Genomic Variation Among and Within Six Juglans Species

Genomic analysis in Juglans (walnuts) is expected to transform the breeding and agricultural production of both nuts and lumber. To that end, we report here the determination of reference sequences for six additional relatives of Juglans regia: Juglans sigillata (also from section Dioscaryon), Juglans nigra, Juglans microcarpa, Juglans hindsii (from section Rhysocaryon), Juglans cathayensis (from section Cardiocaryon), and the closely related Pterocarya stenoptera. While these are ‘draft’ genomes, ranging in size between 640Mbp and 990Mbp, their contiguities and accuracies can support powerful annotations of genomic variation that are often the foundation of new avenues of research and breeding. We annotated nucleotide divergence and synteny by creating complete pairwise alignments of each reference genome to the remaining six. In addition, we have re-sequenced a sample of accessions from four Juglans species (including regia). The variation discovered in these surveys comprises a critical resource for experimentation and breeding, as well as a solid complementary annotation. To demonstrate the potential of these resources the structural and sequence variation in and around the polyphenol oxidase loci, PPO1 and PPO2 were investigated. As reported for other seed crops variation in this gene is implicated in the domestication of walnuts. The apparently Juglandaceae specific PPO1 duplicate shows accelerated divergence and an excess of amino acid replacement on the lineage leading to accessions of the domesticated nut crop species, Juglans regia and sigillata.

Here, we add to the existing J. regia reference genome of the widely planted Chandler cultivar (Martínez-García et al. 2016) by describing draft nuclear reference genomes for five additional members of Juglans (J. nigra, J. hindsii, J. microcarpa, J. sigillata, Reduction of PPO function has been implicated in domestication associated grain colors of three species of Asian rice (Yu et al. 2009), barley (Taketa et al. 2010) and foxtail millet (Inoue, et al. 2015). PPO genes are also implicated in the plant defense response (Thipyapong et al. 2004;Li and Steffens 2002;Richter et al. 2012). The number of PPO genes varies from zero in Arabidopsis (Tran et al. 2012) to 19 in Salvia miltiorrhiza, an important ingredient in traditional Chinese medicine . Previously, genomic resources established that J. regia actually has two PPO genes, JrPPO1 was the first to be found and characterized (Escobar et al. 2008), while the complete genome revealed the presence of another gene, expressed at much lower levels (JrPPO2) and in a narrow range of tissue types (Martínez-García et al. 2016). These genomes are used to more broadly investigate the origins and evolutionary relationship of PPO genes in Juglans. The recently solved structure of JrPPO1 provides a unique perspective to evaluate how evolutionary forces may have influenced protein function.
A detailed and thorough comparative analysis of the genomes of these species is beyond the scope of this paper, but to foster the application of genomics in Juglans research and breeding, we report here the sequencing and assembly of the genomes of five additional Juglans species. Second, we provide accessible pairwise alignments of these genomes annotating synteny and between-species divergence. Finally, an important adjunct to the genome structures and divergences revealed in these resources are surveys of the within species genomic polymorphisms for four species. Beyond the obvious value of detected SNPs as a resource for the development of genotyping tools, genomic variants can be annotated with respect to their potential phenotypic consequences (Cingolani et al. 2012;McLaren et al. 2016), thus serving as a readily accessible source of candidates in functional genomic analyses and gene-oriented breeding and biotechnology. We report the resequencing of samples of independent accessions of two Juglans species with scion breeding programs, J. regia and J. nigra, as well as two species involved in rootstock development, J. hindsii and J. microcarpa. Species differences in the overall levels of genomic polymorphism are documented. Polymorphism is used in conjunction with divergence to infer recent selection possibly associated with domestication. Finally, we demonstrate how these genomic resources can be visualized in support of gene-oriented analyses by employing the widely used and well-supported JBrowse software (Skinner et al. 2009).

DNA Extraction
Nuclei were isolated from adult leaves of each species as previously described (Zimin et al., 2014). Nuclei were lysed by adding N-laurylsarcosine to a final concentration of 1% (w/v) and incubating for 15 min at room temperature. 5M NaCl and 10% (w/v) cetyltrimethyl ammonium bromide (CTAB) were added to final concentrations of 0.7M and 1% (w/v) respectively and the mixture was incubated at 60°for 30 min. DNA was then extracted twice with an equal volume of chloroform:isoamyl alcohol (24:1), precipitated with 2/3 volume of 100% isopropanol, and re-suspended in TE buffer.

