Chromosome-level genome assembly of Protandrena (Anthemurgus) passiflorae (Hymenoptera: Andrenidae), a host-plant specialist bee

Abstract

The passion flower bee, Protandrena (Anthemurgus) passiflorae (Robertson) is a monolectic, host-plant specialist of the passion flower plant Passiflora lutea L. Using a single adult male individual, we generated long-read PacBio HiFi, HiC, and short-read RNA sequencing data to build a well-annotated, chromosome-level genome assembly for this species. The final nuclear genome is 249 Mb with 150 × coverage and with most of the genome scaffolding into 12 chromosomes. The scaffold N50 is 21.4 Mb and the genome has a Benchmarking Universal Single-Copy Ortholog score of 97.2% for 5991 hymenopteran genes. BRAKER3 annotation of the genome identified 12,098 genes and 15,353 total transcripts and found that 20.27% of the genome is made up of repetitive elements. We resolved a mitochondrial genome of 12.7 kb. The P. passiflorae genome represents one of only a few published andrenid bee genomes and one of the first monolectic bees. This new high-quality genome will serve as a valuable resource for investigating the genomic basis of specialization and for providing a useful resource for studying pollinator health and conservation.

Scientists from the USDA-ARS' Beenome100 project report the genome assembly of the passionflower bee, Protandrena passiflorae, a host-plant specialist of a passionflower plant. This genome is the first Protandrena species to be sequenced and assembled at a chromosome scale, and one of a few available in the bee family, Andrenidae. High-fidelity sequencing of the entire genome using PacBio, along with HiC sequencing, and short-read RNA sequencing data led to a comprehensive chromosome level assembly that is >97% complete. This assembly showcases producing efficient, comprehensive, and continuous genome arrangements from a single specimen.

Introduction

Sequencing genomes for every species on Earth has become a major objective of biodiversity science in the 21st century (e.g. Hotaling et al. 2021; Rhie et al. 2021). Similar to the collection, preservation, and curation of specimens in natural history museums, capturing genomic information of diverse species and specimens helps us conserve and understand biological diversity on the planet. Once acquired, reference genomes can be used for a multitude of basic and applied purposes, including evolutionary and functional genomics, conservation genetics, breeding, and systematics. A recent initiative, called Beenome100, aims to generate reference genomes for diverse wild bees in the United States, focusing on species of agricultural, biological, and conservation importance. The project aims to provide genetic resources that will help protect bee diversity and the critical pollination service that they provide and to facilitate future research. This study presents a new genome for this effort.

The passion flower bee, Protandrena (Anthemurgus) passiflorae (Robertson 1902), is a monolectic bee described as a small, black, robust andreniform bee that is ∼7.5–8.5 mm in length (Michener et al. 1994; Neff and Rozen 1995). Originally described as a monotypic genus, Anthemurgus is currently considered a subgenus of Protandrena (Bossert et al. 2022). This species is considered “rare” across North America in Michener et al. (1994) but can be locally abundant in areas where Passiflora lutea L., its host plant, is abundant, particularly in Mississippi (Parys et al. 2018). This species is known from central Texas, Arkansas, Kansas, Illinois, and east to North Carolina in the United States (Fig. 1d; Mitchell 1960; Michener et al. 1994; Neff and Rozen 1995; Holland and Lanza 2008) and was recently collected in Mississippi in 2018 (Parys et al. 2018). Neff and Rozen (1995) described detailed foraging behavior, nests, and larval characteristics, but the bee has received little subsequent study.

Fig. 1.

Lateral (a) and head (b) images of voucher specimen of P. passiflorae, deposited at the NPIC in Logan, UT. P. passiflorae on its host plant, P. lutea (C). Map of P. passiflorae distribution in the United States. Red dots indicate publicly available records while the teal dots indicated the location that the specimens were collected for this study. Photographs by KA Parys.

P. passiflorae is one of the few truly monolectic bee species, known to solely forage on a single host-plant species (Cane 2020). The yellow passionflower, P. lutea L., is endemic to the southeastern United States and occurs in mesic open woodlands and vernal shaded areas from central Texas to Florida, and north to southern Kansas and Pennsylvania (Goldman and MacDougal 2015a). It is also a host plant for several species of Heliconius butterflies (Lepidoptera: Heliconiinae) across the southernmost parts of the range (Cardoso 2008; Estrada and Gilbert 2010).

