Introduction

Walnuts (Juglans) are a genus of perennial trees and shrubs consisting of 21 species (Flora of North America) distributed in North America, South America, Eurasia, Central America, and the Caribbean1. Several members of the genus are important sources of edible nuts and wood, including Juglans regia (Persian walnut) which is grown as a crop in every country of the world with a temperate climate, and Juglans nigra (black walnut), the most valuable North American hardwood. Other North American Juglans species include Juglans major (Arizona black walnut), which grows in hot, arid areas near the border between the United States (U.S.) and Mexico, and Juglans cinerea (butternut), which along with black walnut grows in the eastern deciduous forests of the U.S. All New World Juglans belong to section Rhysocaryon. Aside from J. regia (sect. Dioscaryon), Asian Juglans belong to section Cardiocaryon. Species in this section include Juglans mandshurica, which is native to Korea and China, and Juglans ailantifolia, native to Japan. Species hybrids within Juglans are common, often fertile and vegetatively vigorous, and used as rootstocks for nut production or for timber.

Plant genomes contain a large proportion of non-coding repetitive DNA, including transposable elements, retroelements, non-LTR retroelements, tandem repeats, long and short interspersed nuclear elements, and micro- and mini-satellites2,3. Microsynteny among congeners for these types of repeated sequences is evidence for their conservation over evolutionary time and their role in genome evolution2,3. The characterization of synteny and microsynteny among congeners or other phylogenetically related groups for genes, or groups of genes, is an important objective of comparative genomics4. Flanking regions of microsatellites (SSRs) exemplify a type of synteny for non-genic and non-repetitive sequences that is important because when flanking regions are shared across species, PCR primers can be designed that amplify (presumably) homologous regions. Cross-species amplification of SSRs is used often in studies of species for which no SSRs have been published but SSRs from relatives are available5,6. The extent of genome-scale synteny and sequence conservation within genera at loci containing SSRs is not well understood7, although the sharing of LTR-retrotransposons and other repeat elements may be an important factor driving genome size variation across species and genera8. SSR markers are readily identified from both transcriptome and whole genome data9. Nuclear SSRs (nuSSRs) exhibit codominance, high multiplex potential, and many have high levels of polymorphisms. These characteristics make nuSSRs preferable for high throughput mapping, population genetic analysis, and marker-aided plant improvement techniques such as marker-assisted selection in tree breeding programs10, genetic diversity and gene flow analysis11,12, and quantitative trait loci (QTL) analysis13 in plant species.

The primary role of SSRs in plant evolution is unclear, but studies have shown that the distribution of motif frequencies and microsatellite density across the genome is not equal among species14 and seemingly non-random15. The dominant occurrence of motif patterns, motif repeats, and specific sequences and lengths in plant genomes likely result from selection pressures applied on that specific motif during evolution16. Aside from their use in population genetics and breeding (described above), SSRs can also be used for studies of synteny among species within one family and for comparisons of genome organization and evolutionary relationships across species. Except for Juglans nigra and Juglans regia17,18,19, few efforts have been made to generate SSR primer sets for other valuable Juglans species20.

Our study utilized paired-end sequencing (Illumina Inc., San Diego, CA) to identify and evaluate SSRs in 12 Juglans genomes. We sequenced and assembled genomes of several Juglans species to identify nuclear (nuSSR), mitochondrial (mtSSR), and chloroplast (cpSSR) microsatellites and compare their frequency and distribution within and across various Juglans genomes. We also aimed to evaluate SSR-flanking region sequence similarity across walnut species. Finally, we sought to design unique SSR primers for amplification in a single species or homologous loci in multiple Juglans species.