Library construction and Sequencing
Two types of libraries were prepared from the resulting DNA as follows (using enzymes and buffers from New England Biolabs unless otherwise indicated): Paired end Libraries: DNA (5 mg) was fragmented by sonication in a Diagenode Bioruptor NGS instrument (high power setting, 9 cycles of 15 sec on, 90 sec off). Fragments were end-repaired in a 100 ml reaction in 1X T4 ligase buffer containing 0.4 mM (each) final concentration of dNTPs, 15 U T4 DNA polymerase, 50 U T4 polynucleotide kinase and 5 U DNA polymerase I large (Klenow) fragment. End-repaired fragments were A-tailed in a 50 ml reaction in 1X NEBuffer 2 containing 0.2 mM final concentration of dATP, and 15 U of Klenow fragment (exo-). Paired-end adapter was prepared by heating an equimolar mixture of two HPLC-purified oligos (59-ACACTCTTTCCCTACAC-GACGCTCTTCCGATOT and 59-PHO-GATCGGAAGAGCACAC-GTCT where 59-PHO indicates 59 phosphorylation and O indicates C with a phosphothioate linkage to the next base on the 39 side) in a tube immersed in $500 ml of boiling water and then leaving the tube immersed while the water cooled slowly to room temperature. Annealed adapters were ligated to A-tailed fragments in a 50 ml reaction in 1X Quick Ligation buffer containing 3 mM final concentration of paired-end adapter, and 5 ml of Quick T4 ligase. Adapter-ligation product was size-selected on a 2% agarose gel in 1X TAE run until the bromophenol blue band had migrated approximately 9 cm. Two $1 mm-thick slices were then cut from the gel under blue light transillumination at a position approximately adjacent to the 500 bp ladder band and DNA was extracted from each slice using the MinElute Gel Extraction kit (Qiagen). Concentration of the recovered DNA was estimated using an Agilent Bioanalyzer 2100, and 10 ng of DNA from each slice was used as template in a 50 ml PCR reaction in 1X KAPA HiFi HotStart Ready-Mix (KAPA Biosystems) containing PAGE-purified barcoded primers at 0.5 mM each (forward primer 59-AATGATACGGCGACCACCGAGATC-TACACTCTTTCCCTACACGACGCTCTTCCGATOT and reverse primer either 59-CAAGCAGAAGACGGCATACGAGATG-TAGCCGTGACTGGAGTTCAGACGTGTGCTCTTCCGATOT or 59-CAAGCAGAAGACGGCATACGAGATTACAAGGTGACTG-GAGTTCAGACGTGTGCTCTTCCGATOT where O indicates C with a phosphothioate linkage to the next base on the 39 side). Cycling parameters were 5 min at 95°followed by 10 cycles of 20 sec at 98°, 30 sec at 65°, and 30 sec at 72°, followed by a final 5 min extension step at 72°. DNA purifications following all reactions listed above were performed using PCRClean DX beads from Aline Biosciences according to the manufacturer's instructions.
For J. nigra and J. sigillata an additional short fragment library per species was made using the Illumina TruSeq DNA PCR-Free Sample Prep Kit following the manufacturer's instructions for 550 bp target insert size.
Mate pair Libraries: DNA (15 mg) was treated with 5 ul of PreCR Repair Mix (New England Biolabs) in a 450 ul reaction in 1X Thermo-Pol buffer containing 0.1 mM (each) final concentration of dNTPs and 0.5mM final concentration of NAD+. DNA was then purified by one extraction with phenol/chloroform/isoamyl alcohol (25:24:1) and one extraction with chloroform followed by ethanol precipitation. 4 mg of PreCR-repaired DNA was used as input for the Nextera Mate Pair Sample Preparation kit (Illumina) following the manufacturer's "gel plus" protocol. Size selection was performed with a BioRad FIGE Mapper using a buffer re-circulating pump and the following conditions: 1X TAE buffer; 16 hr run at room temperature; 4.1 V/cm forward and 2.7 V/cm reverse field strength, both with linear ramping from 0.1 sec initial to 0.8 sec final switch time. Gel slices were cut from the gel adjacent to ladder bands at $3kb, $6 kb and $10 kb. Circular ligation products were fragmented by sonication in a Diagenode Bioruptor NGS instrument (high power setting, 5 cycles of 15 sec on, 90 sec off). Fifteen cycles of enrichment amplification were performed.
Completed libraries were pooled as necessary and sequenced in paired-end Rapid Run mode on a HiSeq 2500 (Illumina). Read lengths were 151 bp forward read and 151 bp reverse read. Sequencing results for each library are given in Table S1.

