Genome sequencing and annotation

To understand the evolution of genes related to PGN pathways, we sequenced the genomes of four divergent aphid lineages, including the first genome sequences for subfamilies Fordinae (Geopemphigus sp.), Phyllaphidinae (Stegophylla sp.) and Pemphiginae (Pemphigus sp.), and the second genome sequence for subfamily Chaitophorinae (Chaitophorus sp.) after Sipha flava (NCBI BioProject PRJNA472250). The Geopemphigus genome also represents the first aphid genome for which Buchnera has been lost and replaced with an unrelated endosymbiont [29]. For each genome, we generated 38–58 Gb of genomic data from one short insert size Illumina library (S2 Table) and 5–7 Gb RNA-seq data (S3 Table). After assembly, estimated genome sizes range from 245 Mb—658 Mb (S4 Table). The sequenced genomes show high completeness based on the presence of single-copy Benchmarking Universal Single-Copy Orthologs (BUSCOs) for Hemiptera (n = 2,510) [38], ranging from 97%-99% for both the assemblies (S5 Table) and our gene annotation results (S6 Table). These values are similar to those for other aphid genome sequences [33]. All sequencing data can be found at NCBI under the accession number PRJNA758084.

Aphid HTG incidence varies across aphid species

In order to understand 1) the incidence of A. pisum HTG homologs among different aphid lineages and 2) the evolutionary history of host-encoded PGN genes, we reconstructed the aphid phylogeny using the four newly sequenced aphid genomes, 13 available aphid genomes, and six related insects in suborder Sternorrhyncha included as outgroups (S1 Table). The phylogeny was based on 829 single-copy genes found in at least 80% of the species (Fig 2). Genes with putative activities pertaining to PGN metabolism, including peptidoglycan recognition proteins (PGRPs), lysozymes, and aphid HTGs (bLys, ldcA, amiD, and rlpA), were identified by BLAST search using A. pisum protein sequences as queries (S7 Table). With the exception of Daktulosphaira vitifoliae [39], the outgroup species contain bacterial symbionts, none closely related to Buchnera.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Ancestral reconstruction of aphid genes that putatively interact with Gram-negative bacterial PGN.

The aphid phylogeny was constructed using maximum likelihood from a concatentated nucleotide alignment of 829 core genes shared by all 23 insect species. Bootstrap supports after 1,000 replicates are shown, with red branches indicating supports < 70%. All symbionts are Buchnera except those of Geopemphigus sp, indicated by an asterisk, which contains a novel Bacteroidetes symbiont. Host species containing known co-obligate or uncharacterized facultative symbionts are indicated with an obelisk (†) or a double obelisk (‡), respectively. Each species code represents the combined first two letters of the genus and species. The ancestral state (gray) and placement of gene gains (black), duplications (black), and losses (white) were determined using Count (38). Numbers in shapes indicate the number of gene copies involved in the event. Shapes indicate distinct gene families, and colored borders denote the four RlpA paralog types, named after the A. pisum rlpA genes.

https://doi.org/10.1371/journal.pgen.1010195.g002

We first evaluated the distribution of genes encoding enzymes that target PGN and genes that represent conserved components of insect innate immunity (S7 Table). PGRPs sense bacterial infection by binding and/or cleaving PGN and activating downstream pathways that stimulate the humoral immune response, which includes increased production of bacteriostatic lysozymes [40]. Lysozymes, characterized by their N-acetylmuramidase activity, include members from four different glycoside hydrolase families: chicken- and invertebrate-type (c- and i-types, respectively; GH22), goose-type (g-type; GH23), viral-type (v-type; GH24), and Chalaropsis-type (ch-type; GH25) [40]. Both PGRPs and c-type lysozymes are evolutionarily conserved in insects [40,41]. We found no genes encoding PGRPs or c-type lysozymes among aphid or phylloxerid genomes, but their presence in the scale insect, whitefly [42], and psyllid [43,44] genomes support their loss in a shared ancestor of the Aphidomorpha infraorder (Fig 2). The absence of these key immune genes may have facilitated the establishment of endosymbioses in aphids [45], though their absence from phylloxerids, which lack symbionts [39] and presence in other outgroups that harbor obligate symbionts suggest that alternative forces underlie their loss from the Aphidomorpha. We observed little variation in the number of invertebrate-type (i-type) lysozymes among most insects investigated (Fig 2), indicating a functional difference between these and c-type lysozymes. Major exceptions are Cinara cedri (Lachninae) and Eriosoma lanigerum (Eriosomatinae) which harbor eight to nine i-lysozyme paralogs compared with three or four paralogs in other species. Finally, we found bLys in all Aphidomorpha genomes, indicating that this gene family was acquired concomitantly with the loss of PGRPs and c-type lysozymes (Fig 2). The ch-type lysozyme domain of bLys is most closely related to domains in prophages of Wolbachia-like bacteria [13,46] and is minimally expressed within bacteriocytes, implying a role outside of symbiosis [13]. The multiple independent HGTs of ch-type lysozymes to all domains of life [46] and the unique N,O-diacetylmuramidase activity described for some ch-type lysozymes [47,48] suggest that aphid bLys is more likely involved in host defense against bacterial pathogens than in symbiont PGN metabolism.

Unique to superfamily Aphidoidea are the HTGs rlpA, ldcA, and amiD, all of which were acquired in a shared ancestor of this superfamily (Fig 2). With the exception of an amiD homolog within the D. citri genome, none of the bacteriocyte-expressed HTGs found in aphids were detected outside of the Aphidoidea (Fig 2). In the case of AmiD, phylogenetic analysis revealed that the D. citri homolog clusters separately from the aphid copies, indicating that D. citri amiD was acquired from a bacterial source independently of aphid amiD (S1 Fig).