Results and Discussion

Library preparation and paired end sequencing data, frequency of motif types

We used paired-end libraries to generate 977,379,442 reads (940 GB) of raw sequence data from all Juglans genotypes in our study. Quality control filtering yielded 787,083,476 reads (723 GB). Thus, 77% of data were used for downstream analyses (Table 1). The genomes of 12 walnut genotypes were sequenced and assembled using the draft nuclear genome, organellar genome, and transcriptome of J. regia21 as a reference to guide assemblies. Nuclear and organellar genomes of all 12 Juglans genotypes and the J. regia transcriptome were analyzed to identify simple sequence repeat (SSR) motifs. Motifs which repeated from two to four times and over a minimum length of 10 bp were selected for analysis (Supplementary Table S1). In Juglans, di-nucleotide and tri-nucleotide SSRs comprise 98% of total SSRs (Fig. 1). The AT/TA motif was the most frequent motif in the J. regia reference genome and across all of the 12 newly sequenced Juglans genomes (Fig. 2), but the frequency of this motif varied considerably; in J. mandshurica, J. regia and J. major it was 21%, 18%, and 4% respectively (Figs 1, 2, Supplemental Table S1). Our estimates of motif frequency were influenced by sequence quality and depth of coverage; for example, the frequency of a common motif was not consistent among different samples of J. regia or within sections of the genus. In general, the GC/CG motif was among the least represented of the dinucleotide repeats within Juglans genomes. The percentage of GC/CG motifs was relatively high in J. nigra (27%) from section Rhysocaryon, but much lower in another Rhysocaryon, Juglans major (10%). The frequency of GC/CG was surprisingly consistent in three Cardiocaryon species: J. mandshurica (17%), J. cinerea (17%), J. ailantifolia (17%), and among the samples of J. regia in section Dioscaryon (10%). Recent data indicated this specific repeat motif is rare in most hardwood trees, Capsicum species, Arabidopsis, rice, and wheat17,22. In Capsicum, the AG/CT repeat was the most abundant SSR22; this motif was also common in the J. regia reference, but not in the other Juglans genotypes (Fig. 2A), possibly because of low depth of coverage. The AAT/ATA motif was the most prominent tri-nucleotide motif found in Juglans. In monocots such as rice (Oryza sativa) and wheat (Triticum spp.), CCG/CGG comprised the greatest proportion of tri-nucleotide repeats22.

Table 1 Frequency of motifs and nuSSRs in all evaluated genomes.
Full size table
Figure 1
figure 1

Frequency of SSR motifs in all examined genomes. (DNR, di-nucleotide; TNR, tri-nucleotide; TTR, tetra-nucleotide).

Full size image
Figure 2
figure 2

Motif type and frequency. (A) Abundance of di-nucleotide and (B) tri-nucleotide motifs. (C) Total numbers of nuSSRs within the nuclear genome of Juglans spp. and the Persian walnut (J. regia) transcriptome.

Full size image

Simple sequence repeat (SSR) motif frequency (among di-nucleotide, tri-nucleotide, and tetra-nucleotide motifs) decreased sharply with increased motif length in both the reference and the newly sequenced genomes (Fig. 1). For example, there were twice as many AT as AAT, and four times as many AG repeats as AAG (Supplementary Table S1). Di-nucleotide motifs account for 88.7% of nuSSRs while tri- and tetra-nucleotide motifs were significantly less abundant at 9.2% and 2%, respectively (Fig. 1). These data were similar to those from Arabidopsis, cucumber (Cucumis sativus), potato (Solanum tuberosum), tomato (Solanum lycopersicum)22, sorghum (Sorghum bicolor)16, and several hardwood tree genomes17. Simple sequence repeats (SSRs) with Tri-nucleotide motifs were the most common in monocot genomes23. Evaluation of nuclear genome motif frequency variation in other species showed that Arabidopsis (Arabidopsis thaliana), cucumber (Cucumis sativa)22, pine (Pinus taeda L.)24 and ten other hardwood tree species17 also showed a decline in motif numbers when motif lengths increased.

