Assembly and annotation of a chromosome-level reference genome for the endangered Colorado pikeminnow (Ptychocheilus lucius)

Abstract

Advancements in genome sequencing technology have brought unprecedented accessibility of high-throughput sequencing to species of conservation interest. The potential knowledge gained from application of these techniques is maximized by availability of high-quality, annotated reference genomes for endangered species. However, these vital resources are often lacking for endangered minnows of North America (Cypriniformes: Leuciscidae). One such endangered species, Colorado pikeminnow (Ptychocheilus lucius), is the largest North American minnow and the top-level native aquatic predator in the Colorado River Basin of the southwestern United States and northwestern Mexico. Over the past century, Colorado pikeminnow has suffered habitat loss and population declines due to anthropogenic habitat modifications and invasive species introductions. The lack of genetic resources for Colorado pikeminnow has hindered conservation genomic study of this unique organism. This study seeks to remedy this issue by presenting a high-quality reference genome for Colorado pikeminnow developed from Pacific Biosciences HiFi sequencing and Hi-C scaffolding. The final assembly was a 1.1 Gb genome comprised of 305 contigs including 25 chromosome-sized scaffolds. Measures of quality, contiguity, and completeness met or exceeded those observed for Danio rerio (Danionidae) and 2 other Colorado River Basin leuciscids (Meda fulgida and Tiaroga cobitis). Comparative genomic analyses identified enrichment of gene families for growth, development, immune activity, and gene transcription; all of which are important for a large-bodied piscivorous fish living in a dynamic environment. This reference genome will provide a basis for important conservation genomic study of Colorado pikeminnow and help efforts to better understand the evolution of desert fishes.

Introduction

Leuciscidae (Actinopterygii: Cypriniformes) is a highly diverse family of Holarctic fishes which includes North American minnows (Schönhuth et al. 2018). Leuciscids have diversified to occupy nearly all freshwater habitats in North America (Gotelli and Pyron 1991; Martin and Bonett 2015; Burress et al. 2017), including the desert southwest where several species grow to atypical size (>40 cm length) and occupy unique niches in the Colorado River and its tributaries (Minckley et al. 1986). However, despite their abundance and speciosity, leuciscids remain underrepresented in efforts to sequence vertebrate genomes for evolutionary study and conservation genomic applications (Fan et al. 2020; Rhie et al. 2021).

The Colorado pikeminnow (Ptychocheilus lucius), a Colorado River endemic, has several traits which are exceptional among leuciscid species. It is the largest native North American minnow and has a near-exclusively piscivorous diet as an adult, which is rare among minnows (Vanicek and Kramer 1969). It was historically the top-level aquatic predator in a highly depauperate native fish community (Minckley and Marsh 2009). Several of its life history traits (long lifespan, slow maturation process, and migratory behavior associated with spawning) are common among imperiled southwestern fishes (Tyus and McAda 1984; Tyus 1990). Population declines due to changes in flow regimes caused by dams, threats from nonnative species, and habitat loss have led to its listing as endangered under the US Endangered Species Act of 1973 (Miller 1961; Minckley and Deacon 1968).

Contemporary population trends, environmental uncertainty, and anthropogenic habitat modifications portend an uncertain future for Colorado pikeminnow (Comte et al. 2022; Pennock et al. 2022). Its contemporary distribution is restricted to the Green, Upper Colorado, and San Juan River subbasins (Fig. 1; Dibble et al. 2023). The San Juan River population continues to experience low recruitment following its extirpation and subsequent reestablishment (Ryden and Ahlm 1996; Franssen et al. 2016; Clark et al. 2018). The Green and Upper Colorado River populations are self-sustaining, but recent population declines have made persistence uncertain (Osmundson and White 2017; Dibble et al. 2023).

Fig. 1.

Approximate current and historical distributions of Colorado pikeminnow (Ptychocheilus lucius) along major rivers of the Colorado River Basin in the United States and Mexico. The Green, San Juan, and Upper Colorado River subbasins which contain the present-day Colorado pikeminnow populations are highlighted. The top-left inset shows the location of the Colorado River Basin in North America. The Colorado pikeminnow illustration is shown with copyright permission from © Joseph R. Tomelleri.