Assemblies
Genomes were assembled following the method used in (Martínez-García et al. 2016). Scaffolds originating from uncollapsed heterozygous sequence were expected. We partially addressed this issue by identifying and completely removing nested redundant scaffolds as follows: we aligned all scaffolds shorter than 50 Kb to each other using bwa (Li and Durbin 2009) and MUMmer (Kurtz et al. 2004), and used the show-coords program within MUMmer to identify scaffolds that were completely contained by and nearly identical to other, longer scaffolds.
Pacbio sequencing and re-assembly of J. regia Light coverage of longer reads (PacBio) was obtained in an effort to improve the existing J. regia genome assembly. DNA was extracted from adult leaves from the same tree used for the original J. regia genome (Martínez-García et al. 2016) and subsequently converted into a sequencing library using the method previously described (Zimin et al. 2017). In total, 814,584 PacBio sequence reads were obtained totaling 6 billion bp ($10 fold coverage of the genome). These were combined with previous Illumina data and assembled using the MaSuRCA hybrid assembly method described in Zimin et al. (2017). The resulting unannotated assembly is included here as an additional resource and to confirm results inferred from micro-synteny to the original assembly.

k-mer Analysis
For each species, 31-mer histograms were computed using the software jellyfish (Marçais and Kingsford 2011) on the paired end Illumina reads. The command jellyfish 'count (-m 31 -s 1G-bf-size 200G)' was used to generate each database, and jellyfish histo was used to compute each histogram. The program jellyfish query (-s) was used to obtain the depth of 31-mers for specific sequences. Custom scripts were used to perform the set operations required for the PPO depth analysis. Genome sizes were estimated from 31-mer histograms using the method described in (Sork et al. 2016).
Pairwise genome alignment Pairwise genome alignment was conducted with the nucmer nucleotide alignment software component of the mummer v4.0 software package. Each genome participated as a query and as a reference in the alignments. Genome alignment was conducted using the command 'nucmer-prefix = ref_qry ref.fasta qry.fasta' recommended for aligning draft genomes to draft genomes. Alignments were then filtered using the nucmer command 'delta-filter' to select the best weighted set of non-overlapping alignments to the query that cover the reference sequence. Alignments were then processed using the 'show-coords' command with a minimum reference sequence length of 1000 bp. Genome wide coverages of the alignments were calculated as the number of aligned query bases over the total number of bases in the reference genome over 1000bp. Genome wide divergences were calculated as the number of mismatches, from the nucmer percent identity, over the number of aligned query bases.

Core Gene Annotation and PPO Analysis
To annotate a "core" set of genes expected to be present in each assembly, version 2.5 of the Core Eukaryotic Genes Mapping Approach (CEGMA) (Parra et al. 2007) and version 2.0 of Benchmarking Universal Single-Copy Orthologs (BUSCO) (Simão et al. 2015) were run on each genome separately. CEGMA was run using default parameters and BUSCO was run in 'geno' mode using the 'embryophyta_odb9' profile.
The PPO genes annotated using GMAP (Wu and Watanabe 2005) to align the two available J. regia PPO1 and PPO2 sequences (Martínez-García et al. 2016) to each of the additional genome assemblies analyzed in this paper. The inferred nucleotide sequences for each copy were obtained from the alignments. A few genomes contained more than one copy of a gene, in each case, a copy confirmed by synteny (on the same scaffold) was considered the ortholog.

Phylogenetic tree construction
To construct a phylogenetic tree from pairwise genome alignments, estimates of pairwise divergence were calculated from the average genome alignment divergence values and corrected using the method of Jukes and Cantor (1969). The unrooted tree was then constructed using the Neighbor-Joining method (Saitou and Nei 1987) as implemented by the 'neighbor' program in the 'Phylip' package (Felsenstein 1986). The outgroup P. stenoptera was excluded from this method due to the low coverage observed in the pairwise alignments.
Genome wide phylogenetic trees were also constructed using the filtered and curated multiple alignments of single copy BUSCO genes present in all seven species. For each BUSCO gene annotated as single copy in all seven species, multiple alignments of both protein and nucleotide sequences were constructed with the software MUSCLE (Edgar 2004). Gap columns were subsequently filtered for phylogenetic reconstruction, and if an alignment consisted of 50% or higher gap columns, it was completely filtered. Filtered alignments were concatenated, resulting in a total input dataset of 373,615 sites. The phylogenetic tree was inferred by Maximum Likelihood using PhyML (Guindon and Gascuel 2003). To test for phylogeny 100 bootstrap replicates were performed.
Divergence times for all branching points in the topology were calculated using the Maximum Likelihood method based on the model of Tamura and Nei (1993). A chronogram was subsequently estimated using the method of Tamura et al. (2012) using the Juglans crown group ancestor as the calibration point. The 95% confidence intervals shown were computed using the method described in Tamura et al. (2012).
A gene tree for the PPO family was constructed from the orthologous nucleotide sequences annotated in each genome. From the nucleotide sequences a multiple sequence alignment was created using MUSCLE (Edgar 2004). A maximum likelihood phylogenetic tree was constructed using PhyML (Guindon and Gascuel 2003) using 100 bootstrap replicates. Lineage specific K a and K s values were then estimated using this tree using the method of (Zhang et al. 2006).