nuSSR primers in Juglans genomes

A total of 824,170 loci containing SSR were identified in this study. We designed 205,000 SSR primers from the nuclear genome and 6,000 from the transcriptome (Table 1). Total SSR motif numbers from the reference genome and transcriptome were considerably higher than from other sequenced Juglans species (Fig. 2). For example, the number of SSRs in the J. regia reference genome was five times greater than that of J. mandshurica and seven times greater than J. ailantifolia (Table 1). The fewest SSRs were found in J. × intermedia (233) and J. × cinerea (2,380). Relatively few SSR loci were found in J. major and J. nigra (6,727 and 9,027) (Table 1). J. nigra had three times more reads than J. major, but only about 1.5 X more SSRs. J. quadrangulata had more reads than J. ailantifolia but J. ailantifolia had 5x more SSRs. We identified a total of 189,208 di-nucleotide, 13,826 tri-nucleotide, and 1,990 tetra-nucleotide SSR loci among all 12 Juglans genomes examined (Supplementary Table S2a,b). The compiled Juglans SSR primer database resulting from this research will provide a rich resource for Juglans markers, enable development of in-depth linkage maps, and allow the fine-mapping of QTLs.

Sequence similarities and synteny among SSR-flanking regions

We compared SSR flanking regions for sequence similarity and the presence of SNPs. BLASTN results indicated the pairwise mean sequence similarity of SSR flanking regions ranged from 91% to 96% across all Juglans genomes (Fig. 3). The species exhibiting the lowest average similarity to all other samples were J. major and J. mandshurica (91%, e-value = 1.006E-10) and the greatest pairwise similarity was between a J. regia backcross and J. × intermedia (96%). Similarity between pairs of J. regia samples was consistently about 95%. Similarity between all J. regia samples and related hybrids varied from 94% to 95%.

Figure 3
figure 3

Predicted flanking regions of SSRs from all the walnut species were compared for pairwise mean sequence similarity and mean e-value using BLASTn.

Full size image

We compared the similarity of each sample’s SSR-flanking region sequences based on BLASTN results. Similarity scores were used for neighbor joining analysis, and the resulting dendrograms showed Juglans species sorted into three main groups consistent with their conventional assignment to sections within the genus (Fig. 4A). A Cardiocaryon clade contained J. ailantifolia and J. mandshurica, J. cinerea, and J. × cinerea. Juglans regia genotypes and hybrids sorted into a Dioscaryon clade, while J. nigra and J. major sorted into a Rhysocaryon clade. Evolutionary analysis based on SSR-flanking region alignment (Fig. 4B) also sorted Juglans genotypes into their respective section with the exception that the closely-related J. ailantifolia and J. mandshurica could not be distinguished from one another25. The placement of J. × cinerea, and J. × intermedia based on a Maximum Composite Likelihood model (MCL), which estimates evolutionary divergence, was not the same as their placement based on BLASTN results (Fig. 4A,B). The sample identified as J. × cinerea was a hybrid of unknown but probably complex pedigree that included butternut and Japanese walnut. In our analysis, it sorted onto a branch between J. ailantifolia and species in other sections of the genus (Fig. 4B). Based on chloroplast data (Fig. 4C), J. × cinerea sorted closely with J. ailantifolia, probably because the hybrid contains a J. ailantifolia chloroplast. Our results are in the agreement with phylogeny results obtained using ITS (Internal Transcribed Spacer) and matK (megakaryocyte-associated tyrosine kinase)26. Our results showed that SSR flanking region sequences are a reliable method for phylogenetic evaluations of Juglans species and may represent an improvement over the marker systems used in previous studies because a large number of SSR flanking regions are available within genomes, are most often found in non-coding DNA (and so are likely to be neutral with respect to selection), and are scattered across the entire genome. About 16,000 flanking regions contributed to our phylogeny. Qi et al.27 reported that the majority of the genome (91%) is covered by neither tandem repeats nor indels, but much of the remaining genome (about 8%) is associated with SSR flanking regions. If these flanking regions are distributed across a random 8% of the genome, they may be useful targets for future analyses utilizing fosmid libraries or segmented genomes.

Figure 4
figure 4

Phylogeny: (A) Phylogeny analyses based on mean similarity of all flanking region of nuSSRs obtained from BLASTN results. (B) Estimate of evolutionary divergence based on SSR-flanking region sequences. (C) Neighbor joining tree based on frequency of SSR motifs within the chloroplast.