Aphid lineages vary in whether HTGs have been lost or duplicated (Fig 2). RlpA subfamilies are maintained to different extents among aphid lineages, with each genome containing at least one type (Fig 2). We examined the phylogenetic relationships of aphid RlpA proteins and found that each sequence falls within one of four paralogous subfamilies: RlpA4-like, RlpA3-like, RlpA1-like, and RlpA2/5-like (S2 Fig). RlpA4-like proteins are the most basally branching subfamily and RlpA2/5 is the most recent branch, supporting the initial analysis of A. pisum rlpA genes [13]. Ancestral reconstruction indicates that the most recent aphid common ancestor had all rlpA subfamilies, implying that the underlying gene duplications occurred in a common ancestor of the Aphidoidea (Fig 2). In contrast, amiD and ldcA have been lost independently five and two times, respectively (Fig 2). The repeated loss of these genes provides an opportunity to test our hypothesis—if aphid HTGs contribute to symbiont PGN metabolism, the loss of aphid HTGs should correlate with the absence of interacting symbiont genes.

Aphid amiD and ldcA coincide with the loss of symbiont PGN

We anticipated that the symbionts of host species lacking amiD and/or ldcA genes might display greater loss of genes from PGN-related pathways due to a functional link between these host genes and symbiont cell wall metabolism. To test this idea, we looked for patterns of coincidence between host and symbiont PGN gene repertoires among all available sequenced aphid species. First, we measured the incidence of 138 bacterial genes, collectively involved in PGN metabolism or two potentially related pathways, within the primary symbiont genome of each aphid species (Fig 3 and S8 Table). As noted previously, Buchnera from aphids of tribe Macrosiphini (subfamily Aphidinae) harbor the most complete set of PGN metabolism genes among Buchnera species, maintaining a nearly complete lipid II synthesis pathway and the minimal enzyme functions necessary for cell wall synthesis (glycosyltransferase and d,d-transpeptidase) (Fig 3) [21,27]. Though Macrosiphini Buchnera generally maintain genes encoding two of the three essential activities for cell wall remodeling, an amidase (amiB) and at least one lytic transglycosylase (LT; mltA, emtA) gene, they often lack the gene typically encoding the third, endopeptidase activity (mepM) (Fig 1). In comparison, Buchnera of the tribe Aphidini (subfamily Aphidinae) display more frequent loss of murF, murC, and/or murE of the lipid II synthesis pathway. Most other subfamilies (Chaitophorinae, Phyllaphidinae, Lachninae, Eriosomatinae, and Pemphiginae) exhibit near-complete loss of all PGN pathway genes, while Buchnera from Hormaphis cornu (subfamily Hormaphidinae) represents an intermediate, maintaining a single cell wall synthesis gene (mrcB, encoding the bifunctional glycosyltransferase/d,d-transpeptidase PBP1B) and partial lipid II synthesis pathway (Fig 3). Notably, no aphid primary symbionts, including Buchnera of most aphids and Candidatus Skilesia alterna (referred to hereafter as Skilesia) of Geopemphigus, maintain any of the PGN recycling genes used in our searches (S8 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Incidence of genes underlying PGN metabolism and related pathways from 17 aphid host-symbiont species.

Filled cells indicate gene presence, while empty cells indicate gene absence. For aphid genes, numbers indicate the number of gene paralogs identified per genome. For symbiont genes, filled cells all indicate a single copy per genome. All symbionts are Buchnera except those of Geopemphigus sp, indicated by an asterisk, which contains a novel Bacteroidetes symbiont. Host species containing known co-obligate or uncharacterized facultative symbionts are indicated with an obelisk (†) or a double obelisk (‡), respectively. Each species code represents the combined first two letters of the genus and species (S1 Table). Colored circles indicate genes with dual-membership in the pathway of the same color, while black indicates dual-membership in amino acid metabolism.

https://doi.org/10.1371/journal.pgen.1010195.g003

Next, we considered whether the presence or absence of host PGN genes coincides with the degree of symbiont PGN gene loss. The ubiquity of aphid i-type lys and bLys genes in host genomes suggests no involvement of these genes with symbiont PGN metabolism (Figs 2 and 3). While the incidence of rlpA paralog types varies among aphid lineages, all species contain multiple rlpA genes, implying no relationship with symbiont gene repertoire. This is less surprising when the functional basis of RlpA’s role in bacterial cell division is considered. Bacterial RlpA proteins participate in cell division largely via the N-terminal sporulation-related repeat (SPOR) domain (Pfam 05036) that binds “denuded” (stemless) PGN glycan chains that accumulate at the septum [49,50]. In contrast, aphid RlpA proteins contain only the catalytic double-ψ ß-barrel (DPBB) domain (Pfam 03330) of RlpA [12,13]. While the RlpA DPBB domain is unique among LTs for its ability to cleave denuded PGN [51,52], the absence of SPOR domains from aphid RlpA proteins suggests that these enzymes likely do not participate in symbiont cell division. Furthermore, aphid RlpA proteins contain one or more eukaryotic inhibitor cysteine-knot (ICK)/knottin domains that are unrelated to PGN metabolism [12,13], further distancing RlpA from a role in this pathway. Collectively, these observations suggest that the DPBB domain may serve an alternative function in aphid RlpA proteins.