Specialist bees are often thought to be more efficient pollinators than generalist visitors, but female P. passiflorae foraging for pollen are thought to act as pollen thieves on P. lutea, rarely ever contacting the stigmas during the process (Parker 1981; Neff and Rozen 1995). Host-plant specialists are common in several groups of bees, but P. passiflorae is the only known bee to be oligolectic on a species in the family Passifloraceae (Neff 2003; Rasmussen et al. 2020). The pollen collection behavior of P. passiflorae appears to be unique to P. lutea, with no foraging observed on sympatric species with similar phenologies, Passiflora incarnata L. and Passiflora affinis Engelm. (Neff and Rozen 1995; Goldman and MacDougal 2015b,c).

Here, we present a high-quality reference genome for P. passiflorae, the first genome sequence for the genus Protandrena and subgenus Anthemurgus and one of only a few available genomes outside of the nominal genus Andrena. Two genomes are available for the genus Perdita, which is in the same subfamily (Panurginae), but a different tribe. This genome is important because of the taxonomic position of the bee and its unique biology as a specialist pollinator of passion flowers. This genome will facilitate understanding the genetic basis of oligolecty/monolecty within bees, help elucidate andrenid phylogeny and evolution and contribute to conservation of an endemic plant. Insect genomic initiatives like the i5k consortium, Ag100Pest, AgriVectors, and now Beenome100 (https://beenome100.org) are highlighting the benefits of creating high-quality and improved genomic resources and building open access tools to make them more accessible (Robinson et al. 2011; i5K Consortium 2013; Giraldo-Calderon et al. 2015; Childers et al. 2021; Saha et al. 2021; Isaacs-Thomas 2022).

Methods and materials

Sample collection

Individuals of P. passiflorae were collected by net directly off blooms of P. lutea in Cleveland, Mississippi, located in Bolivar County (Fig. 1; 33.73521, −90.73244). The collection of 5 live males, which are haploid, occurred on the July 21, 2019, and bees were placed individually into cryo tubes, placed on ice, returned to the laboratory, and placed directly into a −80°C freezer. We have designated specimen SIMRU20858 as the official specimen voucher and it has been deposited into the National Pollinating Insects Collection (NPIC) at the USDA-ARS Pollinating Insects Research Unit in Logan, UT (Fig. 1a-c; unique voucher# SIMRU36530).

The distribution map (Fig. 1d) was generated using R version 4.3.1 (R Core Team 2023) and the following R packages: cowplot v. 1.1.3 (Wilke 2024), ggrepel v. 0.9.5 (Slowikowski 2024), ggspatial v. 1.1.9 (Dunnington 2023), googleway v. 2.7.8 (Cooley et al. 2020), maps v. 3.4.2 (Becker et al. 2023), rnaturalearth v. 1.0.1 (Massicotte and South 2023), rnaturalearthdata v. 1.0.0 (South et al. 2024), sf v. 1.0.16 (Pebesma 2018; Pebesma and Bivand 2023), and tidyverse v. 2.0.0 (Wickham et al. 2019), with available distribution data (GBIF 2024). Images of the voucher specimen (Fig. 1a and b) were taken using a Keyence VHX-7000.