Single Nucleotide Polymorphisms
Reads were aligned to the genome using bwa mem (Version 0.7.13; Li and Durbin 2009). Aligned reads were subsequently converted into bam format using samtools view (Version 1.3; ). The bam-formatted files were sorted using samtools sort. A multi-sample vcf file was then produced for SNP calling using samtools mpileup. SNPs were called using the multi-sample allele calling algorithm of bcftools call (Version 1.2; Li 2011) with reads from each sample distinguished. SNP filtering was performed using bcftools view. To classify a subset of diploid sites, histograms of aligned read depth were used to inform the minimum and maximum coverage cutoffs for a SNP in each species (Table S2). Nucleotide diversity p (Nei and Li 1979)

Estimation of genome size and relative heterozygosity
For each target genome, we deeply sequenced paired end and mate-pair libraries (Table 1; Table S1). Every base of each genome was represented an average of 100 times in the whole genome shotgun Illumina sequence data from paired end libraries. For the very long insert matepair libraries used for linking, we looked at the physical coverage based on the input DNA molecule, which is much longer than the reads. For each genome, we sequenced libraries in long insert size ranges of about 3000, 5000, and 10000 bp, targeting at least 100-fold physical coverage in all cases (Table S1).
Prior to assembly, we characterized the genome using the distribution of all short subsequences of fixed length k (k-mers) in the unassembled paired end Illumina reads. This was performed using k-mer histograms constructed from the paired data using jellyfish (Marçais and Kingsford 2011) with word size (k) of 31. The histograms for each genome (Figure 1), display three distinct peaks. The extreme peak at the origin of the depth axis, representing approximately 1% of the distinct 31-mers, are very rare k-mers in the data. These are attributed to sequence errors. The two peaks of interest, together comprising the largest area of each histogram, reflect the bi-modal distribution expected from a heterozygous diploid genome. The area under the right (deeper coverage) "diploid" peak, represent 31-mers shared between the homologous chromosomes. The area under the left (lower coverage) "haploid" peak, represents 31-mers found in only one of the two homologous chromosomes. Among the genomes, the deeper "diploid" peaks occur consistently at twice the depth as the lower coverage "haploid" peaks. However, different genomes exhibited noticeable variation in the relative proportions of the genome found at haploid depth, reflecting the variation in the level of heterozygosity among the species. The greater number of differences between chromosomes leads to a larger fraction of haploid-depth 31-mers. Estimates of genome size based on the 31-mer histograms, ranged from a low of 498 Mbp for J. nigra to a high of 594 Mbp for P. stenoptera ( Table 2). The only cytological estimate of genome size for Juglans is 606 Mbp for J. regia (Horjales 2003). Within the family Juglandaceae, the other available C-value estimate is a genome size of 808 Mbp for pecan (Carya illinoensis).

Genome assemblies
The draft genome assemblies for the species sequenced and assembled for this paper are characterized in Table 2. The statistics are compared to the original Juglans regia v1.0 assembly. It is notable that variance in assembly size is clearly higher than the variance in estimated genome sizes (Table 1). In particular, there are two instances where the genome was much larger than expected: Juglans microcarpa and Pterocarya stenoptera. These two species are also outliers for a number of additional measures which we describe below. An unannotated v1.5 J. regia assembly, incorporating light coverage of long PacBio sequences for improved contiguity (N50 of 639 kbp vs. 242 kbp), is included as an additional resource and to validate inferences made from the v1.0 J. regia assembly.