Aphid ldcA exhibits a mixed distribution among host species harboring symbionts with and without PGN. The uniform conservation of ldcA in the Aphidinae subfamily, where Buchnera maintains most PGN pathway genes (Fig 3), supports a relationship between ldcA and symbiont PGN metabolism, at least in this lineage. Aphid ldcA is also present in H. cornu, whose Buchnera harbors a partial PGN pathway (Fig 3). However, aphid ldcA is present in three aphid subfamilies with symbionts lacking PGN pathway genes (Eriosomatinae, Lachninae, and Pemphiginae) and is absent in two (Chaitophorinae, Phyllaphidinae). While the role of ldcA in the former lineages is unclear, Buchnera from each of the latter species has undergone significantly more gene loss relative to symbionts of other aphids [27]. Aphid ldcA is also absent in Geopemphigus sp. (Fordinae), where excessive genome reduction in Buchnera is thought to have led to its evolutionary replacement with the Bacteroidetes symbiont Skilesia [29]. Thus, our data suggest that the two independent losses of ldcA in these clades are associated with extensive loss of Buchnera genes, including most PGN genes (Figs 2 and 3).

In contrast, the incidence of aphid amiD in different lineages strongly suggests a link to symbiont cell wall metabolism. The loss of host amiD from most subfamilies with highly reduced Buchnera genomes (Chaitophorinae, Phyllaphidinae, Eriosomatinae, and Pemphiginae) or Buchnera loss (Fordinae) suggests that the gene is expendable in the absence of Buchnera PGN (Fig 3). The putative ability of AmiD to cleave stem peptides from anhydro-MurNAc suggests a dependence on the prior activity of LTs [53], which are distinct from lysozymes in their production of cyclic anhydro-MurNAc. As LTs are found only in Buchnera from Aphidinae aphids (emtA and/or mltA) and Skilesia (mltG), the presence of aphid amiD largely correlates with LT incidence, with the exception of C. cedri (Fig 3). In tribe Aphidini, the loss of host amiD coincides with that of murF, murC, and/or murE in symbionts—while the encoded enzymes putatively function in two distinct pathways of PGN metabolism (remodeling and lipid II synthesis, respectively), all four enzyme activities involve PGN stem peptides (Fig 1), implying the possibility of a functional relationship. Alternatively, a lineage-specific functional redundancy of host amiD with symbiont amiB could have rendered amiD obsolete. The presence of both aphid amiD and ldcA in H. cornu coincides with incomplete PGN gene loss in its symbiont (Fig 3), suggesting a stronger reliance on host HTGs for symbiont PGN remodeling. Finally, though Buchnera from Cinara cedri (Lachninae) lacks all PGN pathway genes, the host maintains both amiD and ldcA. Many Lachninae aphids harbor secondary symbionts that have taken on essential roles for host survival [54–56]—aphid PGN genes may regulate PGN metabolism in these secondary symbionts instead of Buchnera.

We also evaluated the symbiont gene repertoire of pathways related to but outside of PGN metabolism. Cell division is closely coordinated with cell wall synthesis and remodeling [34], while assembly of flagella basal bodies FBBs involves interaction with the cell wall via the ß-N-acetylgucosaminidase activity of FlgJ [35,36]. Like the PGN pathway, loss of both cell division and FBB genes is more extensive in all subfamilies relative to the Aphidinae (Fig 3). However, there is no apparent coincidence in these losses with those of PGN pathway genes that would implicate a dependence of either pathway on the cell wall. For example, though symbionts of both C. viminalis (Chaitophorinae) and Stegophylla sp. (Phyllaphididae) lack virtually all PGN genes, the former maintains many cell division genes, similar to the tick-borne pathogens of the genus Ehrlichia [21], while the latter lacks all of them (Fig 3), a unique phenomenon among obligately intracellular bacteria [21,26]. Similarly, Buchnera of C. cedri, E. lanigerum (Eriosomatinae), and P. obesinymphae (Pemphiginae) lack the PGN pathway but maintain many FBB genes (Fig 3). Furthermore, cell wall synthesis and remodeling genes that are known components of the E. coli divisome are not preferentially maintained in aphid symbionts that lack the PGN pathway but maintain cell division genes (Fig 3) [35]. Additionally, all Buchnera FlgJ enzymes lack the PGN hydrolase domain, effectively demonstrating a decoupling of FBB assembly to cell wall metabolism. Thus, the Buchnera cell wall generally does not appear to support cell division or FBB assembly.

Aphidini ldcA genes are influenced by positive selection

Aphid HTGs present in species harboring symbionts undergoing gene loss in the PGN pathway or lacking it altogether may have evolved alternative functions through natural selection. Because ldcA is present in species with varying degrees of PGN pathway degradation, we were able to test which conditions might lead to positive selection. The most complete PGN pathways are found in tribe Macrosiphini, followed by tribe Aphidini—PGN pathways are near-absent in subfamily Hormaphidinae and entirely absent in all other subfamilies with Buchnera. We tested for lineage-specific diversifying selection in the aphid ldcA gene family using BUSTED [57], selecting different sets of test lineages based on the criteria described above. For example, we asked whether diversifying selection has influenced ldcA in members of the tribe Aphidini, where the loss of aphid amiD and Buchnera murF, murC, and/or murE may signal the absence or modification of stem peptides from Buchnera PGN in this lineage, leading to a potential shift in LdcA function, which also putatively targets stem peptides. We found evidence for positive selection only when all lineages or Aphidini-specific ldcA genes were selected as the test group (Table 1), suggesting that host ldcA is undergoing adaptive evolution in at least one member of the Aphidini. These results are especially interesting in the light of our recent biochemical characterization of A. pisum LdcA, which shows that the aphid enzyme is multifunctional and plays a novel role in PGN degradation relative to its homolog in E. coli [22]. In contrast, we did not detect evidence for positive selection in ldcA when test groups included: all Macrosiphini species, species with symbionts lacking PGN genes and lacking secondary symbionts (P. obesinymphae and E. lanigerum), species with intermediate PGN gene loss (H. cornu), and species in which ldcA has been duplicated (Sitobion miscanithi, Rhopalosiphum padi, C. cedri, and P. obesinymphae) (Table 1). Additionally, we were unable to detect evidence for positive selection of amiD when all genes or only duplicated genes were selected as test groups (Table 1). Thus, tribe Aphidini stands out in showing evidence of positive selection for ldcA. However, the lack of evidence for positive selection in some of our BUSTED tests may reflect lack of power due to small sample sizes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. BUSTED tests for diversifying selection among aphid ldcA and amiD gene families (18 and 9 protein sequences, respectively).