DNA extraction and sequencing

DNA from a single adult male specimen (specimen# SIMRU36530) was used for all genome sequencing. For long-read sequencing, high molecular-weight (HMW) DNA was extracted from part of the abdomen using a Qiagen MagAttract HMW DNA kit (Qiagen, Hilden, Germany) and DNA quality was assessed using a Denovix DS-11 fluorometer/spectrophotometer (Denovix Inc, Wilmington, DE) for purity and quantification and a Fragment Analyzer (Agilent, Santa Clara, CA) for size assessment. The HMW DNA was sheared to 15–20 kb using a Megaruptor 3 (Diagenode, Denville, NJ, USA) and a single SMRTbell library was prepared using the SMRTbell 3.0. library preparation kit following manufacturer's protocols for low input samples (Pacific Biosciences, Menlo Park, CA). The final library was sized with Ampure PB beads (Beckman Coulter, Brea, CA) to remove fragments <3 kb in length. The entire library was quantified using Qubit HS DNA reagents, sized on an Agilent Fragment Analyzer to determine molar concentration and then bound and loaded for sequencing on a single cell on a Pacific Biosciences Sequel IIe system. A HiC library was prepared from abdomen tissue. The proximity-ligated sequencing library was prepared from crosslinked tissue following standard protocols (adapted from e.g. Lafontaine et al. 2021) using restriction enzymes DdeI and DpnII and subjected to PE 150 bp sequencing on an Element Biosciences Aviti system (Element Biosciences, San Diego, CA).

Genome assembly, scaffolding, and quality control

We assembled the P. (Anthemurgus) passiflorae genome using a combination of long-read sequencing and proximity-ligation sequencing. After filtering out adapters from raw HiFi reads using HiFiAdapterFilt v. 0.2.3 (Sim et al. 2022) under default parameters, we performed genome assembly using HiFiASM v. 0.16.1-r375 (Cheng et al. 2021) with the “–n-hap 1” specification. This preliminary contig assembly was assessed for quality and completeness using BlobToolKit v 2.6.1 (Challis et al. 2020) and the benchmark of single-copy orthologs (BUSCOs; Waterhouse et al. 2018). We used GenomeScope2 to estimate genome size and coverage (Ranallo-Benavidez et al. 2020) and Merqury (Rhie et al. 2020) to assess kmer frequencies. With the HiC data, we scaffolded the contigs into chromosomes using yet another Hi-C Scaffolding tool (YAHS; Zhou et al. 2023), which also generates HiC contact maps via the Juicebox tool set. We manually curated the scaffolded assembly in Juicebox (​Robinson et al. 2018), then converted the curated scaffold using the Juicebox “juicebox_assembly_converter.py” script. Following conversion of the scaffolded assembly, we re-ran Blobtools and BUSCO to assess assembly quality and identify taxonomic origin of scaffolds. Scaffolds were assigned to their closest taxon using BLAST (Altschul et al. 1997) and Diamond (Buchfink et al. 2015) searches to NCBI nt and UniProt databases, respectively, within BlobToolKit. We removed any scaffolds that were neither arthropod nor “no-hit.”

Mitochondrial genome

Following the MitoHiFi v2 workflow (Uliano-Silva et al. 2023), we identified the mitochondrial genome of Andrena chekiangensis (Lin et al 2023; NCBI accession: NC_042768.1) as the closest reference genome. We used Mitos (Bernt et al. 2013) within MitoHiFi to identify and remove the mitochondrial genome from the cleaned, scaffolded whole genome assembly.

Repeats

In order to characterize repetitive regions within the P. passiflorae genome, we implemented RepeatModeler v2.0.4 (Flynn et al. 2020) to identify de novo transposable elements. Next, we soft masked de novo and known transposable elements from the draft genome assembly using RepeatMasker v4.1.5 (Smit et al. 2024; www.repeatmasker.org/). Both RepeatModeler and RepeatMasker are available within the Dfam consortium software container (https://github.com/Dfam-consortium/TETools/; Storer et al. 2021).

RNA extraction and sequencing

We sequenced RNA from 2 tissues to aid in genome annotation. Specifically, total RNA was extracted separately from abdomen and thorax tissues through homogenization of snap frozen tissue in Tri Reagents and then extraction using the Zymo Direct-zol Magbead Total RNA kit (Zymo Research, Irvine, CA) on a Kingfisher Flex 96 system (ThermoFisher, Waltham, MA). From total RNA, a poly(A) cDNA library was generated using the NEB Ultra II RNA Library Prep Kit (New England Biolabs, Ipswich, MA) using oligo(dT) selection. The final RNASeq libraries were sequenced on a PE 150 bp sequencing run on an Element Biosciences Aviti instrument (Element Biosciences, San Diego, CA).