The precarious nature of remnant Colorado pikeminnow populations identifies a need to develop genomic resources which facilitate the estimation of population genomic parameters with improved precision and accuracy. For example, an annotated reference genome can aid in identification of selectively neutral loci which meet key assumptions for many conservation genomic analyses (Allendorf et al. 2010). Reduced representation sequencing methods routinely used in conservation genomic studies also benefit from reference genomes to ensure data sets fit analytical model assumptions, such as linkage disequilibrium among genetic loci (Campbell et al. 2018; O’Leary et al. 2018). Additionally, genomic resources facilitate studies from evolutionary perspectives which can help identify genomic structural variants associated with environmental adaptation (Wellenreuther et al. 2019; Mérot et al. 2020), facilitate differential gene expression studies to quantify molecular responses to physiological or ecological stimuli (Connon et al. 2018), and identify species’ physiological limitations (Komoroske et al. 2021).

Unfortunately, reference genomes which can maximize the information gained from genome-scale studies remain rare for western North American leuciscids. Colorado pikeminnow is no exception, with prior studies relying upon legacy genetic markers such as allozymes, microsatellites, and mtDNA assays (Morizot et al. 2002; Borley and White 2006; Martin et al. 2015). This deficit is remedied herein by combining long- and short-read sequence data to assemble and annotate a chromosome-scale Colorado pikeminnow reference genome. Comparisons to other Cypriniformes species were conducted to demonstrate genome contiguity and completeness, as well as to investigate potential gene family expansions of ecological importance for Colorado pikeminnow.

Materials and methods

Tissue collection

A male Colorado pikeminnow was collected from the 2006 cohort at the Southwestern Native Aquatic Resources and Recovery Center in Dexter, New Mexico, USA. This cohort is an F2 descendant from wild fish collected from the Colorado and Green rivers in 1981 and 1991 and was spawned to serve as a backup broodstock for the species’ recovery program (Diver et al. 2019). The 2006 cohort has an observed heterozygosity of 0.862 as determined from 24 microsatellite loci (Diver et al. 2019). The selected individual was first anesthetized using MS-222 for 2 blood draws into 4-mL vacutainer tubes coated with K2EDTA as anticoagulant. Blood samples were placed on dry ice immediately after collection and shipped overnight to sequencing facilities. The fish was then euthanized with an overdose of MS-222 and dissected to obtain 1 cm³ subsamples of 11 tissues for transcriptome characterization: brain, eye, gall bladder, gill, heart, kidney, liver, muscle, skin, spleen, and testes. These tissues were preserved in RNAlater and stored at 4°C for 24 h prior to RNA extraction.

DNA extraction and sequencing

One blood sample was shipped to the University of Oregon Genomics and Cell Characterization Core Facility for high molecular weight (HMW) DNA extraction using a Circulomics Nanobind CBB kit (Pacific Biosciences, Menlo Park, CA, USA). A genomic library was then prepared from 10 μg DNA using the PacBio HiFi Prep Kit following the manufacturer’s protocols. HiFi sequencing was conducted on 2 SMRT cells in a Pacific Biosciences Sequel II sequencer. The other blood sample was shipped to North Carolina State Genomic Sciences Laboratory (NCSU GSL) for Hi-C library preparation and sequencing. This library was prepared using the Proximo^™ Hi-C kit for animals (Phase Genomics, Seattle, WA, USA) following the manufacturer’s protocols for nucleated red blood cells and sequenced on a single lane of an Illumina NovaSeq 6000 SP flowcell (2 × 150 bp reads).

Genome assembly

HiFi data were filtered to remove reads containing PacBio adapter sequence with HiFiAdapterFilt v2.0.0 (Sim et al. 2022). Illumina adapter sequence was trimmed from Hi-C reads using TrimGalore v0.6.7 (Martin 2011). Genome size was predicted from the k-mer profile of the HiFi data by counting 32-mers in Meryl v1.3 (Rhie et al. 2020) and analyzing the resulting histogram in GenomeScope v2.0 (Ranallo-Benavidez et al. 2020). Contigs were assembled in Hifiasm v0.16.1-r375 using default settings and Hi-C integration (Cheng et al. 2021, 2022). Scaffolding with Hi-C data was accomplished using the Juicer v1.6 and 3D-DNA v201008 pipelines (Durand, Shamim, et al. 2016; Dudchenko et al. 2017). A Hi-C contact map was produced using the trimmed Hi-C reads, DpnII restriction cut site locations, and the assembled contigs as Juicer inputs. The contact map was provided to the 3D-DNA pipeline to identify draft scaffolds that were subsequently visualized in Juicebox v1.11.08 (Durand, Robinson, et al. 2016). Two misjoins were manually corrected in Juicebox, and the updated contact map was input into 3D-DNA to construct the scaffolded assembly.