https://doi.org/10.1371/journal.pgen.1010195.t001

Both Macrosiphini and Aphidini Buchnera build cell walls

The observation that Buchnera of most Aphidini aphids lack murF, murC, and/or murE (Fig 3), representing three critical steps in lipid II biosynthesis (Fig 1), raises the question of whether these symbionts are capable of synthesizing a cell wall at all [26]. The loss of these particular genes may be coincidental, reflecting the breakdown of cell wall synthesis altogether rather than evidence of functional linkage. To test this hypothesis, we developed a method to fluorescently label Buchnera PGN via incorporation of ethynyl-d-alanine (EDA), a d-alanine analogue compatible with click chemistry [58,59], by adapting a host feeding approach previously used to label symbiont PGN in the mealybug system [17]. Aphids were reared from birth on individual leaves immobilized in agarose gel containing EDA until reaching seven days of age (Fig 4A). To validate our approach, Buchnera cells were isolated from pea aphids (tribe Macrosiphini) reared on EDA, fixed, and stained with azide-linked Alexa Fluor 488 (AF488) via click reaction. Buchnera derived from hosts fed EDA exhibited green fluorescence at the cell periphery while control aphids raised on ethynyl-l-alanine (ELA) did not (Fig 4B), indicating 1) transport of EDA from host plant xylem to phloem, and 2) selective incorporation of EDA from plant phloem into Buchnera cell walls. Furthermore, the observed fluorescence was definitively linked to Buchnera cell wall stem peptides by treating cells with PGN-degrading enzymes prior to fixation and counter-staining with the PGN backbone-specific probe wheat germ agglutinin CF-640R conjugate (WGA640). Hen eggwhite lysozyme (HEWL) and E. coli AmiD target distinct PGN features: the backbone and stem peptides, respectively. We observed a decrease in fluorescence of both EDA-AF488 and WGA640 following HEWL treatment and of EDA-AF488 alone for AmiD (Fig 4B and 4C and S9 Table), firmly associating AF488 fluorescence with PGN stem peptides.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Labeling of A. pisum LSR1-c Buchnera cells with EDA by host feeding.

A) Aphids were reared on plant leaves immobilized in agar containing ethynyl-d-Ala (EDA) or the control l enantiomer from birth until the 4^th instar larval stage, after which Buchnera cells were purified from aphid homogenate. B) Cells were fixed and stained with azide-linked Alexa Fluor 488 to label PGN stem peptides (PGN stem peptides), wheat germ agglutinin CF-640R conjugate to label the PGN glycan backbone (PGN glycans), and the counter-stain DAPI to label DNA (DNA counter-stain). Hen eggwhite lysozyme (HEWL; PGN glycan-targeting) and E. coli AmiD (PGN stem peptide-targeting) were used to demonstrate incorporation of EDA within Buchnera PGN stem peptides. Images were falsely colored to aid visualization. The white scale bar represents 5 μm. C) For each combination of enzyme and probe, the average fluorescence intensity (AFI) per area was measured for each of n Buchnera cells using ImageJ software (S9 Table). Significant differences between means are indicated by compact letter displays above each boxplot—means with no letters in common are significantly different by the Dunn’s Multiple Comparison test (p-value < 0.01 following adjustment for false discovery).

https://doi.org/10.1371/journal.pgen.1010195.g004

Next, we applied the same cell wall labeling strategy to R. maidis aphids (tribe Aphidini) and pea aphids either cured of or containing the facultative symbiont Serratia symbiotica. Symbiont identities were confirmed using diagnostic FISH microscopy alongside the clickable cell wall probes (Fig 5). As expected, no cell wall fluorescence was observed in symbionts fed ELA. However, Buchnera cell walls were labeled with EDA in both aphid species (Fig 5A), demonstrating that even species lacking aphid amiD and Buchnera murC, murE, and murF can synthesize a cell wall. Additionally, pea aphids harboring Serratia show only minor incorporation of EDA into Serratia symbionts relative to Buchnera (Fig 5B), suggesting that Serratia differs from Buchnera in their primary source of d-Ala.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. FISH and cell wall labeling of Macrosiphini and Aphidini aphid symbionts.

A) Pea aphids, both cured (Austin-c; left column) and uncured (Austin; middle column) of secondary Serratia symbionts, and R. maidis (Rma-NM-1; right column) were fed ELA (left-side) or EDA (right-side) via their host plants as described in Fig 4A. Embryos removed from 4^th instar larvae were fixed, treated with azide-linked Alexa Fluor 488 (AF488), hybridized with Buchnera- and Serratia-specific Cy5- and Cy3-labeled riboprobes, respectively (S10 Table), and counter-stained with DAPI. B) Enlarged cutout images (white boxes in A) of EDA-labeled, uncured A. pisum show overlapping AF488 fluorescence (i) with Buchnera cells (ii) and, to a minor extent, Serratia cells (iii). A bacteriocyte containing Serratia is highlighted with a dashed line in each image. The white scale bars represent 20 μm. Images were falsely colored to aid visualization.

https://doi.org/10.1371/journal.pgen.1010195.g005

Aphid collection

Aphids were collected from multiple locations between June, 2020 and May, 2021 (S1 Table). Aphids were removed from leaves or galls, flash frozen in liquid nitrogen, and stored at -80°C. Excess wax was removed from Geopemphigus aphids by dipping individuals in Dawn Professional Grade Liquid Detergent (0.1% v/v), then rinsing with deionized water. Multiple individuals regardless of developmental stage were pooled per sample, with each Geopemphigus and Pemphigus sample derived from a single gall, and thus expected to contain members of a single all female clonal lineage. Geopemphigus sp. is an undescribed species corresponding to the “straight galls” found on Pistacia texana [29].