Core gene annotation
Both the Benchmarking Universal Single-Copy Orthologs (BUSCO) (Simão et al. 2015) and the Core Eukaryotic Genes Mapping Approach (CEGMA) (Parra et al. 2007) were used to estimate the completeness of the new assemblies. For all of the species, the final assemblies compared favorably with the original J. regia reference genome (Martínez-García et al. 2016) ( Table 3). The CEGMA results are given for 248 eukayotic genes. The most general metric for assembly completeness, the percentage of CEGs annotated as complete or partial ranges from 93.55% (J. nigra) to 96.37% (J. sigillata), with J. regia falling in the middle of that range at 94.76%. The BUSCO analysis is aimed at a more targeted phylogenetic range and includes more genes. The BUSCO results are given for 1440 single copy Embryophyte genes. A comparable measure of completeness, including fragmented annotations, has a slightly narrower range from 94.24% (J. microcarpa) to 96.18% (J. regia). BUSCO additionally estimates the fraction of single copy orthologs that appear duplicated in the target genome. The genomes of J. microcarpa and P. stenoptera also stood out as having high values for this statistic, more than twice the average of the other genomes.
Pairwise Genome Alignments and Genomewide Phylogenies Pairwise alignments were constructed as a resource to evaluate microsynteny and sequence conservation (divergence) between pairs of species. The alignment methodology used was asymmetric and the alignments between a pair of genomes differed depending upon which genome was used as the query and which was the reference. The filtered alignments consisted of a tiling path of aligned segments with respect to the reference genome. The aligned coverage and corresponding divergence estimates for each possible query reference assignment of seven genomes, presented in Table 4, supports the recognized section level classifications of the species (Table 4). Coverage was typically quite high, 80-90% of sites (dark gray), for pairs of Juglans species within the same section, and dropped to as low as 60% (light gray) between Juglans species pairs in different sections. Alignments between Juglans species and the Pterocarya stenoptera outgroup had the lowest coverage. Pairwise divergence estimates varied for the most part inversely with alignment coverage. However, this relationship breaks down for the distant comparisons involving P. stenoptera that have the largest fractions of unaligned bases.
We estimated two classes of genome-wide phylogenetic trees from our data. The most broadly informed tree uses pairwise Jukes-Cantor divergence estimates obtained from the genome alignments given in Table 4. These divergence estimates were calculated from the concatenated forward and reciprocal alignments between pairs of species. We excluded P. stenoptera from this tree because of obvious ascertainment bias; the aligned coverage was much more restricted to the most conserved regions of the genome. The resulting neighbor joining tree (Figure 2a) can be compared to two additional trees constructed using only orthologous single copy genes and incorporating the outgroup. These trees were estimated using the set of BUSCO orthologs present as a single copy in all seven genomes from which a good multiple alignment could be obtained. Concatenated multiple alignments of 244 gene sequences were used to reconstruct the maximum likelihood phylogeny in Figure 2b. When restricted to these highly conserved genes, the total divergence in the genus Juglans was notably much less than captured in the unrooted tree constructed from the pairwise genome alignments. Genome wide phylogenetic trees were consistent with the accepted section level classifications of Dioscaryon, Rhysocaryon, and Cardiocaryon. A chronogram was constructed (Figure 2c) to estimate section level divergence times and compare them to previous results Dong et al. Genomic diversity in J. regia, J. hindsii, J. nigra and J. microcarpa Table 5 shows the numbers of SNPs identified in J. regia, J. hindsii, J. nigra and J. microcarpa and corresponding estimates of the nucleotide diversity. These can be compared to the relative heterozygosity rankings in Table 1. As inferred from the 31-mer analysis (above) J.
n Table 1 Selected targets and the basic statistics for the unassembled genomes. The genome sequences assembled in this study all have 2n = 32 chromosomes. The Illumina sequence reads and the corresponding genome sizes estimates from the 31-mer analysis of the paired end reads are given. Qualitative levels of heterozygosity, 1 = highest 7 = lowest, are based on quantitatively ranking the 31-mer distributions by relative proportion of the two peaks 1 The genome size estimates for these the genomes are derived from the paired end sequence using 31-mer histograms as described in methods.
nigra and J. microcarpa harbor more SNPs and have higher estimates of expected heterozygosity that do J. regia and J. hindsii. The 31-mer analysis indicates our J. microcarpa genome is more heterozygous than J. nigra. The lower nucleotide diversity estimate for J. microcarpa may be due to the lower re-sequencing coverage obtained compared to J. nigra. As demonstrated below these SNPs are not only resources for genotyping in breeding efforts, they can be critical resources in the identification of impact of natural and artificial selection on and around specific genes associated with traits of interest. It is also important to recognize that these SNP resource capture the majority of the diversity in the species and in particular that in ongoing breeding programs.