Full size image

Although our NGS data underwent adapter trimming before analysis, we cannot know how many SNPs in the flanking regions were true polymorphisms and how much of the sequence homology was homoplasious. Utilizing SNP calling with full coverage (100x) yields a 3% chance of having a read with SNP-like changes due to sequencing or assembly error27. In our data, the percentage of flanking regions that were identical to J. regia was low and varied among species 0.3 to 1.1%; thus, we were unable to design primers that were completely conserved across all species studied (Figs 5 and 6).

Figure 5
figure 5

Genome-based SSR alignment. Typical alignment of SSR-flanking regions within J. regia and related hybrids).

Full size image
Figure 6
figure 6

Percent flanking sequence identity. Percent of identical flanking-SSR sequences across Juglans genomes. Overall is the comparison of all genotypes versus all other genotypes.

Full size image

Screening of SSR loci, development of unique shared nuSSR markers and genome duplication

To identify primer pairs that amplify only a single locus, primer sequences which occurred at multiple locations in the genome were filtered out as not useful for most applications28. We used a criterion of 100 percent identity in both primers for identifying duplicated loci, although it is likely that primers with less than 100% identity would cross-amplify non-target sites. Of the ~205,000 SSR-containing loci, we were able to design unique primers with >40% GC content within the flanking regions for ~60,000 loci. About 39,000 of these primer-containing loci were validated by e-PCR (Supplementary Table S4). The majority of validated single-locus primers identified in each species did not match a locus in any other Juglans species.

Of the 39,000 primer pairs, only 710 (1.8%) were 96–100% identical to a locus found in all Juglans (corresponding to a single nucleotide difference within a 20 bp primer), and no single primer sequence had a 100% match across all species studied (Fig. 6). In some cases, apparently identical loci in different species were similar enough in their flanking regions that we could design an upper primer with 100% identity and a lower primer with high homology (>95%). The use of wobble codes and adjustment of melting temperatures might make these primers practical as universal SSR primers for Juglans, or at least for all the species we evaluated. The greatest number of shared primer sets belonged to J. ailantifolia and J. mandshurica, with 37 shared primer pairs displaying 100% identity at a locus that was present only once in each species. Primers were designed for 63 regions conserved with 100% identity across at least two Juglans species (Supplementary Table S5). In all cases, complete primer sequence conservation was observed between only two species. Regions of shared homology with J. regia, the only species in the genus with a reference genome, could be useful in understanding genome evolution in Juglans and assist in comparative map development, especially if the SSRs associated with them are polymorphic in J. regia.

Based on pairwise comparisons among J. regia genomes, about half of the SSR regions in Persian walnut are evolving, and the other half remain highly conserved (Fig. 6). The proportion is the same, more or less, when comparisons are made within each section (for example, in J. ailantifolia versus J. mandshurica or J. major vs. J. nigra), but comparisons across sections (J. regia vs. J. mandshurica or J. regia vs. J. major or J. cinerea or J. nigra) showed that about 75% of the flanking regions have low similarity (30–60% identity). Across sections, very few loci (less than 1%) show high identity. This demonstrates why it is difficult to identify primers that will amplify well across the entire genus. Heterologous amplification has been demonstrated in Juglans, although the cross-species utility of primers varies by locus and by species6. The success of heterologous amplification using SSR primers depends generally upon the evolutionary distance between the original species and the tested species, with decreasing success as genetic distance increases29.