Sequencing and annotation

Genomic DNA was extracted from aphid samples using the DNeasy Blood & Tissue Kit (Qiagen) following the manufacturers’ protocol for DNA purification from insects using a morar and pestle and adding RNase cocktail (5 μl; Invitrogen) along with proteinase K. DNA concentration was measured using the Qubit dsDNA BR Assay Kit (ThermoFisher). To improve the assembly and gene annotation, we also extracted RNA from the four species. For Geopemphigus sp. and Stegophylla sp. samples, RNA was extracted using the RNeasy Plus Mini Kit (Qiagen) following the modifications described by Smith and Moran [75]. Genomic DNA was removed from RNA samples using the RQ1 RNase-free DNase Kit (Promega) and repurified using the RNeasy MiniElute Cleanup Kit (Qiagen). For P. obesinymphae and C. viminalis samples, RNA was extracted using the Quick-RNA MiniPrep Kit (Zymo Research) according to the manufacturers’ protocol. RNA concentration was measured using the Qubit ssRNA BR Assay Kit (ThermoFisher). Whole genome DNA (350 bp insert size) and poly-A enrichment mRNA libraries were prepared and sequenced on an Illumina NovaSeq 6000 instrument (2x150) by Novogene Corporation Inc. For Geopemphigus sp., we downloaded the already available genome sequencing data via NCBI under NCBI SRA accession number SRR6849201 [29]. For H. cornu, we manually translated the available nucleotide sequences of all protein-coding genes (S1 Table).

Our bioinformatics workflow is summarized in S3 Fig. First, we removed low quality sequences and sequencing adaptors from raw reads for both DNA and RNA sequencing data using Trimmomatic version 0.38 [78]. We first assembled the genomes with reads from DNA sequencing data using SPADES version 3.14.1 [79] with kmer sizes (-k) = 21, 33, 55, 77, 99, 127. Then we further improved the genome continuity by using the paired-end information from RNA-seq data. We mapped RNA-seq data to SPADES scaffolds using HISAT2 version 2.2.1 [80] and scaffolded the SPADES assemblies using P_RNA_scaffolder [81] with default settings.

To remove potential contamination and retrieve symbiont genomes, we first aligned scaffolds to NCBI NT database (last accessed January 27, 2021) using BLASTN megablast with “-culling_limit 5 -evalue 1e-25” in BLAST+ version 2.11.0 [82]. Then we mapped DNA sequencing data to the assemblies to estimate sequencing depth using bwa version 0.7.17 [83]. We created blobplots based on the BLAST results and BAM files using BlobTools version 1.1.1 [84]. Sequences that were annotated as human or bacterial were removed from downstream analysis. Sequences that were annotated as Buchnera, Bacteroidetes, Rhizobiales, Rickettsiales, or Pantoea were extracted as the symbiont genomes.

Lastly, as we detected potential contaminations from Pemphigus and ants in the Chaitophorus genome assembly (reflecting the origin of these samples from field sites where these organisms coexisted), we aligned Chaitophorus scaffolds to Pemphigus scaffolds using BLAST with e-value = 1e-10. Chaitophorus scaffolds were removed if 1) the scaffolds were ≥ 1000 bp and have at least 1000 bp matches with 95% identity to Pemphigus scaffolds, or 2) the scaffolds were < 1000bp and showed at least 95% identity with any length mapped to Pemphigus scaffolds. Scaffolds assigned as Hymenoptera by BlobTools were also removed from downstream analysis. We evaluated the completeness of genomes using BUSCO version 4.1.4 [38] with the Hemiptera gene set (n = 2,510).

For gene annotation, we first soft-masked the assemblies using Tantan [85] in Funannotate version 1.8.8 [86]. We used BRAKER2 version 2.1.5 [87,88] with both RNA evidence and protein evidence from closely related species (—etpmode). For RNA evidence, we mapped RNA-seq data to assemblies using HISAT2. For protein evidence, we downloaded all annotated proteins from aphid genomes on RefSeq and the gene annotation of E. lanigerum [33], as it is the closest species available. Lastly, we evaluated the quality of gene annotations using BUSCO Hemiptera gene sets.

Host phylogenetic analysis

To construct the host phylogeny, we first extracted the longest isoform for each annotated gene across 23 species. We assigned genes to orthologous groups using OrthoFinder version 2.5.2 [89] with default parameters. Orthologous groups exist in at least 80% of the species were used for phylogeny. For each orthologous group, we aligned amino acid sequences using MAFFT version 7.407 L-INS-i algorithm [90] and removed gappy regions using Gblocks version 0.91b [91] with default parameters. Lastly, we concatenated the resulting alignements and constructed phylogenetic trees using IQ-Tree version 2.0.6 [92] with 1,000 ultrafast bootstrap replicates. Bemisia tabaci, Diaphorina citri, and Pachypsylla venusta were used as outgroups.