The evolutionary history of the polyphenol oxidase (PPO) genes In Martínez-García et al. (2016) the reference genome sequence of 'Chandler' was used to first identify and characterize two distinct polyphenol oxidase (PPO) genes (PPO1 and PPO2) in Juglans regia. The genes were on separate genomic scaffolds, neither linked to a chromosome. Here we use comparative genomics resources to characterize PPO genes in six Juglans species and the outgroup Pterocarya stenoptera. In all seven species, we observe a copy of both PPO1 and PPO2 in close proximity, in the same relative orientation, on the same assembly scaffold (Figure 3). In J. cathayensis and P. stenoptera we found extra copies of PPO2. For these extra copies, we used k-mer depth to determine that the extra copies were due to assembly artifacts and represent alleles of the heterozygous gene (Table S5). In J. microcarpa we found an apparent allelic copy of PPO1 interrupted by an insertion.
To confirm our findings with micro-synteny, we examined the directed pairwise genome alignments for reciprocity and found that the genomic region containing PPO1 and PPO2 was conserved and co-linearly aligned (syntenic) across the two tandem genes (Figure 4). Taken together the results of protein homology, micro-synteny, and k-mer depth, suggest that single functional copies PPO1 and PPO2 genes are in fact tandem in all Juglans species examined, consistent with an ancestral gene duplication. PPO1 and 2 are classical type 3 Cu-binding proteins that catalyze the oxidation of mono and ortho-di-phenolic substrates. As noted in Martínez-García et al. 2016 and confirmed here, they are differentially expressed in different J. regia tissue types with PPO1 being the most abundantly expressed of the two genes in terms of expression in a wider range of tissues (Supplementary table 6). Not surprisingly, PPO1 was the first characterized and shown to be able to convert both monophenolic substrates like tyrosine and diphenolic substrates like DOPA into melanin pigments (Escobar et al. 2008). PPO1 is upregulated by jasmonic acid (Escobar et al. 2008) and in response to walnut blight infection (Khodadadi et al. 2016). The higher expression PPO1 does not appear to be limited to Juglans regia (Supplementary table 7). Structural alignment of primary sequence of PPO1 and PPO2 show correspondence between all relevant structural features. Both proteins have two transit peptides indicating chloroplast localization. Superposition of PPO2 modeled with the recently solved structure of PPO1 reveal conservation of the active site cavity indicating that PPO2 would also display mono and diphenolic activity (Supplementary figure 2). However, the replacement of a few residues in the mouth of the active site results in a change in the electrostatics of the surface that could result an altered range of substrate specificity. Using an outgroup we can infer these consequential changes likely happened on the lineages leading to PPO1 (Supplementary figure 3).
The pairwise genome alignments also identified a 10kbp insertion in the J. regia 1.5 assembly that was not present in the other genomes. We annotated the inserted sequence as a repetitive element, noting that the complete sequence is present as insertions in both the J. regia and J. sigillata genomes at multiple unrelated loci. The three loci in J. regia and one in J. sigillata all appeared to be lineage specific locations. Those in J. regia are apparently homozygous based on examination of the raw Illumina reads. No complete elements were detected in other Juglans genomes. The sequence of the inserted repeat adjacent to PPO1 contains a truncated open reading frame annotated with a DNA-binding domain FAR1 (PF03101). The FAR1 domain functions as a transcription factor (Hudson et al. 2003) in other contexts. While the potential functional impact of the inserted repeat in requires further empirical study, the sequence is a likely cause for the break in the v1.0 assembly between the PPO genes.
An outward taxonomic search for the ancestor of the gene duplication led us to a PPO2 ortholog in the genome of Valley Oak (Quercus lobata) (Sork et al. 2016). The complete Q. lobata genome contained no corresponding ortholog of PPO1. A phylogenetic analysis of these PPO genes was undertaken and a maximum likelihood gene tree was Figure 1 The 31-mer histograms of our paired end sequence data. Each histogram shows a bimodal distribution typical of diploid heterozygous genome. The relative fraction of the distribution under the left (haploid) peak is proportional to the genome heterozygosity. Using the relative proportions of the two peaks the genomes can be ranked by their heterozygosity (Table 1).
n Table 2 Assembly statistics for our six genomes. The original v1.0 J. regia assembly, constructed using similar methods, is included for comparison. As an additional resource and for validation purposes, we also included a v1.5 J. regia assembly which incorporates light coverage of PacBio sequences for improved contiguity

