v2 chip (4483324, Life Technologies), and subsequently sequenced according to the Ion PGM Sequencing 400 Kit user guide (4482002, Life Technologies). After standard filtration of the raw reads (polyclonal and low quality removal), the Ion Torrent sequencing yielded 204,990 sequences for the 170 bp pool and 353,683 sequences for the 340 bp pool. They were deposited in the NIH Short Read Archive (SRA accession numbers SRX1595681 and SRX1595679).

Alpha satellite sequence filtering

All sequences with an average Phred score lower than 25, a length outside the range 162–182 bp for monomers and 324–364 bp for dimers, and sequences without the XmnI digested sites at the extremities (5′-NNTTC … GAANN-3′) were not considered for further analysis. Alpha satellite sequences were identified with a BLAST search against a reference alpha satellite sequence of Chlorocebus aethiops (AM23721) [60]. Using default BLAST parameters, all sequences exhibiting a hit longer than 80 bp for monomers and 160 bp for dimers were considered as alpha satellite sequences and conserved for the following analysis. All sequences were then reoriented if necessary in order to match the orientation of the reference alpha satellite sequence. The orientation information was preserved for investigations regarding reading biases.

Processing of dimeric sequences was performed as follows. When an XmnI site was present in the middle of these sequences, it was used for separating both monomers, providing the so-called left and right monomers located on the 5′ side and on the 3′ side of the sequence, respectively. Dimers that did not contain any XmnI site in the middle were aligned against a synthetic sequence formed by two consecutive copies of the reference sequence using the Needleman-Wunsch algorithm [61] to identify the monomer limits and split them into left and right monomers according to the same rule as described above. All pairs with at least one monomer outside the 162–182 bp range were discarded. Pairing information was conserved to study association between left and right monomers.

Alpha satellite sequence characterization

Monomeric sequences were compared using their 5-mer composition in order to identify putative alpha satellite groups without direct alignment. For each set of monomers, the 5-mer frequency table was analyzed using a principal component analysis (PCA) to reduce the space complexity and enable data visualization on the first factorial planes. Sequences were classified into groups by using a hierarchical clustering method (HCA) based on the Ward criterion [62] applied to the Euclidean distances calculated from the 100 first principal components of the PCA. Because of the size of the monomer dataset, direct classification of the sequences using HCA was not possible. Instead, HCA was applied on 2,500 randomly selected sequences which were used to train a linear discriminant model. This model has been finally used to classify all the other monomers. The dimer dataset was analyzed in two different ways: 1) monomers extracted from dimers without XmnI sites were classified by using an HCA based on a PCA, 2) monomers extracted from dimers with a XmnI site have been classified by using a LDA trained to recognize the C1-C4 groups.

Because of the size of the datasets, the phylogenetic trees, the consensus sequences and the sequence distance analysis were conducted with different subsets of randomly selected sequences, using a homemade python script. The selected sequences were aligned using MUSCLE [63] and analyzed with SeaView [64]. The phylogenetic trees were built by using the Neighbor joining algorithm and the Kimura 2-parameters distance. Reliability of nodes was assessed using 100 bootstrap iterations. The relatively low bootstrap values observed in the trees can be explained by a limited number of family specific sites, i.e., the informative sites, into the alignments. Nevertheless, the same clustering of the families and the same relationship between these families have been observed with all the trees generated with different randomly selected sequences.

CENP-B and pJalpha boxes were searched with the patterns TTCGTTGGAARCGGGA and TTCCTTTTYCACCRTAG respectively [40] by using the program Fuzznuc [65] and allowing 2 mismatches. All statistical analyses were conducted with R [66]. Our R scripts and other programs are available upon request.

The S1-S5 monomers used in Fig. 7a have been isolated from the sequences described in [17]. All these monomers have been extracted by using the homologous position of the XmnI digestion site as a starting point (XmnI phase) in order to be aligned with the monomers of Cercopithecus solatus. Unfortunately, no full length S3 monomer was available in this phase. To obtain the Genbank accession numbers and the alignment of the used S1-S5 monomers, see [17] and Additional file 3: Text S1. Human monomers from old and ancient families (M1, R1-2, V1, H1-H4) used in Fig. 7b have been isolated from the human Xp chromosome sequence (Genbank ID NT_011630) by using the homologous position of the XmnI digestion site as a starting point. Monomers have been assigned to a family according to their location along the sequence and the annotations provided in [19] (see Additional file 4: Text S2 for alignment).