Genome annotation

Following annotation approaches used in previous bee genome studies (e.g. Sless et al. 2022a; Schweizer et al. 2024), we predicted gene annotation using the BRAKER3 workflow (Gabriel et al. 2024). This workflow predicts genes using GeneMark-ETP and AUGUSTUS, using both the P. passiflorae thorax and abdomen RNAseq data, in conjunction with the OrthoDB v11 orthologous protein database for Arthropods (Kuznetsov et al. 2022). After annotation with BRAKER3, we performed functional annotation using Interproscan software v5.60–92.0 (Jones et al. 2014; Blum et al. 2020).

Phylogenetic inference

We constructed a phylogeny of bees by extracting ultraconserved elements (UCEs) from the iyProPass1 draft genome assembly and a selection of other genomes. To conduct this analysis, we downloaded all publicly available bee genomes from NCBI using the NCBI Datasets command line tools v. 15.6.0 (download date: February 23, 2024) and pruned the initial set of genomes to remove duplicate species, resulting in a final selection of 123 bee genomes plus one outgroup wasp genome (Pemphredon lugubris) (Kapheim et al. 2015; Sadd et al. 2015; Brand et al. 2017; Beadle et al. 2019; Kapheim et al. 2019; Wallberg et al. 2019; Heraghty et al. 2020; Kapheim et al. 2020; Waiker et al. 2020; Wee et al. 2020; Zhou et al. 2020; Shell et al. 2021; Sun et al. 2021; Zhang et al. 2022; Falk et al. 2022a,b; Sless et al. 2022b; Boyes et al. 2023; Ferrari et al. 2023; Jones et al. 2023; Koch et al. 2023; Pope et al. 2023; Taylor et al. 2023; Araujo et al. 2024; Falk and Crowley 2024; Koch et al. 2024; Santos et al. 2024; Schweizer et al. 2024). We then extracted UCE loci from genomes with Phyluce v. 1.7.3 (Faircloth 2016), following “tutorial 3” from the Phyluce documentation (https://phyluce.readthedocs.io/en/latest/tutorials/tutorial-3.html). UCEs were extracted by aligning the principal v2 UCE probe set for Hymenoptera from (Branstetter et al. 2017a) to the target genomes and by slicing out 1 kb of flanking sequence from either side of each UCE locus. For the probe alignment step, we set both the “min-coverage” and “min-identity” parameters to 80%.

Following UCE extraction, we used a series of Phyluce commands to generate a concatenated UCE matrix for phylogenetic analysis. We ran the phyluce_assembly_match_contigs_to_probes script using the principal version of the probe set and default parameters. Each locus was aligned with MAFFT v. 7.475 (Katoh and Standley 2013) and trimmed with Gblocks v0.91b (Talavera and Castresana 2007) using reduced stringency parameters (b1 = 0.5, b2 = 0.5, b3 = 12, b4 = 7). We then filtered the complete set of loci to have 98% taxon completeness and concatenated the alignments into a supermatrix. A phylogenetic tree was inferred using IQ-Tree v. 2.4 (Minh et al. 2020) and a general time reversible model with gamma among site rate variation (GTR + F + G4). To measure support, we conducted 1,000 ultrafast bootstrap replicates (UFB; Hoang et al. 2018) and 1,000 SH-like approximate likelihood ratio text replicates (Guindon et al. 2010).

Results and discussion

Assembly

DNA sequencing generated 2,571,591 Sequel reads, of which only 363 (0.014%) were removed with HiFiAdapterFilt. The estimated genome coverage using GenomeScope2 was 150×. The initial HiFiASM assembly consisted of 53 contigs, a L50 of 6, N50 of 19.799 Mb, and a genome size of 251.030 Mb (Fig. 2a). Scaffolding with HiC further improved the contiguity of the genome, placing contigs into 49 scaffolds with an N50 of 21.4 Mb. BlobToolKit identified 11 scaffolds of non-Arthropod origin, mapped to the plants P. lutea, Passiflora edulis, Acacia ligulata, Populus alba, and Gossypium hirsutum, the endosymbiotic bacterium Wolbachia pipientis, and the fungus Meira sp. Upon removal of these scaffolds, the final genome consisted of 38 scaffolds totaling 249,183,375 bp, an N50 of 5 Mb, L50 of 21,425,975 bp, N90 of 11 Mb, and L90 of 13,250,199 bp (Fig. 2b). The assembly has fairly complete Hymenoptera BUSCO, with 97.2% single-copy complete, 0.55% fragmented, 0.30% duplicated, and 2.82% missing (Fig. 2b). HiC scaffolding identified 12 chromosomes (Fig. 2c).