The fossil record places the radiation of the Juglandaceae to the Paleocene30 (approximately 56 to 66 MYA) and whole genome duplication may have occurred at this time, when the haploid number of chromosomes changed from 8 to 16. Within Juglans, it is unknown what percent of the genome is highly conserved across the genus. Factors leading to conservation, divergence or loss of flanking regions of SSRs in Juglans are probably similar to those in other species, but in general they are not well understood. The size of the genome, and number of retrotransposons and other mobile elements (which is not known for Juglans), may have an important role in genome duplication. Nevertheless, it seems clear that at least 710 (of 60,000) SSR loci in Juglans genomes are older than the most recent speciation in their lineage, as they were found across all Juglans. Interestingly, about 708 loci were duplicated in J. regia, i.e., for these 708 SSRs there was 100% homology in the flanking region at two loci in the J. regia genome (Supplementary Table S8). These 708 loci may be so ancient as to predate the genome duplication event for Juglans. It is also possible their sequence identity is homoplasious. In some cases, the number of identical flanking regions was as great as 19, but duplicated or multiple copies of a flanking region was generally rare. The frequency of multiple copies of a flanking sequence in a genome was greater in J. nigra than in J. regia or J. cinerea, and considerably greater in J. mandshurica and J. ailantifolia than in other species we analyzed (Supplementary Table S8). A low percent of the genome was a shared flanking region with 100-percentage identity across 12 Juglans genomes; perfect sequence identity was found between two species only. Shared loci between two species most likely originated with the common ancestor of the two species in which they were found, and unique SSRs must have arisen later in evolutionary time. The relationship between SSR duplication within a genome and conservation of SSRs across genomes within a genus is not understood but could ultimately shed light on genome evolution. We suggest that loci with sufficient identity to share primers across two Juglans species probably arose prior to speciation, about 1 to 3 million years ago.

Motif frequency in cpSSRs and mtSSRs

The extent of within species variation in chloroplasts and mitochondria in Juglans is poorly described, although there is some data based on reasonable sample sizes and sample distribution31. The reference genome and twelve additional walnut genotypes sequenced in this study revealed 230 chloroplast and 330 mitochondrial SSRs in the walnut genomes. Total density of SSRs in the chloroplast genome was significantly lower than mitochondrial genomes (p = 0.001). There are no cpSSRs in Juglans chloroplasts known to be polymorphic within a species, emphasizing the low level of genetic variation found within Juglans chloroplasts. All chloroplast polymorphisms identified within Juglans species so far are indels or SNPs, many of which result in restriction site differences32,33. Many of these polymorphisms are in SSR-rich regions, but the documented polymorphism is not within the SSR. Comparison of all Juglans genomes showed both cpSSRs and mtSSRs exhibited very slow rates of evolution, which means that in Juglans, cpSSRs are good tools for high-level taxonomic evaluations34,35,36,37,38.

Di-, tri-, and tetra-nucleotide motifs were detected in both chloroplast and mitochondrial genomes, although their presence and frequency varied considerably among species. Motifs longer than three nucleotides in organellar genomes were generally rare and were termed “complex”. There were between three and six loci per chloroplast with complex nucleotide repeats, depending on walnut species (Supplemental Table S1). Complex motifs present in chloroplasts were unique for each species (Fig. 4C, Supplemental Table S1). The total numbers of motifs identified in chloroplast and mitochondrial genomes were small compared with nuSSRs but differences in the numbers of motifs among species and differences in the presence or absence of motifs among species may be great enough that they could be used for understanding plastid evolution (Supplementary Tables S1, S6, S7). The cpSSRs studied were rich in AT motifs. Among dinucleotide SSRs, AT/TA repeats were the most common (46%). Trinucleotide SSRs (ATT/ATA) were also present, but they were rare (1 to 3 motifs per genome). Trinucleotide variations appeared to be species dependent, with 1 to 2 motifs identified per species. J. regia and J. mandshurica displayed two ATT repeats and one ATA repeat, but J. nigra, J. major, and J. cinerea had only the two ATT repeats. J. ailantifolia and J. × cinerea (which likely contains a J. ailantifolia chloroplast) had only one ATT repeat.

All J. regia and J. regia hybrids (J. × intermedia and J. × quadrangulata) had similar repeat motifs in their chloroplast, and for this reason they formed a distinct clade (Fig. 4C). For example, the ATAAA/TTTATA motif was only found in J. regia and hybrids of J. regia (J. × intermedia and in J. × quadrangulata), as expected, assuming the female parent of the hybrid was J. regia. The GATAA motif was found in J. ailantifolia, J. mandshurica, and J. × cinerea only, but the AATA motif was found in these species and J. cinerea (Fig. 4C, Supplemental Table S1). J. × cinerea was not joined with J. cinerea, which means the J. × cinerea sample contained a J. ailantifolia chloroplast, which is commonly observed (Fig. 4C). Chloroplast motifs were similar in J. major, J. nigra, and the J. regia backcross, with the exception that the J. major chloroplast did not contain a TAAA motif. Differences in read depth may have resulted in the apparent absence of motifs. The J. regia backcross was joined with J. nigra because J. nigra was used as a female in the first cross (Supplemental Table S1). Yi-heng et al.39 reported ATAAA motifs in J. regia and J. sigillata and AAGAT repeat motifs in J. cathayensis, J. hopeiensis and J. mandshurica. Whether these motifs are found in all lineages of these species is not yet established.