Species
Assembly constructed from which divergence values were estimated. The gene tree shows that subsequent to the duplication event, the PPO1 subgroup diverged more rapidly from the common ancestor than did the PPO2 subgroup (Supplementary figure 1). Tajima's relative rate test (Tajima 1993) comparing JrPPO1 to JrPPO2 using QlPPO as an outgroup yielded a significant excess of differences along the JrPPO1 lineage (P , 0.0003). The most pronounced difference in rates occurs on the lineage immediately following to the common ancestor of section n Table 4 Pairwise genome alignment statistics (top) The percent coverage is calculated for each ordered pair as the percentage of the reference genome covered by the aligned query genome. (bottom) Divergence is calculated for each ordered pair of aligned query to reference genomes. For both metrics, the highest values belonged to pairs of genomes within the same Juglans section n Table 3 CEGMA Core gene results for the genome assemblies of all six Juglans species and the outgroup P. stenoptera. CEGMA: Complete and Partial record the number and fraction of all 248 ultra-conserved CEGs present in the assembly as a complete or partial annotation respectively. Partial annotations use a more liberal cutoff that includes all complete annotations. BUSCO: The number and percentage of 1440 single copy Embryophyte genes present in the assembly. These results are further broken down into single-copy and duplicated genes. Ã Summary results for v1.0 and v1.5 assemblies were the same for both analyses Juglans (1) and (2), when nine times as many changes (46 vs. 5; P , 1e-6) occurred to the PPO1 gene compared to PPO2. A recent acceleration in the nucleotide divergence rate in Juglans PPO1 is inferred when these rates of nucleotide change are compared to the lineages derived immediately from the PPO common ancestor. Immediately after duplication, only 1.27 times as many changes (P , 0.003) happened on the PPO1 lineage compared to PPO2. High ratios of nonsynonymous to synonymous substitution were observed in PPO1 on the lineages leading to J. regia and to J. sigillata (Section Juglans), especially the lineage to their common ancestor where K a = 0.03 and K s = 0.01. High ratios of nonsynonymous to synonymous substitutions (K a /K s ) indicate selection. The relatively low level of polymorphism in J. regia at the PPO1 locus relative to divergence compared and to the same quantities measured in and around PPO2 are consistent with the 'hitchhiking effect' (Maynard Smith and Haigh 1974) of recurrent directional selection and suggest that the impact of domestication continues to the present. In the resequencing data from 27 J. regia samples (Table 5; Table S2) we observed reduced polymorphism at the PPO1 locus: 8 segregating sites compared to 20 at PPO2. Using an HKA-like test for selection (Hudson et al. 1987), the reduction in polymorphism was determined to be significant (P , 3x10 26 ) when polymorphism is compared to the estimated divergence in section Juglans: 53 nucleotide changes for PPO1 and 13 nucleotide changes for PPO2.
Addressing the question of when the ancestral PPO duplication occurred informs which additional taxa may contain descendants of the duplicated genes. Subject to the caveats associated with chronogram analyses, lineage lengths on the PPO gene tree ( Figure S1) indicate that the ancestral PPO gene duplication occurred near the basal split of the Juglandaceae and Fagaceae families. Estimates based on fossil records place the ancestor of Juglandaceae at 71-96 MYA (Xiang et al. 2014), which is an upper estimate for the age of the gene duplication. A lower estimate for the age of the gene duplication would be the divergence of Pterocarya and Juglans during the late Paleocene/Lower Eocene approximately 54MYA (Manchester 1987). Consistent with these findings, using the same methodology, single copy orthologs of JrPPO1 and JrPPO2 were annotated in an unpublished assembly of Pecan (Carya illinoinensis var. Pawnee)) (Jenkins et al. 2015) obtained from the Hud-sonAlpha Institute (hudsonalpha.org). This additional observation shows that the duplication also predated the split of Carya and Juglans during the Paleocene 60MYA (Manchester 1987). n Table 5 The count of single nucleotide polymorphisms and a corresponding estimate of nucleotide diversity from re-sequenced population samples from four Juglans species. The individual accessions are described in Supplementary