Fig. 2.

Genome assembly and quality metrics for the P. passiflorae genome. a) BlobToolKit snail plot of the initial draft genome contig assembly. The main plot is divided into 1,000 size-ordered bins around the circumference with each bin representing 0.1% of the 251 Mb assembly. The distribution of sequence lengths is shown in dark gray with the plot radius scaled to the longest sequence present in the assembly (31,838,323 bp, shown in red). Orange and pale-orange arcs show the N50 and N90 sequence lengths (17,524,680 and 3,032,719 bp), respectively. The pale gray spiral shows the cumulative sequence count on a log scale with white scale lines showing successive orders of magnitude. The blue and pale-blue area around the outside of the plot shows the distribution of GC, AT, and N percentages in the same bins as the inner plot. A summary of complete, fragmented, duplicated, and missing BUSCO genes in the hymenoptera_odb10 set is shown in the top right. b) Same as (a) but for the final scaffolded genome. c) HiC contact map of the P. passiflorae genome, with 12 chromosomes ordered from largest to smallest, from the top left to bottom right. d) Cumulative length of scaffolded, clean assembly for scaffolds ordered largest to smallest. The 2 lines indicate taxonomic assignment to Arthropoda (light blue) or no match (dark blue.).

The P. passiflorae genome has many quality metrics that meet or exceed the standards of the Earth BioGenome Project, such as a contig N50 well over 1 Mb (N50 = 19.8 Mb), and single-copy complete BUSCOs over 90% (Lawniczak et al. 2022). Additionally, the size of the scaffold genome (249 Mb) is well within the range of size of available genomes on NCBI (see Fig. 1 of Sproul et al. 2023 for meta-analysis) for Hymenoptera.

Repetitive DNA

RepeatModeler and RepeatMasker analyses classified then masked 20.27% of the P. passiflorae genome for repetitive elements. Retroelements (1.31%) and DNA transposons (1.63%) were the most common categories of classified repeats, with an additional 15.63% of the genome in unclassified repeats (Table 1).

Table 1.

Repetitive elements identified in the P. passiflorae genome using RepeatModeler and RepeatMasker.

	Number of elements	Length occupied	Percentage of sequence
Retroelements	7,084	3,263,493	1.31
SINEs:	652	103,829	0.04
Penelope:	0	0	0
LINEs:	2,666	1,224,956	0.49
CRE/SLACS	0	0	0
L2/CR1/Rex	753	193,026	0.08
R1/LOA/Jockey	1,297	822,839	0.33
R2/R4/NeSL	0	0	0
RTE/Bov-B	51	71,760	0.03
L1/CIN4	0	0	0
LTR elements:	3,766	1,934,708	0.78
BEL/Pao	127	125,971	0.05
Ty1/Copia	307	162,647	0.07
Gypsy/DIRS1	1,699	1,287,539	0.52
Retroviral	154	29,778	0.01
DNA transposons	13,961	4,060,460	1.63
hobo-activator	1,871	340,388	0.14
Tc1-IS630-Pogo	5,750	1,147,960	0.46
En-Spm	0	0	0
MULE-MuDR	587	148,042	0.06
PiggyBac	1,354	286,604	0.12
Tourist/Harbinger	38	4,199	0
Other (Mirage, P-element, Transib)	104	58,297	0.02
Rolling-circles	1,325	394,230	0.16
Unclassified	189,456	38,952,193	15.63
Total interspersed		46,276,146	18.57
Small RNA:	353	602,377	0.24
Satellites:	0	0	0
Simple repeats:	45,724	2,742,034	1.1
Low complexity:	9,081	485,074	0.19