The assembly was screened for duplicated scaffolds using the “clean” function of funannotate v1.8.15 (Palmer and Stajich 2023). Scaffolds were then classified using Kraken2 v2.1.3 (Wood et al. 2019) and the Kraken2 “standard plus protozoa, fungi, & plant” (PlusPFP) RefSeq index dated 2023 October 9. Scaffolds mapping to algal, bacterial, plant, and yeast sequences were removed using custom scripts (removeContamination.sh and matchHeader.pl). HiFi reads were mapped back to the final assembly with Minimap2 v2.24-r1122 (Li 2018, 2021) for the purpose of calculating and visualizing genome assembly statistics in Quast v5.2.0 (Gurevich et al. 2013) and BlobToolKit v3.5.4 (Challis et al. 2020). Genome completeness was assessed using BUSCO v5.5.0 (Manni et al. 2021) (actinopterygii_odb10 database: 26 genomes, 3,640 genes).

Repeat modeling and masking

RepeatModeler v2.0.3 was used to develop a custom repeat library for Colorado pikeminnow, with the -LTRStruct option enabled (Flynn et al. 2020). The custom library was screened using BLAST (Altschul et al. 1990) to remove coding sequences that aligned to the UniProt-SwissProt database (Poux et al. 2017), downloaded 2023 November 30. The custom library was combined with the Repbase v27.08 vertebrate database (Bao et al. 2015) and input into RepeatMasker v4.1.2 to generate a soft-masked genome and GFF file (Smit et al. 2015).

RNA extraction and sequencing

RNA was extracted using the Macherey-Nagel NucleoSpin RNA Mini Kit (Macherey-Nagel, Düren, Germany). The manufacturer’s protocol was modified to include mechanical tissue disruption. Up to 20 mg of tissue was placed in a vial containing 2.8 mm ceramic beads and submerged in the Buffer RA1/β-mercaptoethanol mixture from the NucleoSpin RNA Mini Kit. Tissues were subjected to 2 rounds of 2-min disruption at 20–30 Hz on a Qiagen TissueLyser II (Qiagen, Germantown, MD, USA). Sample blocks were rotated 180° between rounds to promote homogenization. Samples then entered the NucleoSpin RNA Mini Kit manufacturer’s protocol for RNA extraction at the lysate filtering step. RNA extracts were quantified with a Qubit RNA Broad Range kit on the Qubit 3.0 Fluorometer (Life Technologies, Carlsbad, CA, USA) and diluted to a standard concentration of 20 ng/µL using nuclease-free water. RNA libraries were prepared from 500 ng input RNA following the poly-A capture protocol of the Illumina Stranded mRNA Prep kit (Illumina, Inc., San Diego, CA, USA) and dual indexed with 10 bp barcodes. Sequencing was performed at the NCSU GSL on a single lane of an Illumina NovaSeq 6000 SP flowcell (2 × 150 bp reads).

Transcriptome assembly

Raw RNA sequence reads obtained from all tissues were first subjected to error correction using the k-mer-based program Rcorrector v1.0.5 (Song and Florea 2015). Read pairs containing at least 1 read deemed unfixable by Rcorrector were discarded. TrimGalore v0.6.7 (Martin 2011) was then applied with default settings to remove low-quality bases and Illumina adapters. Processed reads were then aligned to the SILVA v138.1 LSUParc and SSUParc databases (Quast et al. 2013) using bowtie2 v2.4.5 (Langmead and Salzberg 2012) to remove ribosomal RNA reads. The –very-sensitive-local and –nofw alignment options were applied in bowtie2. Finally, poly-A/T tails of at least 5 bp were trimmed from the 3′ and 5′ ends of all reads in Prinseq v0.20.4 (Schmieder and Edwards 2011). Read pairs were discarded if 1 read was <35 bp after trimming. Processed RNA reads were reference aligned to the genome assembly using the –very-sensitive, –dta, and –rna-strandedness RF options in HISAT2 v2.2.1 (Kim et al. 2019) and sorted using samtools v1.15.1 (Danecek et al. 2021).