Quality and Completeness of the Draft Genomes
For the six new assemblies, gene space completeness, as estimated by both BUSCO and CEGMA, was comparable to the original J. regia V1.0 assembly. This result is consistent with the similarity in methodology used to obtain the assemblies. Heterozygous diploid genomes are a challenge for genome assemblers. The observation that the inflated assemblies were positively associated with the genomes with higher estimated heterozygosity fits a hypothesis that heterozygosity is the underlying cause for assembly inflation. To varying degrees, these genome assemblies consist of regions that are a diploid consensus and regions that are haploid alleles. The unintended haploid regions are a consequence of divergent haplotypes that were neither collapsed by the assembler nor filtered downstream. This haploid allele inflation is most notable in the assemblies of P. stenoptera and J. microcarpa. These two species had the highest level of heterozygosity in the unassembled reads (Table 1) and their assemblies had the highest levels of inflation over their estimated genome size (Table 2). These two genomes also had the highest levels of duplicated, single copy BUSCO orthologs (Table 3). We presume that additional allelic copies of the single copy genes are present in these genomes.

New Juglans genomes are useful for geneoriented analysis
The potential for gene-oriented analysis of the genomic variation in these Juglans reference sequences, their pair-wise alignments, and in the polymorphism data, is evident in our investigation of the PPO genes: 1. The number, spacing and orientation of PPO1 and PPO2 is conserved in the Juglans genomes and Pterocarya stenoptera. 2. In a more distant lineage, Quercus, a single PPO gene was found in the genome. 3. The more rapid divergence, compared to PPO2, and the lack of an ortholog in Quercus is consistent with a "derived" PPO1. 4. As has been reported for other seed crops, excess K a /K s in the divergence of PPO1 on the sect. Dioscaryon lineage supports the view that domestication often involves selection on traits determined by PPO activity, e.g., biosynthesis of phenols, color, pest resistance, etc.

Accessibility of the new genomes
To analyze the PPO genes in Juglans we applied a variety of bioinformatics tools to the new Juglans genomics resources. But the accessibility and utility of the different software varies greatly. Many software tools cannot be usefully applied on a genomic scale, nor can their results be readily inspected. To demonstrate how the genomic resources described in this paper can be browsed quickly to gain the basic comparative, gene-oriented analyses we developed a simple method to load these data into the widely used and well-supported JBrowse software (Skinner et al. 2009)  FAR1-like repeat, an approximately 10kbp insertion in J. regia that is not present in the other genomes. At the bottom of the figure SNPs from the J. regia vcf file are displayed in gray as a density histogram. We imagined a scenario in which a researcher wished to identify, e.g., using a BLAST search, a scaffold in one or more Juglans species that may have homology to a sequence of a known gene that could be from any species. First an identified scaffold can be loaded into JBrowse. Then all alignments of the other Juglans species (including P. stenoptera) can be loaded. Other resources in a large number of formats (e.g., gff, bam, vcf, bed) can be loaded in the coordinate system of the target scaffold, including genotyping data from arrays or GBS. The entire resource of reference sequences, pair-wise alignments and SNP gff files can easily fit on a personal computer, allowing the exploration structures and divergence of all the Juglans genomes and their within species polymorphism. A user can also add annotation as discoveries are made. High quality gene annotation, when it becomes available will further enhance these resources for Juglans researchers and breeders.

Alignments & SNPs
The genome alignments included here are a resource for divergence annotation and for identifying micro-synteny. We used the divergence estimates to construct genome wide phylogenies from the data. We used the micro-synteny annotation to validate a tandem duplication hypothesis for the history of the PPO gene family. While we did not construct a multiple alignment, for many analyses, pairwise alignments will be sufficient to polarize analyses of polymorphism or obtain site specific divergence values. Once these genomes are fully annotated, determining which genes harbor variants likely to have phenotypic effect and exhibit patterns consistent with strong selection will be of great interest, particularly for those lineages undergoing domestication.
For J. regia, J. nigra, J. microcarpa, and J. hindsii, we computed a set of filtered genome wide SNPs and quantified the variation within each species. We observed high levels of nucleotide diversity for J. microcarpa and J. nigra and low levels in, J. regia and J. hindsii consistent with initial estimates from 31-mer histograms. The lowest level of heterozygosity was observed in J. hindsii. This is consistent with the limited natural range and small population of the endemic species. The restricted demographic history for J. hindsii was first proposed by McGranahan et al. (1988). The low heterozygosity observed here confirms previous estimates based on RFLP markers (Fjellstrom and Parfitt 1994) and SSRs (Ross-Davis and Woeste 1998).
The SNPs included here are rich resource of potential markers of interest to breeding programs. In the design of genotyping platforms to greatly augment scion and rootstock breeding, SNPs can be selected based on quality of the SNP call, sequencing depth, allele frequency, sample size, depth, and physical linkage. The relatively complete, unphased genotypes creates a foundation for developing, implementation and validation of GBS tools. and PPO2 region of scaffold896 in J. regia v1.5. The gene regions for PPO1 and PPO2 are aligned to the same scaffold in assemblies as divergent as the outgroup P. stenoptera. An apparent excess divergence in J. regia coincides with a lineage specific insertion of a 10kbp FAR1 domain containing repeat. At this scale only SNP density is visible. Zooming in would reveal the 8 sites overlapping PPO1 and the 20 sites overlapping PPO2.