Summary statistics by chromosome category are shown in Table 1 (with previously reported ZAL2/ZAL2^m data included for reference) and detailed statistics for each locus can be found in Additional file 1. In total, we sequenced ~10 kb from 27 autosomal loci (nine pairs of BAC-end sequences and nine additional loci), as well as ~6 kb from the 12 sex-linked loci, three of which were from the W chromosome (1.7 kb). The majority of the sequence was intergenic or intronic with a small fraction of protein coding positions (~2%), and thus likely to reflect neutral levels of genetic variation. For autosomal and Z-linked loci, we sampled 9 - 12 birds with a minimum of four TS and four WS birds. For W-linked loci we sampled 23 - 24 females, including at least 14 TS and nine WS birds.

To understand patterns of diversity as they relate to chromosome size, we grouped our autosomal data into three categories: macrochromosomes, intermediate chromosomes, and microchromosomes. Based on the 267 identified polymorphic autosomal sites (excluding three tri-allelic SNPs), we observed a trend of increasing diversity with decreasing chromosome size (Figure 1). Overall, the amount of nucleotide polymorphism across the sparrow genome is highly variable. The nucleotide diversity across the classes of autosomes varied up to 8-fold, with the lowest diversity observed on the macrochromosomes (π ± SD = 1.98 × 10^-3 ± 3.3 × 10^-4) and the highest diversity on the microchromosomes (π ± SD = 1.59 × 10^-2 ± 1.0 × 10^-3, Figure 1B). In order to quantify the correlation between chromosome (autosome) size and genetic diversity, we used all the individual estimates of nucleotide diversity and predicted chromosome sizes based on synteny with the assembled zebra finch genome to calculate Spearman's rho, a non-parametric measure of statistical dependence, for these two variables. By this measure there was a statistically significant negative correlation between nucleotide diversity and chromosome size (rho = -0.54, p < 0.005, see also Additional files 2 and 3).

To provide independent support for the relationship we observed between nucleotide diversity and chromosome size, we calculated the pairwise divergence between available pairs of complete BAC clone sequences representing alternative haplotypes from three macrochromosomes and a microchromosome. Consistent with our results from the sequencing of multiple small loci in multiple individuals, nucleotide diversity was lower on the macrochromosomes than the sampled microchromosome. Specifically, after excluding low-quality sites and annotated protein coding exons and UTRs, the π values for the macrochromosomes were 0.00120 ± 0.0006 (145,536 sites, orthologous to chicken chromosome 1[GenBank:AC235523 and AC236253]), 0.00172 ± 0.00086 (142,469 sites, orthologous to chicken chromosome 2 [GenBank:AC237008 and AC236908]), 0.00081 ± 0.0004 (143,722 sites, orthologous to chicken chromosome 5 [GenBank:AC236607 and AC237119]), whereas the π on the sampled microchromosome was 0.01632 ± 0.00816 (106,615 sites, orthologous to chicken chromosome 17, [GenBank:AC235993 and AC235934]). Thus, in this small independent data set the highest levels of nucleotide diversity were observed on a microchromosome.

For the sex chromosomes we identified 12 Z-linked SNPs and one polymorphic site on the W chromosome. As expected, the Z and W sex chromosomes showed reduced diversity compared to the sampled autosomal loci (π ± SD = 5.0 × 10^-4 ± 9.0 × 10^-5 and 5.0 × 10^-5 ± 5.0 × 10^-5, respectively; see Table 1 and Figure 1) and were lower than all three classes of autosomes. Our estimate of π for the W chromosome was 10-fold lower than that of the Z chromosome and between 40- and 300-fold lower than the diversity on the autosomes, depending on which chromosome type was used for comparison. Note that the high variance in the estimate of π for the W chromosome was due to the observation of just a single SNP.

Within the inversion, nucleotide diversity did not differ significantly between the ZAL2 and ZAL2^m (Mann-Whitney U = 1603.0, N₁ = 58; N₂ = 58, p = 0.575), nor did it differ from that of either sex chromosome (ZAL2/Z, MWU = 190.0, N₁ = 58; N₂ = 9, p = 0.123; ZAL2^m/Z, MWU = 175.0, N₁ = 58; N₂ = 9, p = 0.051; ZAL2/W, MWU = 82.0, N₁ = 58; N₂ = 3, p = 0.836; ZAL2^m/W, MWU = 86.0, N₁ = 58; N₂ = 3, p = 0.965). Because of the limited number of loci sampled on the sex chromosomes compared to the ZAL2 and ZAL2^m, the lack of a significant difference between the inverted interval and the sex chromosomes should be interpreted with caution. Nucleotide diversity within the inversion on both the ZAL2 and ZAL2^m was, however, significantly lower than observed on other macrochromosomes (ZAL2: Mann-Whitney U = 123.5, N₁ = 58; N₂ = 12, p < 0.001; ZAL2^m: MWU = 94.5, N₁ = 58; N₂ = 12, p < 0.001).