Annotation

Extrinsic protein evidence, RNA sequence alignments, and the soft-masked genome were used to predict genes in the Colorado pikeminnow genome. Protein evidence was utilized from fishes of superorder Ostariophysi that had annotated chromosome-scale genome assemblies available in the NCBI RefSeq database. These included Mexican tetra (Astyanax mexicanus; GCF_023375975.1), goldfish (Carassius auratus; GCF_003368295.1), common carp (Cyprinus carpio; GCF_018340385.1), zebrafish (Danio rerio; GCF_000002035.6), channel catfish (Ictalurus punctatus; GCF_001660625.3), Chinese sucker (Myxocyprinus asiaticus; GCF_019703515.2), fathead minnow (Pimephales promelas; GCF_016745375.1), and razorback sucker (Xyrauchen texanus; GCF_025860055.1). Proteins curated for vertebrate species in the UniProt-SwissProt database were also included.

Gene prediction was conducted in the BRAKER3 v3.0.6 pipeline (Gabriel et al. 2023), which combines the BRAKER1 and BRAKER2 pipelines to utilize both RNA sequence data and extrinsic protein evidence for genome annotation (Hoff et al. 2016, 2019; Brůna et al. 2021). Briefly, BRAKER3 first combined RNA sequence alignments using StringTie (Kovaka et al. 2019). The assembled transcripts of putative protein-coding genes were compared against the protein database to identify probable genes and generate protein hints with GeneMark-ETP (Buchfink et al. 2015; Brůna et al. 2021, 2023). The protein hints were used to train AUGUSTUS (Stanke et al. 2006, 2008). Finally, genes predicted by GeneMark-ETP and AUGUSTUS were combined using TSEBRA (Gabriel et al. 2023) and output in GFF format (Pertea and Pertea 2020).

Putative gene functions were identified first using InterProScan v5.65-97 (Jones et al. 2014). All default analyses were run, with the Phobius v1.01 (Käll et al. 2004), SignalP v4.1 (Petersen et al. 2011), and TMHMM v2.0 (Sonnhammer et al. 1998; Krogh et al. 2001) analyses also enabled. The eggNOG-mapper v2.1.12 online database was queried with default settings to conduct functional annotation of protein sequences using orthology assignments (Cantalapiedra et al. 2021). The “annotate” function in funannotate v1.8.15 was then used to combine the vertebrate BUSCO data set with the outputs of InterProScan (XML format), eggNOG, and BRAKER3. The final annotation was evaluated in BUSCO v5.5.0 (Manni et al. 2021), and summary statistics were calculated using the agat_sp_statistics.pl script from AGAT v1.2.0 (Dainat 2023).

Comparative genomic analysis

The Colorado pikeminnow genome was aligned to 2 other southwestern North American leuciscids, spikedace (Meda fulgida; GCA_030578275.1) and loach minnow (Tiaroga cobitis; GCA_030578255.1), to identify syntenic regions and genomic structural rearrangements (Alexandre et al. 2023). Alignments were conducted using the default (-ax -asm5) options in Minimap2 v2.24. Colorado pikeminnow scaffolds were reverse complemented as necessary using a custom script (revcomList.pl) to facilitate alignment to other genomes. Synteny was evaluated using SyRI v1.6.3 (Goel et al. 2019) and visualized in plotsr v1.1.1 (Goel and Schneeberger 2022).