Chloroplast and mitochondrial SSRs were monomorphic, in sharp contrast with the nuclear genomes (Supplemental Table S1). In all 12 mitochondrial genomes studied, tri-nucleotide repeats were the most prevalent (42%), followed by tetra-nucleotide (35%) and di-nucleotide (23%) motifs. Complex motifs were absent from most species’ mitochondria, but complex motifs were present in the mitochondria of a few species (e.g., AGCA and TATC) (Supplemental Table S1). Most mitochondrial genome SSR motifs were conserved across all studied genomes; however GAA repeats were not found in J. major, J. cinerea or J. nigra. The remaining walnut species contained GAA motifs in addition to TCTT. The TAAA motif was found in J. ailantifolia, J. cinerea, J. × cinerea and J. mandshurica only. Although no complex motif was common to all the mitochondrial genomes we studied, there were several mtSSR motifs that were absent from some species and present in all the others.

Data Archiving

All primer sequences were included as supplementary files.

Conclusion

We developed databases containing lists of nuclear, chloroplast and mitochondrial loci containing SSRs from a total of 12 genotypes of six Juglans species. The depth and density of our marker database will assist researchers seeking to fill gaps in linkage-based genetic maps and improve the resolution of plant breeding approaches. Pairwise similarity of SSR flanking regions reflected known phylogeny. The further development of flanking sequences as sequence tagged sites would increase their utility. In particular, the maternally inherited cpSSRs showed high levels of sequence conservation and are useful for high-level taxonomic evaluations, including analyses of geographical origin, identification of distinct genetic lineages, and studies of dispersal. Mitochondrial motif patterns exhibited a lack of sequence diversity but showed variability in the number of repeats across species.

Material and Methods

A flow chart of the methods process is available in the Supplementary files (Supplementary Fig. S1).

Sample collection and DNA extraction

Tissue samples of 12 Juglans genotypes representing six species and four species hybrids were provided by the Hardwood Tree Improvement and Regeneration Center at Purdue University (HTIRC; www.htirc.org) (Table 1). Genomic DNA was extracted using a CTAB-based extraction method40.

Library preparation, genomic sequencing, and sequence assembly

We constructed DNA sequencing libraries for each of the 12 walnut taxa and sequenced them using a single lane of paired-end reads at the Purdue Genomics Core (https://www.purdue.edu/hla/sites/genomics/) using the Illumina HighSeq2500 (Illumina Inc., San Diego, CA). Raw sequence reads were trimmed using Trimmomatic41. Contiguous sequence fragments were assembled de novo for each species with SOAP-denovo42 using trimmed and filtered reads. Whole genome, transcriptome, and organellar sequences from the Persian walnut genome were downloaded from the Walnut Genome Database (http://dendrome.ucdavis.edu/ftp/Genome_Data/genome/Reju/)21. Chloroplast genomes were constructed by assembling short reads to the Persian walnut chloroplast reference sequence21 using BWA43. Duplicates were flagged and sorted using the Picard tools software44 and SNPs were called using the Haplotype Caller tool from the Genome Analysis Toolkit (GATK)45. A similar approach was used to assemble the mitochondrial genomes.

