Can ecological interactions drive evolutionary outcomes? Evidence from insect host shifts between parasitic and non-parasitic plants

https://doi.org/10.1098/rspb.2026.0100

Abstract

• Phytophagous insects have specialized on virtually every plant lineage. Parasitic plants, however, are uncommon hosts. Among insects, only a single lineage of weevils, the Smicronychini, has successfully radiated on both parasitic and non-parasitic plants in a large panel of distantly related Asterid families. This unusual pattern suggests that major host plant shifts have occurred over the course of their diversification. Through the analysis of a phylogenomic dataset, we reconstruct for the first time their evolutionary history and ancestral host plant associations. Our results show that independent host plant shifts occurred both from parasitic to non-parasitic hosts and between distinct parasitic lineages. These results suggest that host shift mechanisms can be driven by ecological opportunities provided by plant-plant interactions. This first evidence of extreme insect host plant shifts apparently mediated by parasitic plant-plant interactions emphasizes the core importance of ecological interactions as driving forces behind insect host plant shifts.

Figures & Tables

Figure 1: Host repertoire of Smicronychini weevils. From top to bottom: diagrams represent parasitic interactions between Smicronychini weevils and their host plants. Host plant families that make up the host repertoire of Smicronychini are shown below each corresponding diagram, with examples of associated weevils and known galls. Credits: BZ & JH, except for Orobanchaceae (Clémence Massard) and Asteraceae (@sinaloasilvestre, iNaturalist) pictures; diagrams with ©BioRender.

Figure 2: Phylogenomic tree based on the AHE dataset. A. Maximum likelihood tree of the AHE dataset. Nodes poorly supported (SH-aLRT < 80 or UFBS < 95%) are respectively highlighted with gray triangles and circles. For graphical purposes, some branches were shortened to half of their length (//). Pictures of weevils are linked to corresponding species. B. Partitioning scheme of a theoretical AHE loci. Flanking and inserted partitions (f1-3) are highlighted in grey and coding partitions (c1-3) are highlighted in green. The region targeted by the probes (P) is shown in red but does not appear in the final supermatrix. Credits for weevil pictures:  J.H.

Figure 3: Ancestral host plant reconstruction (ASR) of Smicronychini weevils and their current ecologies. The phylogenetic tree topology corresponds to the extended dataset ML tree pruned to one branch per species. The AHE backbone tree is represented with plain branches while less robust COI data are represented with dashed branches. Reported gall-inducing species are marked with a “x” between branch tips and species names. Pie charts represented on each node of the phylogeny correspond to the estimated likelihood of each character state computed by the ace function. The colors of pie charts and background shading correspond to host plant character states used in the ASR as detailed in the bottom-left table. Colored shading represents the most probable ancestral state estimated at each node and left blank for unknown states and uncertain ancestral states. Photo credits: B.Z. & J.H.

Appendices

Figure S1: Flowchart of the phylogenomic pipeline developed to process AHE data. Each step, numbered from 1 to 10 as referenced in the main text, is represented, from raw reads to the complete supermatrix. External tools and custom scripts (marked with an *) are given, along with a diagram representing each step of the pipeline.

Figure S2: Complete grafted tree of the extended dataset. ML tree generated with MFP+MERGE model on IQTREE v2.3.2 using the AHE topology as backbone. To increase computation time, -allnni option was not used. Bootstraps show SH-aLRT and UFBS values respectively.

Figure S3: Grafted tree of the extended dataset with one specimen per species. ML tree generated with MFP+MERGE model on IQTREE v2.2 using the AHE topology as backbone. Bootstraps show SH-aLRT and UFBS values respectively.

Table S1: Detailed interactions of species represented in the extended dataset. Informations marked with an asterisk (*) are considered unsure. Parasitic mode (PM): np = non-parasitic, hemip = hemiparasitic, p = holoparasitic.

Table S2: Complete specimen data. Sample IDs highlighted in red were removed from the final dataset, those highlighted in yellow were merged to an AHE sample as mentionned in "COI" column.

Zenodo supplementary files

01_PREPROCESS: Commands and files used to gather AHE probes targeted sequences from Anthonomus grandis genome

see README.sh for a detailed description of files and commands used to obtain them.

02_SCRIPTS: Scripts used during phylogenomic analysis.

Most scripts are used by the shell pipeline "04-FcC_HMMC_PB_elongated_probes.sh". Other softwares used by the pipeline that are available online are not listed but can be found inside the pipeline script or in figure S1.

bed and gapped_refs folders contain necessary files for the pipeline to run, see 01_PREPROCESS material to see how they were generated.
bed - genomic data from Anthonomus grandis reference used by the pipeline steps 7-8
gapped_refs - probe consensus sequences targeted on Anthonomus grandis reference genome, split fasta from consensus_probes_gap.fasta