OrthoFinder v2.5.5 (Emms and Kelly 2019) was executed using default settings to identify orthogroups from the proteomes of 6 species from suborder Cyprinoidei with annotated genomes in the NCBI RefSeq database: common carp, goldfish, zebrafish, fathead minnow, and speckled dace (Rhinichthys osculus; GCF_029890125.1). Large orthogroups and gene family expansions were identified following the methods of Alexandre et al. (2023). First, a custom script (findLargestOrthogroups.pl) was employed to identify the 10 largest orthogroups in Colorado pikeminnow with at least 1 ortholog present in the outgroup genome (zebrafish). Secondly, gene family expansions were discovered by using another custom script (findExpandedOrthogroups.pl) to identify orthogroups with gene counts in Colorado pikeminnow that were at least 5× greater than in zebrafish. The outputs of these 2 analyses are respectively called the “largest orthogroups” and “expanded orthogroups” data sets. Orthologs were extracted from the zebrafish proteome and evaluated in STRING-DB v12.0 (Szklarczyk et al. 2021) using the “search proteins by sequences” function to identify Gene Ontology (GO) terms and evaluate functional enrichment in Colorado pikeminnow. Statistical significance of enrichment was assessed using P-values corrected by a 5% false discovery rate.

Results and discussion

The HMW DNA extraction yielded 36.2 μg DNA with 71.9% of fragments being >30,000 bp in length. PacBio HiFi sequencing yielded 2,893,341 reads totaling 36.3 Gb of sequence data. HiFi sequence data had a read N50 of 13,744 bp and a maximum read length of 76,826 bp. The Hi-C library yielded 218,603,599 reads totaling 66 Gb sequence data, of which 218,308,942 reads (65.4 Gb) were retained after quality filtering. RNA sequence reads for 11 tissue types totaled 340,027,426 reads (102.7 Gb), of which 268,869,255 reads (74.1 Gb) were retained postquality filtering. RNA extraction yield and number of sequence reads per tissue type are provided in Supplementary Table 1.

K-mer analysis predicted a genome size between 1,040,088,571 and 1,041,803,284 bp (Supplementary Fig. 1). This was smaller than the 1,232,280,000 bp length predicted from flow cytometry analysis of red blood cells (Gold and Li 1994). The initial assembly of 337 contigs fell between these expected values at 1,103,106,296 bp in length (Supplementary Table 2 and Supplementary Fig. 2). The final genome length, after Hi-C scaffolding and removal of contaminants, was 1,100,610,714 bp (Fig. 2). The Hi-C contact map is provided in Supplementary Fig. 3. The final product was a highly contiguous genome with 97.5% of its total length being contained in 25 scaffolds, coincident with the haploid number of chromosomes expected for Colorado pikeminnow (Jenkin et al. 1992). Mean sequencing depth of HiFi reads for the final assembly was measured at 35× (Supplementary Fig. 4).

Fig. 2.

A snail plot was prepared in BlobToolKit v3.5.4 to highlight postscaffolding genome assembly statistics, genome-wide GC content, and BUSCO scores.

Genome assembly statistics were compared to the loach minnow, spikedace, and zebrafish reference genomes in Table 1. The Colorado pikeminnow genome compared favorably with these species in measures of N50, N90, L50, and L90. Colorado pikeminnow had the third-greatest genome length of these 4 species and exhibited the second-largest N50 (41.8 Mb), L50 (11 scaffolds), and L90 (23 scaffolds) values. Its N90 value (31.5 Mb) was the largest of the 4 species evaluated. BUSCO values also compared favorably with those for spikedace, loach minnow, and zebrafish (Table 2). Colorado pikeminnow had the greatest numbers of complete BUSCOs (N = 3564; 97.9%) and single-copy BUSCOs (N = 3502; 96.2%). It had the lowest number of fragmented BUSCOs (N = 22; 0.6%) and second-fewest duplications (N = 62; 1.7%) and was tied with loach minnow for the fewest missing BUSCOs (N = 54; 1.5%).

Table 1.

Genome length summary statistics calculated for Colorado pikeminnow in Quast v5.2.0.