Insect PGN pathway genes were identified using BLASTP against a custom database of insect PGN genes, including aphid HTGs. To search for bLys genes, the ch-type lysozyme domain of A. pisum bLys was used as a query in a TBLASTN search of all insect genomes. For each search with the exception of RlpA, hits were aligned using MUSCLE [93]. For RlpA, hits from the search of each A. pisum paralog were pooled, manually dereplicated, aligned with MUSCLE, and then manually adjusted based largely on the position of Cys residues in the ICK domains. Alignments were then trimmed using trimAl [94]. Gene trees were constructed for AmiD and RlpA protein sequence alignments using IQ-Tree 2 [92] with Transfer Bootstrap Estimates [95], with VTR4 and VTR5 models selected for amiD (S1 Fig) and rlpA (S2 Fig) gene families using ModelFinder [96]. Gene incidence was determined by counting the number of sequences per species in each alignment following manual inspection of alignment quality and sequence integrity.

We constructed a gain-loss-duplication model to explore the origin and fate of PGN pathway genes, including HTGs, lysozymes, and PGRPs, among aphids and close relatives. Using Count [97], we first performed rate optimizations for gene gain, loss, and duplication using flexible gain-loss and duplication-loss ratios for all lineages and a Poisson distribution at the tree root. Next, we used Dollo parsimony to determine the positions of gene gains and losses within the aphid phylogeny. Fig 2 incorporates the results of the reconstruction for each gene family with the exception of rlpA3- and rlpA1-like genes—since H. cornu branches at the base of Aphidoidea, parsimony maps the acquisition of these genes after the divergence of H. cornu from other aphids, but a phylogenetic analysis of RlpA protein sequences places RlpA3- and RlpA1-like paralogs as basal to RlpA2/5-like paralogs, suggesting that H. cornu lost the ancestral rlpA3- and rlpA1-like genes following the acquisition of rlpA2/5-like genes by gene duplication (S2 Fig).

Analyses of Buchnera genomes

Sixty-nine bacterial genomes, including 51 Buchnera, 11 S. symbiotica, Skilesia, and five outgroup genomes, were used to identify orthologous gene groups belonging to the PGN and related pathways (S11 Table). As the assembled Buchnera genome sequence of Sitobion miscanthi was not available, we obtained its genome sequencing data from NCBI SRA (accession number: SRX5767526) [98], cleaned reads using Trimmomatic and assembled the Buchnera genome using SPADES under the same parameters described above. All other genomes are available online (S11 Table). First, all genomes were annotated using DFAST version 1.2.13 [99]. Coding sequences containing internal stop codons or frameshifts, likely representing pseudogenes, were removed from annotation files using a custom bash script (available at https://github.com/lyy005/peptidoglycan_related_genes_in_basal_aphids). Second, orthologous groups of DFAST-annotated bacterial proteins were determined using OrthoFinder version 2.5.2 [89]. The inflation parameter was set at 10 to account for the high phylogenetic distance among selected bacterial species. Seven proteomes downloaded from EnsemblBacteria, including Bacteriodes fragilis YCH46 (ASM992v1), Escherichia coli MG1655 (ASM584v2), Buchnera from A. pisum (ASM960v1), B. pistaciae (ASM772v1), and Schizaphis graminum (ASM736v1), and S. symbiotica from A. pisum str. Tucson (ASM18648v1), were included to facilitate retrieval of the gene names represented by each orthogroup. A SmartTable containing all E. coli MG1655 gene names and synonyms was downloaded from EcoCyc.org [100] and a custom R script, written with R version 3.6.1 [101] and operated in RStudio version 1.1.463 [102], was used to 1) ensure uniform gene nomenclature between E. coli and Buchnera, 2) identify the incidence of 138 genes involved in PGN metabolism or related pathways (S8 Table), and 3) count the number of genes per orthogroup per species (available at https://github.com/lyy005/peptidoglycan_related_genes_in_basal_aphids). The heatmaps shown in Fig 3 were constructed with the pheatmap package [103] using the same R script.

Test of positive selection

To test whether certain aphid HTGs are under positive selection, we prepared codon alignments of each HTG protein family that match the protein alignments described above using PAL2NAL [104]. We then employed the BUSTED method (57) to test for positive selection through the Datamonkey web server [105].

Aphid rearing for symbiont cell wall labeling

Clonal lines of A. pisum, including the parental lines LSR1-c and Austin-c, previously cured of secondary symbionts (Austin-cured) [106], as well as Austin aphids uncured of secondary symbionts (Austin), were reared on broad bean (Vicia faba) at 20°C constant temperature with a 14L/10D daily light cycle. The clonal R. maidis line Rma-NM-1, collected from Sorghum halepense in Austin, Texas in 2017, was reared on barley (Hordeum vulgare) under the same temperature and light conditions as A. pisum. To incorporate d-alanine probes into Buchnera cell walls, aphids of either species were reared on leaves of their respective host plants immersed in a probe-agar solution. Individual leaves were immobilized in liquid 1.5% agar containing 35.4 mM of either EDA (d-proargylglycine; Alfa Aesar) or ELA probe (l-propargylglycine; Santa Cruz Biotechnology) within petri dishes (50 x 10 mm) and incubated for 24 hours. Two adult aphids (≥ 10 days old) were placed on each leaf and allowed to reproduce for 24 hours. Adults were removed and offspring reared for 3–4 days on the same leaves before transferring aphids to fresh probe-fed leaves and rearing for another 3–4 days. Seven-day-old aphids were used for all downstream experiments.