We previously reported extreme genetic differentiation between the ZAL2 and ZAL2^m chromosomes as a result of suppressed recombination between the alternate chromosome types 22 23 , which is also apparent in comparisons of TS and WS individuals. In order to compare this genetic differentiation to that present elsewhere in the genome, we calculated F_ST between the groups of TS and WS birds at all the sampled loci (Figure 2 and Additional file 1). Most F_ST values for the autosomal and sex-linked loci clustered around zero, and we found no significant signal of population structure across the genome except within the ZAL2/2^m inversion interval (Figure 2). Additionally, none of the haplotype networks constructed from individual loci unlinked to the ZAL2/2^m polymorphism revealed any clear indication of population structure (data not shown), indicating the genetic differentiation observed between TS and WS birds within the ZAL2/2^m inversion is exceptional with respect to the rest of the genome.

Tests based on allele frequency distributions can yield insights into demographic, population genetic and evolutionary processes acting upon a particular genomic region. Previously, we reported a skew towards intermediate frequency alleles within the ZAL2/ZAL2^m system and a positive Tajima's D 22 23 . To establish whether this pattern was unique to the ZAL2/ZAL2^m system within the sparrow genome, we calculated Tajima's D for all sampled loci (Figure 3, Additional file 1). As with the F_ST values, the Tajima's D associated with the ZAL2/ZAL2^m chromosomal polymorphism was an outlier compared to the other regions of the white-throated sparrow genome. No Tajima's D values were found to be statistically different from neutral expectations, owing greatly to the conservative nature of the statistic. However, a one-tailed t-test demonstrates that the Tajima's D value for the ZAL2/2^m system is statistically different from values reported elsewhere in the genome (t = 27.72, p < 0.0001).

Recombination is expected to reduce LD and, in general, LD decreases with increasing distance between sites. To compare LD associated with the ZAL2/2^m chromosomal polymorphism to that observed in the rest of the genome, we pooled r² values from informative pairwise comparisons within autosomal loci, plotted their distribution as a function of distance (Figure 4A) and calculated their averages (Additional file 4). The r² values from our autosomal data were generally low, as measured by ZnS = 0.14, a summary statistic which is based on the average correlation between pairs of sites 29 . Among the autosomal sites, LD decayed rapidly such that only limited LD was observed even within 500 bp (Figure 4A and 4B). Consistent with the overall low levels of LD, the four-gamete test revealed evidence for recombination across all our autosomal BAC-paired end loci (data not shown). Note that due to the limited number of informative pairwise comparisons on the Z (n = 3), no similar analyses could be performed on that chromosome. In contrast, we previously observed perfect LD between the majority of pairwise comparisons within the ZAL2/ZAL2^m inversion, independent of distance between sites and extending to > 100 Mb (Figure 4C and 4D). Thus, the strong LD associated with the chromosomal polymorphism was exceptional compared to other white-throated sparrow chromosomes.

On average across all sparrow autosomes, one segregating site is observed every ~180 bp (π = 5.62 × 10^-3 ± 4.7 × 10^-4). This estimate is similar to the diversities calculated in other natural populations of passerine birds, such as Ficedula flycatchers, π = 2.7 - 3.6 × 10^-3, 20 30 , Carpodacus finches, π = 5.7 - 8.5 × 10^-3 31 , the great reed warbler, π = 1.2 × 10^-3 32 and the blue tit, π = 1.8 × 10^-3 32 . Nucleotide diversity across the sparrow autosomes varies substantially, spanning a range of nearly two orders of magnitude between individual loci (π = 3.0 × 10^-4 - 2.9 × 10^-2). The microchromosome group on average was associated with the highest level of nucleotide diversity and was 8-fold higher than the macrochromosomes. However, as pointed out in the Results section, the level of nucleotide diversity was not perfectly correlated with chromosome size. For example, the diversity of the loci orthologous to microchromosomes 24 and 28 were comparable to that observed on the macrochromosomes. This result could be attributed to a number of factors, such as local variation in recombination rates, or may reflect inaccurate assignment of these white-throated loci to chromosome class sizes caused by differences in genome organization between this species and the zebra finch or chicken.

The negative correlation between nucleotide diversity and chromosome size class observed in our study was previously predicted for avian genomes based on two well-established observations. The first is that the rate of recombination on avian microchromosomes can be 5-10 times greater than on macrochromosomes 9 10 . The second observation is that regions of high recombination tend to harbor greater genetic variation 7 . Nevertheless, previous studies that tested for this correlation in nucleotide diversity in the chicken genome reported conflicting results 13 14 33 . Future studies will therefore be necessary to establish whether the negative correlation between nucleotide diversity and chromosome size we observed for the white-throated sparrow autosomes is a general characteristic of avian genomes.