Gene annotation

The BRAKER3 workflow using orthologous arthropod protein databases and the P. passiflorae RNAseq data annotated 12,098 genes and 15,353 total transcripts (Supplementary File 1). The mean gene length is 8,564 bp, with a mean of 7 exons per gene. In comparison to another Andrenidae genome, Perdita meconis (Schweizer et al. 2024), which was assembled from HiFi data only, the gene length for P. passiflorae is slightly longer (P. meconis: 3,623 bp) and the mean number of exons for P. passiflorae is slightly larger (P. meconis: 4 exons per gene). Sequencing of additional Andrenidae assemblies will expand our understanding of gene structure variation in this diverse group of bees. The inclusion of the RNAseq data reduced the total number of annotations compared to a BRAKER3 workflow using only the orthologous arthropod protein database (19,553 genes, 21,177 transcripts; results not shown). Interproscan run on the RNAseq and protein workflow was able to assign functional annotation to 10,958 of 12,098 genes and 14,203 of 15,353 transcripts.

Phylogeny

We recovered an average of 2,455 UCE loci from 124 genomes (range: 1,085–2,528 loci), including iyProPass1, and generated a concatenated dataset containing 1,657 UCE loci, 2,478,627 bp of sequence data and 1,688,610 informative sites. We recovered a robust phylogeny of bees that includes all bee families except Stenotritidae (Fig. 3). All nodes in the tree except one were maximally supported with 100% bootstrap and SH-like support values and relationships among bee families and major lineages match recent phylogenomic studies with more balanced taxon sampling (Branstetter et al. 2017b; Bossert et al. 2022; Almeida et al. 2023). Within Andrenidae, P. passiflorae was correctly placed as sister to the recently published genome of Perdita meconis Griswold and together they formed a clade sister to Andrena. P. passiflorae and P. meconis both belong to the subfamily Panurginae, but are in different tribes, Protandrenini and Perditini, respectively. This represents ∼54 million years of divergence (Bossert et al. 2022), which explains the long branches connecting the 2 species. An uncollapsed phylogeny with all 124 genomes is shown in Supplementary Fig. 1. This phylogeny shows that the taxonomic sampling of Andrenidae for full reference genomes remains incomplete and highly biased toward the single genus Andrena. By adding a new tribe to the tree, we fill in an important gap in sampling.

Fig. 3.

Phylogenetic tree based on the analysis 1,657 UCE loci extracted from 124 publicly available genomes, including the target species P. passiflorae. The tree was inferred with IQ-Tree2 and support was estimated using UFB replicates and SH-like approximate likelihood ratio test replicates. Support values are only shown if <100%. Terminal names include GCA numbers for reference. A complete tree without collapsed nodes is provided as Supplementary Fig. 1.

Data availability statement

This Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accession JAZGQT000000000. The version described in this paper is version JAZGQT010000000. Raw reads from HiFi, HiC, and RNAseq have been deposited in the NCBI Short Read Archive under BioProject PRJNA1072036. The reference genome is available on NCBI (GenBank ID: Genome Collections Accession number (GCA_047729815.1); https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_047729815.1/). Most data processing scripts are available on Ag Data Commons (https://doi.org/10.15482/USDA.ADC/25762431.v1), Gene annotations (Supplementary File 1), repeat annotation (Supplementary File 2), the phylotree file (Supplementary File 3), and the output table from repeatmasker (Supplementary File 4) are available as Supplementary files. Two additional scripts are also available as Supplementary files that provide the commands used to find and mask repetitive elements (Supplementary File 5) and annotate with BRAKER3 (Supplementary File 6). Occurrence data used to create Fig. 1c is available online from the Global Biodiversity Information Facility (GBIF); (DOI:10.15468/dl.447nds).

Supplemental material available at G3 online.

Funding

This work was supported by the United States Department of Agriculture, Agricultural Research Service Beenome100 initiative along with USDA ARS in-house projects #6066-21000-001-000-D (Stoneville, MS), #2080-21000-019-000-D (Logan, UT), and #2040-22430-028-000-D (Hilo, HI). RMS was supported by a competitive USDA-ARS administrator-funded postdoctoral award to MGB. Analyses were performed using computing resources funded by USDA Agricultural Research Service’s SCINet initiative (project #0500-00093-001-00-D).