Purification of cell wall-labeled Buchnera

Seven-day-old A. pisum nymphs from six leaves per treatment were transferred to 1.5 ml microfuge tubes, surface-sterilized in bleach (500 μl; 0.5%) for 1 minute, rinsed twice in deionized water (500 μl), and homogenized in buffer A (300 μl; 25 mM KCl, 35 mM Tris, 10 mM MgCl₂, 250 mM sucrose, 250 mM EDTA, pH = 7.5) by mortar and pestle. Homogenates were passed through Sterile Swinnex filter holders (13 mm, EMD Millipore) loaded with UV-irradiated 100 μm nylon mesh filters (EMD Millipore) cut into 1/2 inch diameter circles using a leather hollow punch. Mortar tubes were rinsed with additional buffer A (300 μl) and the rinses filtered and pooled with the initial filtrates. Pooled filtrates were centrifuged for 20 min at 4000 x g, 25°C, and the pellets resuspended in buffer A and filtered again using a 5 μm nylon mesh filters. Filtrates were centrifuged the pellets washed with enzyme reaction buffer (300 μl; 0.1 M NaCl, 25 mM Tris, pH 7.5). Cells were centrifuged once more, the pellets resuspended in enzyme solution (50 μl; 1 μM enzyme, 0.1 M NaCl, 25 mM Tris, pH 7.5), and cell suspensions incubated at 37°C overnight in a thermocycler with the lid set to 60°C. Recombinant E. coli AmiD was recombinantly expressed (see below) while HEWL was purchased (EMD). Treated cells were centrifuged and washed twice with phosphate-buffered saline (PBS; 300 μl).

To fix and stain Buchnera cells, fixative (60 μl; 16% paraformaldehyde, 0.05% glutaraldehyde) was prepared and aliquoted to fresh tubes, to which sodium phosphate was added (12μl; 1 M, pH = 7.4) immediately followed by cell suspension (300 μl). The mixture was inverted twice and incubated at room temperature for 15 min, then on ice for 30 min. Fixed cells were centrifuged at 10,000 x g for 10 min, rinsed twice with PBS (300 μl), permeabilized in PBS + 0.5% Triton X-100 (300 μl) for five min, and rinsed twice more with PBS. Permeabilized cells were blocked with PBS containing 3% bovine serum albumin (BSA; 300 μl) at room temperature for 1 hour, then centrifuged. Cells were reacted with AF488 azide using the Click-iT Cell Reaction Buffer kit in 200 μl volumes of reaction mixture according to the manufacturer’s specifications (Invitrogen). After washing reacted cells with BSA, cells were suspended in PBS (300 μl) and counter-stained with WGA640R (25 μg/ml) and DAPI (10 μg/ml) for 15 min in darkness. Stained cells were washed once more with PBS and suspended in 10 μl PBS, then mounted onto glass microscope slides and coverslips sealed using clear nail polish. Microscopy images were acquired using a Nikon Eclipse TE2000-U inverted microscope equipped with a Hamamatsu ORCA-Flash4.0 V2 camera and Elements software (Nikon; version 4.50.00). Cells were visualized under brightfield (100 ms), GFP (λ_ex = 470 nm, λ_em = 535 nm, exposure time = 2 s), Cy5 (λ_ex = 640 nm, λ_em = 700 nm, exposure time = 500 ms), and DAPI (λ_ex = 395 nm, λ_em = 460 nm, exposure time = 2 s) channels. Fluorescence images were falsely colored green, magenta, and cyan for these channels, respectively. Several multi-channel images were recorded for each sample.

To quantify the effects of enzyme treatment of cellular fluorescence of each fluorescent probe, images were analyzed using ImageJ software [107]. For each image, regions of interest (ROIs) were established around individual cells visualized in the DAPI channel using the freehand selection tool. The area and mean fluorescence across each of the four channels was calculated for each ROI using the “Multi Measure” function within the ROI Manager tool and exported as csv files. Measurements for each image were uploaded to RStudio [102] and analyzed using custom R scripts written with tidyverse package functions [108]. The mean fluorescence intensity per area was calculated for the GFP, Cy5, and DAPI channels, and statistical comparisons made for each using the dunn.test package (S9 Table) [109], which performs a Kruskal-Wallis test followed by the Dunn test to extract significant pairwise differences (p-value < 0.01 after correction for flase-discovery rate). Plots were generated using ggplot2 [110] and compact letter displays added using the cldList function from the rcompanion package [111]. Custom scripts are available at GitHub (https://github.com/lyy005/peptidoglycan_related_genes_in_basal_aphids).

Cell-wall labeling and FISH microscopy in aphid embryos

FISH microscopy largely followed the methods of Koga et al. [112]. Embryos were dissected from several seven-day-old, probe-fed Austin A. pisum and R. maidis aphids in buffer A and temporarily stored in 70% ethanol before fixing in Carnoy’s solution (5 ml; ethanol:chloroform:acetic acid at a 6:3:1 ratio) for ≥ 30 min at room temperature. Fixed embryos were then washed once with 70% ethanol (5 ml) and thrice with 100% ethanol (5 ml) for 5 min per wash with gentle agitation, then bleached overnight in 1% H₂O₂ in ethanol. Bleached embryos were then washed thrice with 100% ethanol and stored at -20°C in ethanol. For hybridization with FISH probes, embryos were washed thrice with PBST (PBS + 0.2% Tween-20; 5 ml each), twice with hybridization buffer (20 mM Tris-HCl (pH 8.0) 0.9 M NaCl, 0.01% SDS, 30% v/v deionized formamide; 5 ml each), then incubated in hybridization buffer containing FISH probes (100 nM per Cy5-ApisP2A and Cy3-PASSisR 16S rRNA probe (S10 Table), 20 ng/ml DAPI; 3 ml) in darkness at room temperature overnight with gentle agitation. For the click reaction, hybridized embryos were washed twice with PBS + 3% BSA (1 ml), then blocked in PBS + 3% BSA (1 ml) for 60 min at room temperature with gentle agitation before replacing the solution with Click-iT reaction containing AF488 azide (500 μl) as described previously. Reacted embryos were washed twice with PBS (1 ml) and gentle agitation, then mounted onto glass depression slides with a drop of SlowFade Diamond Antifade Mountant (Molecular Probes) and fixed with coverslips sealed with clear nail polish. Images in Fig 5 were recorded on a Zeiss LSM 710 confocal microscope.