SSR primer pipeline

We used a modified Perl script from Staton et al.17 to identify microsatellites in our Juglans genomes. Only SSR loci containing perfect repeat units of 2–4 nucleotides were utilized for analysis. We set a minimum-length criterion for SSR analysis; 8–40 repeats for di-nucleotide SSRs, 7–30 repeats for tri-nucleotide SSRs, and 6–20 repeats for tetra-nucleotide SSRs. Our nuSSR primers met the following parameters: a product length of 100–200 bp, a primer size from 18 to 25 bp, annealing temperatures between 55–60 °C, and 40–60% GC content. Simple sequence repeats (SSRs) flanking regions were masked to exclude low complexity regions using Dustmasker46, and primers were designed using Primer3 (v2.3.5)47. Designed SSR primers were denoted ‘nuclear SSRs’ (nuSSRs) to distinguish them from motif sequences. Organellar and transcriptomic genome analyses utilized the bioinformatics pipeline described in Staton et al.17. Organellar SSR primers were designed using Primer 3 in batch mode48. Basic patterns (di-, tri-, and tetra-nucleotide) were identified in all studied genomes. Complex patterns were species-specific, and matched those previously reported for the chloroplast genome39. Unique motifs within the chloroplast genome, and the frequency of di-, tri-, and tetra- nucleotide repeats were counted based on the number of each motif in each species (Supplementary Table S1). Motif frequency values for each genotype were used to perform statistical and pairwise distance analyses of the genotypes using Ward’s method49. Phylogenetic analyses based on motif frequency for chloroplast genome were computed using Ntsys50.

Similarity of SSR flanking regions among Juglans genomes

We performed one-on-one comparisons of all SSR-flanking region sequences identified using BLASTN (BLAST Command Line Applications User Manual 2016). Simple sequence repeat (SSRs) flanking sequences in all species were compared pairwise for sequence similarity and pairwise percent similarity for all comparisons were stored in a database. We then calculated the mean sequence similarity and mean e-value for all BLAST hit results. Mean sequence similarity between SSRs of different species was an indication of how concordant genetic markers and primer sequences of different walnut species were to each other. Pairwise similarity between species for all loci was used for neighbor joining. Mean e-values represent the probability of finding the same sequence in the database by chance, and are directly related to query sequence size. Shorter sequences have higher probabilities of occurring randomly based on mean sequence similarities (Fig. 3). Phylogenetic analysis was performed based on sequence similarity of SSR-flanking regions using Ntsys software50. To find SNP variation (Figs 5 and 6) and the level of heterozygosity across Juglans genomes, SSR flanking regions were aligned with MEGA651. Mean distance within and between groups, genome heterozygosity and SNP variation were calculated with MEGA6.

To estimate the evolutionary divergence between sequences based on SSRs flanking regions, the number of base substitutions per site between sequences was measured. Analyses were conducted using the Maximum Composite Likelihood model (MCL). All positions containing gaps and missing data were eliminated. Evolutionary analyses were conducted in MEGA651, which generated a matrix of mean distances among genomes which was analyzed using neighbor joining to represent the phylogeny. Phylogenetic analysis based on frequency of SSR motifs within the chloroplast was performed based on Neighbor joining methods and a tree was drawn using Ntsys software50.

Filtering of unique SSRs and electronic polymerase chain reaction (e-PCR) validation

Primer sets were designed for all SSRs identified within the 12 Juglans genomes. To avoid untargeted amplification, primer sequences were matched against respective assembled scaffold sequences using BLASTN, and only the primer sets exactly matching a unique genomic region were retained. Primer sequences (forward or reverse) that aligned to multiple regions were filtered out28. The filtered primer pairs were designated ‘unique SSRs’ and were used for further downstream analysis. All remaining primer pair sets were sorted by their repeat number (>10 repeats) and GC content (40–60%). These unique primer sets were then validated by e-PCR52. Although we focused on primers that amplified a single locus, we also evaluated the frequency with which some primer sets could amplify multiple loci with the goal of understanding genome organization.

Filtering shared SSRs among Juglans species

Primer pairs that were shared between two or more species were identified using custom Perl scripts and ClustalW (hybrids were not included in this analysis)53. All previously filtered forward and reverse primer sequences from each species were pairwise matched (>35% sequence identity). Retained primer sets showed 100% sequence identity with primer sets from a different species and displayed the same SSR-flanking region within both species. These SSR regions were considered ‘conserved’ between species.