04-FcC_HMMC_PB_elongated_probes.sh - Semi-automatic pipeline routinely used during analyses
check_het.R - estimates genetic distance
clean_duplicated_seq.sh - removes or makes a consensus of sequences from specimens with multiple sequences depending on genetic distance
count_gap_per_site.py - computes number of gaps per site
drop_short_seq.py - removes sequences shorter than a given fraction of non-gaps
HMMcleanNuc.pl - Nucleotide derived HMMCleaner script: Di Franco, A., Poujol, R., Baurain, D. & Philippe, H. (2019). Evaluating the usefulness of alignment filtering methods to reduce the impact of errors on evolutionary inferences. BMC Evol. Biol., 19, 21.
improve_parts.R - updates partition files based on Anthonomus grandis reference genome.
make_codon_partitions_forProbe_from_charset.py - makes codon partitions from a nexus "charset" file
make_codon_partitions_forProbe.py - makes codon partitions from partition files ".part"
make_gapped_parts.py - makes partition files based on the probe reference considering gaps in probe sequence
make_good_fasta.py - converts block fasta to one line format
PaulBlock.pl - removes columns with more than a given threshold of gaps (only for non-coding parts)
remove_seq_given_ID.py - removes sequences with a strict given ID
remove_seq_given_relaxed_ID.py - removes sequences with a given pattern
seqCat.pl - cf file header
seqConverter.pl - cf file header
trim_partition_file.R - trims sequences overlapping with each other based on bed formatted reference genome

03_TREES: Intermediate and finales files generated by the phylogenomic pipeline

    AHE: Phylogenomic analyses on AHE dataset

AHE_2024_v8.3_70.phylip - Final sequence file obtained with the phylogenomic pipeline, only keeping loci with >70% species
AHE_2024_v8.3_70.nex - Partition file with 1 partition per flanking/coding regions (see Fig. 3B in main text)
AHE_2024_v8.3.70.MFPMERGE* - Tree generated with MFP+MERGE model

    COI: Phylogenetic analyses on COI dataset

COI_2025.v1.fasta - fasta alignment of all cytochrome oxydase I samples cleaned from bad sequences
COI_2025.v1.noemptyseq* - COI tree of all samples, cleaned from empty sequences (corresponding to AHE specimen without COI)

    grafting: Phylogenomic analyses on extended dataset

# Phylogenetic analyses
constraintfile.txt - Backbone generated by AHE analyses
AHE_COI_2025.v1* - Grafted tree with all samples, launched without --allnni iqtree option to reduce computation time
AHE_COI_2025.v2.monosp.nex | .phylip - Nexus and phylip files resulting from manual removal of all but one representative of each species based on AHE_COI_2025.v1.grafting.noallnni and COI tree (for dentirostris and rubricatus), see tips highlighted in red in AHE_COI_2025.v1.grafting.noallnni.treefile.keepsp.pdf for samples kept in v2
AHE_COI_2025.v2* - Grafted tree with one sample per species

# Topology tests
topo1.newick - Same topology as AHE_COI_2025.v2bis.grafting.monosp.treefile
topo2* - Alternative topology assuming the monophyly of North American Smicronyx species
topo3* - Alternative topology assuming the monophyly of North American Smicronyx and Promecotarsus species
topotest.iqtree - results of the topology test showing no significative difference between the three tested topologies

04_ASR: 

Ancestral character state estimations (ASR) are run with ASR_v4.R

topo1 = paraphyletic Asteraceae-feeding species
topo2 = monophyletic Asteraceae-feeding species
Both topologies are equally likely (cf ../03_TREES/grafting/topotest.iqtree)

ACE_HPPM_ER.topo1.pdf - ASR with ace() function on topo1
ACE_HPPM_ER.topo2.pdf - ASR with ace() function on topo2
ACE_HPPM_Likelihoods.csv - Ancestral state likelihoods at each node of ACE_HPPM_ER.topo2.pdf
ACE_HP_ER.topo2.png - ASR with ace() function on topo2, with only host plant families as character states
ACE_PM_ER.topo2.png - ASR with ace() function on topo2, with only parasitic types as character states
ACE_PMNH_ER.topo2.png - ASR with ace() function on topo2, with only parasitic types without hemiparasites as character states on ER matrix
ACE_PMNH_SYM.topo2.png - ASR with ace() function on topo2, with only parasitic types without hemiparasites as character states on SYM matrix
ASR_collapsed_pruned_topo1_HPPM_ER.pdf - ASR with Phytools make.simmap etc. on topo1. NA's were removed and bootstraps below thresholds were left polytomic
ASR_collapsed_pruned_topo2_HPPM_ER.pdf - same on topo2
ASR_pruned_topo2_HPPM_ER.pdf - same but without collapsed nodes
ASR_chrono_topo2_HPPM_ER.pdf - same but on an ultrametric tree
AHE_COI_2025_70coll.v2.grafting.monosp.contree - topo1 collapsed if bootstraps <70
AHE_COI_2025.v2.grafting.monosp.contree - topo1
host_plants.txt - table with host repertoire of each species (see Table S1)
topo2_70coll.rooted.contree - topo2 collapsed if bootstraps were < 70
topo2.rooted.contree - topo2

Raw data & sequences

Raw target capture data and assembled AHE results are publicly available in GenBank (NCBI) under the accession PRJNA1244829: https://dataview.ncbi.nlm.nih.gov/object/PRJNA1244829?reviewer=a4n8mt0a12r7if07871en55mj8. COI sequences generated for the present study are publicly available on BOLD systems under the project name SCRYX and are accessible through this link: https://www.dropbox.com/scl/fo/y2o7ufcuhzm0tiz7441s8/AAMuIDfGhkef7KZ5bDax0WE?rlkey=z1078qn9bf4geer0f1srae81i&dl=0.