Construction of E. coli AmiD protein expression vector

The E. coli amiD gene was cloned into the pET-28b bacterial expression vector as an N-terminally His-tagged coding sequence identical to that used by Uehara et al. [53]. E. coli DH5α genomic DNA was purified using the DNeasy Blood and Tissue kit following the protocol for Gram-negative bacteria (Qiagen). PCR amplification of amiD was performed with Phusion DNA polymerase (NEB), primers 1 and 2 (S12 Table), 1:200 volumes of genomic DNA template per 10 μl reaction, and the following thermocycler program: an initial denaturation (98°C, 1 min) followed by 35 cycles of denaturation (98°C, 15s), annealing (61–66°C, 15 s), and elongation (72°C, 35s) and a final elongation step (72°C, 5 min). Amplified DNA was purified using the QIAquick PCR Purification kit (Qiagen), and digested with NdeI-HF and XhoI restriction enzymes (NEB). The pET-28b vector was digested with the same restriction enzymes and dephorphorylated using Antarctic Dephosphorylase (NEB). Digested amiD insert was ligated into vector by inclubating with T4 DNA ligase (NEB) at 4°C overnight. The ligation reaction was transformed into chemically competent DH5α cells, plated on LB agar plates with kanamycin (50 μg/ml), and grown at 37°C overnight. Colonies were screened for the expected insert size by colony PCR using Taq Polymerase (ThermoFisher) with primers 3 and 4 (S12 Table) and the following thermocycler program: an initial denaturation (94°C, 3 min) followed by 30 cycles of denaturation (94°C, 30s), annealing (46°C, 30s), and elongation (72°C, 80s), and a final elongation step (72°C, 7 min). Positive colonies were grown in LB liquid culture (10 ml) plus kanamycin (50 μg/ml) at 37°C, 220 rpm overnight and plasmids purified using the QIAprep Spin Miniprep kit (Qiagen). Purified plasmids were verified by Sanger sequencing using primers 3 and 4 (S12 Table).

Production and purification of E. coli AmiD enzyme

Chemically competent E. coli Rosetta (DE3) cells were transformed with 1 μl of purified plasmid, plated on LB agar plates with kanamycin (50 μg/ml) and chloramphenicol (25 μg/ml), and grown at 37°C overnight. For each strain, six colonies were used to inoculate six wells of a deep 24-well plate containing LB (6 ml) plus kanamycin (50 μg/ml) and chloramphenicol (25 μg/ml), and the plate shaken at 37°C, 220 rpm overnight. Wells were pooled and the mixture (5 ml) used to inoculate an LB culture (1L) containing kanamycin (50 μg/ml) and chloramphenicol (25 μg/ml) in an Erlenmyer flask (2.8 L). The culture was shaken at 37°C, 200 rpm until an OD₆₀₀ of 0.4–0.6 was reached, upon which the cultures were chilled to 18°C, induced with isopropyl β-D-1-thioglactopyranoside (IPTG, 1 mM), and shaken overnight at 18°C. Cells were harvested by centrifugation at 4000 rpm, 4°C for 30 min and the pellets frozen at -80°C until purification.

To purify proteins, cell pellets (from 500 ml culture) were thawed on ice and suspended in lysis buffer (35 ml; 0.5 M NaCl, 25 mM imidazole, 10% glycerol, pH 8.0, with freshly added 0.4 mg/ml HEWL, 1 mM N-methyl-p-toluenesulfonamide, 1 mM MgCl₂, and 10 μg/ml DNaseI). Cell suspensions were rocked at 4°C for 1 hour, vortexed, and homogenized using a EmulsiFlex-C3 high pressure homogenizer. Lysates were centrifuged at 13,000 rpm, 4°C for 30 min and the supernatants decanted onto Ni-NTA resin (2ml; Qiagen) pre-equilibrated with lysis buffer. The lysate-resin slurry was gently mixed and the flow-through eluted. The resin was washed twice with lysis buffer (25 ml), twice with elution buffer (25 ml; 0.1 M NaCl, 5 mM HEPES, 25 mM imidazole, 10% glycerol, pH 8.0), and once with each of the following elution buffers of increasing imidazole: 50 mM, 100 mM, 200 mM, and 500 mM imidazole (5 ml). Fractions were analyzed by SDS-PAGE (S4 Fig), and those containing proteins of interest were pooled, packed into 3500 MWCO SnakeSkin Dialysis Tubing (Thermo), and dialyzed at 4°C over the course of two days with four exchanges of dialysis buffer (1 L; 0.1 M NaCl, 25 mM Tris, 10% glycerol, pH 7.5, with 5 mM 2-mercaptoethanol added to the first two exchanges). Dialyzed proteins were concentrated using 10 kDa MWCO Amicon Ultra-15 centrifugal filters (EMD Millipore). Protein concentrations were determined from UV absorbance (λ = 280 nm) measured on a NanoDrop Lite spectrophotometer (Thermo) using fully reduced extinction coefficients calculated using the Expasy ProtParam online tool (Swiss Institute of Bioinformatics).