Statistics	Colorado pikeminnow	Spikedace	Loach minnow	Zebrafish
# contigs (≥0 bp)	305	84	551	1,923
# contigs (≥1,000 bp)	305	84	551	1,922
# contigs (≥5,000 bp)	276	73	551	1,824
# contigs (≥10,000 bp)	259	71	551	1,727
# contigs (≥25,000 bp)	200	68	499	1,281
# contigs (≥50,000 bp)	139	64	283	1,091
Largest contig	63,658,627	52,739,976	67,399,795	78,093,715
Total length (≥0 bp)	1,100,610,714	882,128,298	1,320,538,350	1,679,203,469
Total length (≥1,000 bp)	1,100,610,714	882,128,298	1,320,538,350	1,679,202,819
Total length (≥5,000 bp)	1,100,538,438	882,108,831	1,320,538,350	1,678,981,660
Total length (≥10,000 bp)	1,100,417,455	882,098,831	1,320,538,350	1,678,178,846
Total length (≥25,000 bp)	1,099,431,699	882,046,241	1,319,423,137	1,671,157,076
Total length (≥50,000 bp)	1,097,192,067	881,919,014	1,311,861,986	1,664,545,481
N50	41,794,964	34,805,954	48,679,909	52,186,027
N90	31,481,000	26,365,680	30,906,194	339,135
auN	43,728,421	35,515,798	45,453,224	44,676,698
L50	11	11	13	14
L90	23	22	25	405
GC (%)	39.50%	39.33%	40.28%	36.60%
# N's per 100 kbp	16.40	7.12	3.02	279.52
# N's	180,500	62,784	39,837	4,693,316

Statistics are also provided for genome assemblies of 3 related species for comparison: spikedace (M. fulgida; NCBI accession = GCA_030578275.1), loach minnow (T. cobitis; NCBI accession = GCA_030578255.1), and zebrafish (D. rerio; NCBI accession = GCF_000002035.6). All statistics were calculated considering only contigs ≥ 3,000 bp in length unless otherwise noted.

Table 2.

Genome completeness for Colorado pikeminnow was assessed in BUSCO v5.5.0 using the Actinopterygii database version odb10.2019-11-20 containing 3,640 single-copy orthologs.

	Colorado pikeminnow		Spikedace		Loach minnow		Zebrafish
BUSCO	Total	Percent (%)	Total	Percent (%)	Total	Percent (%)	Total	Percent (%)
Complete	3,564	97.9	3,533	97.0	3,557	97.7	3,496	96.0
Single copy	3,502	96.2	3,484	95.7	3,439	94.5	2,861	78.6
Duplicated	62	1.7	49	1.3	118	3.2	635	17.4
Fragmented	22	0.6	32	0.9	29	0.8	60	1.6
Missing	54	1.5	75	2.1	54	1.5	84	2.4

BUSCO scores are provided for genome assemblies of 3 related species for comparison: spikedace (M. fulgida; NCBI accession = GCA_030578275.1), loach minnow (T. cobitis; NCBI accession = GCA_030578255.1), and zebrafish (D. rerio; NCBI accession = GCF_000002035.6).

Repeat element identification

Repetitive elements comprised 609,816,837 bp (55.41%) of the Colorado pikeminnow genome (Supplementary Table 3). Retroelements made up 91.7 Mb (8.33%) of the genome, whereas DNA transposons were 185.5 Mb (16.86%), rolling circles were 27.6 Mb (2.51%), and 287.8 Mb (26.15%) were unclassified elements. The 2 largest categories of retroelements were LINEs (16.1 Mb; 1.47%) and LTR elements (72.5 Mb; 6.59%). Within LTR elements, most were categorized as Gypsy or DIRS1 (58.4 Mb; 5.3%). The top categories of DNA transposons were hobo-Activator elements (47.0 Mb; 4.27%) and Tourist/Harbinger elements (17.1 Mb; 1.55%). Satellites (12.1 Mb; 1.1%) and simple repeats (3.1 Mb; 0.28%) accounted for relatively small proportions of the genome.

Annotation

BRAKER3 identified 32,933 mRNAs representing 25,192 genes that covered 435.16 Mb (39.54%) of the genome (Supplementary Table 4). A total of 379,233 exons (mean = 11.5 per mRNA) and 346,300 introns (mean = 10.5 per mRNA) were found. On average, genes were 17,273 bp in length. Mean exon length was 292 bp and mean intron length was 1,921 bp. After filtering the proteome to retain the longest isoform of each gene, BUSCO analysis of the transcriptome indicated high retention of single-copy orthologs and low levels of duplication [complete = 3,539 (97.3%), complete and single-copy = 3,489 (95.9%), complete and duplicated = 50 (1.4%), fragmented = 12 (0.3%), missing = 89 (2.4%); database = actinopterygii_odb10]. BUSCO results for individual tissue types are provided in Supplementary Fig. 5.