As expected, the white-throated sparrow sex chromosomes showed the lowest levels of genetic diversity. The neutral theory predicts a ratio of 1:3:4 for the genetic diversities of the W, Z and autosomes based on their relative N_es. Z to autosome diversity (π_Z : π_A) is therefore expected to equal 0.75 21 . However, because sex chromosomes are transmitted differently from autosomes, the standard models of how mutation, selection, drift, demography and mating system influence population genetic patterns may not apply 15 . The combined effects of increased HRI and lowered N_e due to variation in male mating success can reduce Z chromosome diversity relative to autosomal diversity 20 21 . Indeed, reports of avian π_Z : π_A are lower than the expected 0.75: in Ficedula flycatchers π_Z: π_A ≈ 0.4 20 and in chicken π_Z: π_A ≈ 0.25 21 . However, because those studies did not indicate the chromosome size class of the autosomal loci used for comparison, the degree to which the π_Z: π_A is skewed is uncertain. Our finding that chromosome size and genetic diversity are negatively correlated suggests that it is appropriate to consider Z chromosome diversity in reference to similarly sized chromosomes (macrochromosomes). In the white-throated sparrow the π_Z: π_A ≈ 0.08 when comparing Z chromosome diversity to average diversity across all autosomes. However, when comparing Z chromosome diversity to average macrochromosome diversity, π_Z: π_A ≈ 0.25, which highlights the importance of considering chromosome type when calculating this ratio.

Under the neutral model, the W chromosome-to-autosome diversity is predicted to be 1:4 or 0.25. However, studies of the avian W chromosomes report far lower diversity values. For example, in a survey of ~3.4 kb from > 150 W chromosomes in seven avian species, Montell et al. 17 did not observe any polymorphisms. Berlin and Ellegren 16 identified a single segregating site in a survey of ~8 kb in 47 chickens from divergent breeds and estimated π_W = 7 × 10^-5, ~1/100 of their π_A estimate 21 . Similarly, we observed low levels of variation on the W chromosome, although it should be pointed out that our estimate of π for the W was based on a single SNP.

Within the white-throated sparrow genome, dramatic HRI is probably not limited to the sex chromosomes. We previously reported patterns of reduced recombination and extensive LD within the ZAL2/2^m system, which bears striking similarities to sex chromosomes 22 23 . Avian macrochromosomes are disproportionately associated with recombination deserts, except close to the chromosome ends 10 . Thus, the low rate of recombination for macrochromosomes is likely to result in intrinsically low diversity values, which are further reduced by HRI on the ZAL2 and ZAL2^m.

Although the ZAL2/2^m system bears many similarities to the XY (ZW) sex chromosomes, including differences in patterns of recombination, disassortative mating, and reduced genetic diversity, the ZAL2/2^m are not sex-linked and, as such, sexual selection 34 , sex-biased demographic history 35 36 , nor mitochondrial linkage 19 can easily explain patterns of evolution within the ZAL2/2^m system. For these reasons, the ZAL2/2^m polymorphism remains an informative point of comparison for the study of population genetic patterns of sex chromosomes and can help distinguish between sex-specific and fundamental molecular evolutionary processes that shape their evolution.

The extent of LD is dependent on two key parameters: N_e and recombination rate 9 37 . While the extent of LD within avian genomes from natural bird populations is largely unknown, the few studies currently available indicate that LD is generally low on avian autosomes as a consequence of high rates of recombination 33 38 39 40 . Megens et al. 33 reported a correlation between LD and chromosome size, such that macrochromosomes exhibit consistently higher LD when compared to microchromosomes. Our autosomal data are consistent with these reports in that on average we found only weak LD even between sites within a few hundred bases of each other, and that the macrochromosomes were associated with the strongest LD (see Additional file 4).

The one striking exception to the general pattern of low LD in the sparrow genome we did observe was within the ZAL2/ZAL2^m inversion. Our previous study of the ZAL2 and ZAL2^m chromosomes revealed extensive LD spanning > 100 Mb, thus representing one of the largest segregating haplotypes ever reported 22 . Given our findings here, as well as the general patterns of low LD reported in other bird autosomes 33 38 39 , we conclude that the LD observed between the ZAL2 and ZAL2^m is unusual within the sparrow genome as well as amongst avian genomes in general.

Our previous studies of the ZAL2/2^m system indicated extensive population structure within the inversion polymorphism 22 23 . The majority of the segregating sites (~70%) were fixed differences and F_ST = 0.94 (p < 0.01) within the inversion between the alternate arrangements 22 . The high proportion of fixed differences between the chromosome arrangements resulted in many intermediate frequency variants, and thus a large and positive Tajima's D value. In this study of autosomal and sex-linked loci, we found no significant population structure between WS and TS birds outside of the ZAL2/2^m inversion. Additionally, the negative Tajima's D values observed for most of the other sampled loci indicate that the effects of balancing selection through disassortative mating are limited to the ZAL2/2^m system.