Oligonucleotide probes

Short oligonucleotide probes (18 nucleotides in length) were designed in order to target specifically the different alpha satellite families identified in Cercopithecus solatus, by systematic prediction of binding frequencies based on the sequencing results. In some instances, when the 18-mer sequence did not allow forming at least 7 GC bp upon hybridization to the complementary strand, length was increased to 19. Sequences and binding frequencies are available in Additional file 1: Figure S3, which also provides details about the positions of locked nucleic acid (LNA) modifications in the probes. These positions were selected based on previous experience in order to achieve a good binding affinity and specificity [44]. When possible, we selected probes that were perfectly complementary to more than 20% of the sequences from the target group and to less than 3% of the sequences from the other groups. Additional file 1: Figure S3 also provides the expected binding frequencies if hybridization is possible despite the presence of one mismatch between the probe and its target. To target three sequences found in high copy number in the monomer dataset, we designed four LNA-modified probes (LNA are written in lower case and classic nucleotides are written in upper case): probe T39G (5′TgTtCtGtTCaTtCaTcTc3′, 5′AlexaFluor488), probe A40C (5′TgTtCtGtGAaTtCaTcTc3′, 3′Digoxygenin), probe C42G (5′TgTtCtCtTAaTtCaTcTc3′, 3′Biotin) and probe TACco (5′TgTtCtGtTAaTtCaTcTc3′) which is complementary to the C1 consensus sequence. LNA-modified probes were purchased from Eurogentec (Seraing, Belgium).

FISH experiments

FISH were performed on metaphase chromosome preparations. Hybridization solutions were prepared by diluting the oligonucleotide probes to a final concentration of 0.1 μM in a hybridization solution consisting of 2X SSC pH 6.3, 50% deionized formamide, 1X Denhardt solution, 10% dextran sulfate, and 0.1% SDS. 20 μL of the hybridization solution were deposited on each slide and covered with a coverslip. The slides were then heated for 3 min at 70°C and hybridized for 1 h at 37°C in a Thermobrite apparatus (Leica Biosystems). Then, each slide was washed twice in 2X SSC at 63°C. Preparations were then incubated in blocking solution (4% bovine serum albumin (BSA), 1X PBS, 0.05% Tween 20) for 30 min at 37°C to reduce nonspecific binding. Then, depending on the combination of probes, the following antibodies were used for subsequent revelations: Alexa 488-conjugated streptavidin (1:200; Life Technologies, Foster City, USA), Cy5-conjugated streptavidin (1:200; Caltag Laboratories, Burlingame, USA), FITC-conjugated sheep anti-digoxigenin (1:200; Roche, Lewes, UK), and Rhodamine-conjugated sheep anti-digoxigenin (1:200; Roche). All antibodies were diluted in blocking solution containing 1X PBS, 0.05% Tween 20, and 4% BSA. Antibody incubation lasted for 30 min at 37°C. All washings were performed in 2X SSC, 0.05% Tween 20. Chromosomes were counterstained with DAPI (4′,6-diamidino-2-phenylindole) by pipetting 40 μL of a 5 μg/mL solution onto the slides, incubating for 5 min and then briefly washing in 1X PBS. Slides were mounted by adding a drop of Vectashield Antifade Mounting Medium (Vector Laboratories, Burlingame, USA) and covering with a coverslip.

Image acquisition and analysis

Metaphases were imaged using an Axio Observer Z1 epifluorescent inverted microscope (Zeiss) coupled to an ORCA R2 cooled CDD camera (Hamamatsu). The Axio Observer Z1 was equipped with a Plan-Apochromat 63x 1.4 NA oil-immersion objective and the following filters set: 49 shift free for DAPI (G365 / FT395 / BP445/50), 38 HE shift free for FITC/Alexa488 (BP470/40 / FT495 / BP525/50), homemade sets for Rhodamine (BP546/10 / FF555 / BP 583/22) and for Cy5 (BP643/20 / FF660 / BP684/24). The light source was LED illumination (wavelengths: 365 nm, 470 nm or 625 nm) except for Rhodamine, for which a metal halide lamp HXP120 was preferred. Immersion oil of refractive index 1.518 at 23°C was used. Color-combined images were reconstructed using ImageJ [67]. At least ten metaphases were visualized for each experiment, which all confirmed the described patterns.