Comparative genomic analysis

Syntenic comparisons to loach minnow and spikedace genomes revealed that approximately 50% of the Colorado pikeminnow assembly was comprised of syntenic regions corresponding to each of these reference genomes (loach minnow = 518.46 Mb, 47.11% assembly, Fig. 3, Supplementary Table 5; spikedace = 562.45 Mb, 51.10% assembly, Fig. 4, Supplementary Table 6). Genomic structural rearrangements were less frequently detected. Inversions were the most detected of all rearrangement categories (loach minnow = 99.03 Mb, 9.0%; spikedace = 116.97 Mb, 10.63%), followed by translocations (loach minnow = 26.84 Mb, 2.44%; spikedace = 15.58 Mb, 1.42%) and duplications (loach minnow = 16.21 Mb, 1.47%; spikedace = 18.82 Mb, 1.71%). Approximately one-third of the Colorado pikeminnow genome failed to align with these 2 genomes (loach minnow = 412.54 Mb, 37.48%; spikedace = 359.15 Mb, 32.63%).

Fig. 3.

Synteny of the Colorado pikeminnow genome with loach minnow (T. cobitis; NCBI accession = GCA_030578255.1) was assessed by first aligning the genomes with the default options in Minimap2 v2.24. Synteny was evaluated in SyRI v1.6.3, and plots were produced in plotsr v1.1.1. Chromosomes are labeled on the y-axis according to NCBI GenBank accession numbers for loach minnow chromosomes.

Fig. 4.

Synteny of the Colorado pikeminnow genome with spikedace (M. fulgida; NCBI accession = GCA_030578275.1) was assessed by first aligning the genomes with the default options in Minimap2 v2.24. Synteny was evaluated in SyRI v1.6.3, and plots were produced in plotsr v1.1.1. Chromosomes are labeled on the y-axis according to NCBI GenBank accession numbers for spikedace chromosomes.

The similar proportions of syntenic regions shared by Colorado pikeminnow with loach minnow and spikedace is concordant with observed phylogenetic relationships indicating that Colorado pikeminnow is approximately equally phylogenetically distant from these 2 species (Schönhuth et al. 2018). Colorado pikeminnow likely diverged from spikedace and loach minnow approximately 36–38 million years ago (adjusted divergence times obtained from the TimeTree v5 database; Supplementary Table 7) (Kumar et al. 2017, 2022). Alexandre et al. (2023) compared spikedace and loach minnow to sunbleak (Leucaspius delineatus) and found that 66.2% (spikedace) and 54.4% (loach minnow) of each species’ genome were syntenic with sunbleak. Each of these 4 species belongs to a different subfamily within Leuciscidae (Colorado pikeminnow = Laviniinae, loach minnow = Pogonichthyinae, spikedace = Plagopterinae, and sunbleak = Leuciscinae). Three of these families (Leuciscinae, Plagopterinae, and Pogonichthyinae) share a more recent common ancestor with one another than with Laviniinae (Schönhuth et al. 2018). Although the 47.1–51.1% synteny that Colorado pikeminnow shares with loach minnow and spikedace seems low, it is congruent with values observed among other Leuciscidae and reflects the divergence time among these species (Alexandre et al. 2023). Other studies have found synteny to be variable among Teleost fish families, as evidenced by a mean 75.3% synteny among 3 African cichlid species (Cichlidae), and 84.4% synteny among 3 livebearer species (Poeciliidae) (Pandey et al. 2020).

The OrthoFinder analysis assigned 202,658 genes representing 95.2% of the total input genes to 25,465 orthogroups. Half of all genes were assigned to 9,004 orthogroups each containing ≥8 genes. All 6 species were represented in 16,578 orthogroups, and 2,083 of these (12.56%) were single-copy orthogroups. OrthoFinder recovered a species tree (Fig. 5) that was concordant with expected phylogenetic relationships (Schönhuth et al. 2018). Statistical support, measured as the proportion of gene trees matching the species tree topology, was high for all nodes (support ≥ 0.89) except the split between speckled dace and fathead minnow (support = 0.59).