White-throated sparrows and a female dark-eyed junco (Junco hyemalis) were collected in mist nets on the campus of Emory University in Atlanta, GA during November and December of 2005-2008. Visual examination of the plumage and a PCR-based assay were used to infer the genotype with respect to the ZAL2/ZAL2^m polymorphism for each bird included in the study 41 . Note that the frequency of WS and TS individuals in our locally sampled population is comparable to the overall frequency reported for the species, i.e. 50:50 (DL Maney, unpublished data). The Emory University Institutional Animal Care and Use Committee approved all procedures involving animals.

PCR primers for autosomal and Z chromosomes were designed in Primer3 42 using white-throated sparrow BAC-end sequences that mapped to unique locations in the zebra finch genome (taeGut1) 43 by MEGABLAST searches (-t 16, -N 2, -W 11, -e 1e-30) 44 . For the W-linked loci, we designed primers from a completely sequenced white-throated sparrow BAC from the W chromosome [GenBank: AC236562] containing the CHD1W locus. To avoid potential amplification of gametologous Z chromosome loci, the W chromosome primers were designed to include at least three mismatches compared to the corresponding Z chromosome sequence. A complete list of PCR primers, their orthologous positions in zebra finch, and orthologous chromosome assignments in chicken are listed in Additional file 5. Each 25 μL PCR contained final concentrations of 1X PCR buffer, 1.5 mM MgCl₂, 20 pM of each primer, 0.2 mM of each dNTP, 1.5 U of Taq or Platinum Taq DNA polymerase (Invitrogen) and ~25 ng of genomic DNA. PCR cycling parameters were as follows: 94° for 5 min, 35 cycles of 94° for 30 sec, 55° for 30 sec and 72° for 1 min, followed by 72° for 7 min. Amplicons were subsequently purified and directly sequenced (Sanger sequencing) [GenBank: GU449819 - GU450219, GU450273 - GU450309 and HM126554 - HM126573].

Nucleotide polymorphisms were automatically called using SNPdetector 45 and manually confirmed prior to further analyses. Annotation of the gene features for each locus was based on the annotation of orthologous zebra finch genomic segments (taeGut1) 43 . Insertion-deletion polymorphisms and sites with more than two segregating alleles were excluded from our analyses, as these events were relatively infrequent and could not be readily incorporated into the subsequent analyses. Like most other avian genomes, the white-throated sparrow karyotype consists of macro-, intermediate and microchromosomes 24 . Given that synteny is highly conserved among even distantly related birds, such as the zebra finch and chicken 43 , we predicted linkage and corresponding chromosome size in the white-throated sparrow based on orthology to the zebra finch and chicken genomes. To examine population genetic patterns between differently sized autosomes, we grouped our sequence data by chromosome type according to the classification convention of the International Chicken Genome Sequencing Consortium that divides chicken autosomes into three size classes: macrochromosomes (GGA 1-5), intermediate chromosomes (GGA 6-10) and microchromosomes (GGA 11-38) 9 .

Where we had sequence from paired BAC ends, genotypes were concatenated and phased as a single locus using Phase v2.1.1 46 . Haplotypes were then split into individual loci for population genetic analysis, except for calculation of LD and construction of haplotype networks. Analysis of polymorphism and tests of neutrality based on allele frequency spectrum, (Tajima's D and Fu and Li's statistics) were performed in DnaSP v5.1 47 .

Using Splitstree v4.10 48 , we generated haplotype networks using the neighbor-joining algorithm and J. hyemalis sequence as an outgroup. Haplotype networks were constructed for individual loci and, when possible, for the concatenated paired BAC-ends. To quantify population structure, F_ST values were calculated in Arlequin v3.11 49 and statistical significance was assessed with exact tests for genetic differentiation. In order to make a direct comparison between F_ST within the ZAL2/ZAL2^m system and the autosomal and sex-linked loci reported here, we calculated F_ST between the TS and WS groups. LD and r² values for all pairwise comparisons between informative sites, including paired BAC-end sequences when possible, were calculated with Haploview v4.2 50 . In addition to inferring recombination from levels of LD, we identified four-gamete pairs within all loci using DnaSP v5.1 47 and estimated the average haplotype block size using the four-gamete rule in Haploview v4.2 50 .

In order to compare levels of diversity among the ZAL2/2 m, sex chromosomes, and macrochromosomes we performed an overall omnibus test followed by planned pairwise comparisons. Because the data set contained many zeros, the data could not be normalized by transformation. We therefore used a nonparametric approach (Kruskal-Wallis ANOVA followed by Mann-Whitney U tests).