Fig. 5.

A species tree was calculated in OrthoFinder v2.5.5 from 16,578 orthogroups that contained all 6 species. Nodal support values represent the proportion of gene trees that support each species tree bipartition. The species tree was rooted by D. rerio, which was algorithmically detected by OrthoFinder as the outgroup taxon.

Significantly enriched GO terms were detected when evaluating the “largest orthogroups” data set. These included a GO process related to gastrulation (GO:0001702; P = 0.0023) and 2 GO functions relating to RNA polymerase II transcription regulatory region sequence (GO:0000977; P = 0.0025) and DNA-binding transcription factor activity (GO:0000981; P = 0.0025). Uncharacterized sequences related to insulin/insulin-like growth factor/relaxin family (CL:44178, P = 0.00058) and immunoglobulin (SM00409, P < 0.00001) were detected among the “expanded orthogroups,” but these gene families were not associated with GO terms.

Analysis of the “largest” and “expanded orthogroups” data sets identified ortholog groups with importance for gene expression, stress response, growth, and development (Supplementary Tables 8 and 9). These functions are important for all organisms, and additional studies should be conducted to examine whether the detected gene families represent lineage-specific expansions once additional genomic resources become available for other desert fish taxa (Lespinet et al. 2002; Fortna et al. 2004). However, there are multiple ecological factors that could explain the importance of these families in Colorado pikeminnow, which is regularly subject to environmental stressors such as elevated water temperatures and drought (Woodhouse et al. 2010; Seager et al. 2013). Significant enrichment of immunoglobulin is relevant in this context, since adaptive immune response is hypothesized to provide a greater relative contribution to immunocompetence in fishes at elevated environmental temperatures (Scharsack and Franke 2022). Regulation of transcription (e.g. GO:0000977 and GO:0000981) is important for all organisms, but could be especially vital for fishes living in fluctuating environments since these genes allow for timely response to environmental stressors via gene expression (de Nadal et al. 2011). Finally, genes regulating growth and development (e.g. GO:0001702 and CL:44178) (Chandhini et al. 2021) could be important because Colorado pikeminnow undergoes an ontogenetic diet shift from consuming aquatic invertebrates to exclusive piscivory (Vanicek and Kramer 1969). Growth both facilitates ontogenetic diet shifts for piscivorous fishes and allows consumption of larger prey (Juanes et al. 1994; Persson and Brönmark 2002). The importance of growth to the Colorado pikeminnow life history strategy has also been supported by observations of immature pikeminnow migrating to waters that allow for an extended growth period during winter and spring months (Durst and Franssen 2014).

The characterization of reference genomes for nonmodel organisms benefits multiple research fields, including both evolutionary and conservation genomics. The Colorado pikeminnow reference genome adds to a growing number of reference genomes for Leuciscidae (Martinson et al. 2022; Meuser et al. 2023), multiple of which are species of conservation concern from the Colorado River (Alexandre et al. 2023; Suchocki et al. 2023). Over time, the accrual of these resources will facilitate better understanding of how species evolve and adapt to their environments (Zhang et al. 2017; Bo et al. 2022). They will help provide novel insights into conservation issues that impact endangered species (Supple and Shapiro 2018; Formenti et al. 2022; Theissinger et al. 2023) and offer potential to examine endangered populations with unprecedented resolution as we determine how to make actionable population management decisions from genome-scale data (Kardos et al. 2021; Schiebelhut et al. 2024).

Data availability

Raw RNA and DNA sequence data are available in the NCBI Sequence Read Archive (SRA) under BioProject PRJNA1003539. The assembled genome is available under GenBank accession JAVCAX000000000. The genome annotation is provided on the Open Science Framework at https://osf.io/w9zuy/ (DOI: 10.17605/OSF.IO/W9ZUY). All scripts utilized in genome assembly are provided at https://github.com/stevemussmann/cpm_genome_assembly (DOI:10.5281/zenodo.10626887).

Supplemental material available at G3 online.

Funding

This work was funded through the National Fish and Wildlife Foundation (project # 8006.21.072189).