A Peptidoglycan Amidase Mutant of Burkholderia insecticola Adapts an L-form-like Shape in the Gut Symbiotic Organ of the Bean Bug Riptortus pedestris

Bacterial cell shapes may be altered by the cell cycle, nutrient availability, environmental stress, and interactions with other organisms. The bean bug Riptortus pedestris possesses a symbiotic bacterium, Burkholderia insecticola, in its midgut crypts. This symbiont is a typical rod-shaped bacterium under in vitro culture conditions, but changes to a spherical shape inside the gut symbiotic organ of the host insect, suggesting the induction of morphological alterations in B. insecticola by host factors. The present study revealed that a deletion mutant of a peptidoglycan amidase gene (amiC), showing a filamentous chain form in vitro, adapted a swollen L-form-like cell shape in midgut crypts. Spatiotemporal observations of the ΔamiC mutant in midgut crypts revealed the induction of swollen cells, particularly prior to the molting of insects. To elucidate the mechanisms underlying in vivo-specific morphological alterations, the symbiont was cultured under 13 different conditions and its cell shape was examined. Swollen cells, similar to symbiont cells in midgut crypts, were induced when the mutant was treated with fosfomycin, an inhibitor of peptidoglycan precursor biosynthesis. Collectively, these results strongly suggest that the Burkholderia symbiont in midgut crypts is under the control of the host insect via a cell wall-attacking agent.

Bacterial morphology is also influenced by internal factors, such as the bacterial growth phase, and by external factors, including nutrient availability, environmental and chemical stresses, physical constraints, predation, and life inside other organisms (Young, 2006;van Teeseling et al., 2017). Bacteria that colonize environments inside other organisms, such as endosymbionts or pathogens, sometimes adapt forms that markedly deviate from their regular shape in the free-living state. A well-known example of this cell shape alteration is in the nitrogen-fixing legume-rhizobium symbiosis when rhizobial symbionts colonize the symbiotic organ, the root nodule, developed by the host plant. Rhizobia are typical rod-shaped alpha-or beta-Proteobacteria in the free-living state in soil and under in vitro culture conditions (Oke and Long, 1999). The symbiotic bacteria that colonize the plant cells of the root nodule differentiate into a specialized nitrogen-fixing form called a "bacteroid". In some legumes, bacterial cell division is inhibited during the formation of bacteroids, whereas cell growth and genome replication continues, resulting in polyploid, extremely enlarged bacterial cells that may be elongated, branched, or spherical (Oke and Long, 1999;Mergaert et al., 2006;Czernic et al., 2015). Defensin-like membrane-attacking antimicrobial peptides called "Nodule-specific Cysteine Rich peptides (NCRs)" are strongly expressed in root nodules (Mergaert et al., 2003), and induce morphological and physiological alterations in rhizobium bacteroids (Mergaert, 2018;2020).
Article ME20107 Intracellular symbionts are also common in insects, which carry them in specific organs called bacteriomes composed of symbiont-infected cells or bacteriocytes (Baumann, 2005). Bacteriocyte-associated endosymbionts are generally unable to grow in a free-living state outside of their insect host, but have free-living close relatives in alpha-, beta-, and gamma-Proteobacteria or Bacteroidetes. These intracellular insect symbionts are often large and have markedly different morphologies from their rod-shaped free-living relatives. Similar to the rhizobium bacteroids in legumes, bacteriocyte symbionts may be spherical or elongated and sometimes enlarged to extreme sizes. Previous studies reported that these symbionts have very unusual, irregular, and pleomorphic shapes, forming balloon structures with invaginations or even rosette-like forms (Login et al., 2011;Shigenobu and Wilson, 2011;Hirota et al., 2017;Lukasik et al., 2017;Okude et al., 2017;Bublitz et al., 2019). These differentiated bacterial forms of insect and plant endosymbionts may adapt space-filling shapes to optimally occupy the greatest volume available in their host cells.
The bean bug Riptortus pedestris possesses a gut symbiotic bacterium, Burkholderia insecticola, in the cryptbearing posterior region of the midgut (the 4 th section of the midgut, M4), which harbors more than 100 million symbiont cells (Takeshita and Kikuchi, 2017;Kaltenpoth and Flórez, 2020;Ohbayashi et al., 2020). Eggs and hatchlings of R. pedestris are aposymbiotic (i.e. symbiont-free) and the insect acquires B. insecticola specifically from ambient soil during its development (Kikuchi et al., 2007;. A bacteria-sorting organ of the midgut, called the "constricted region", is crucial for the acquisition of symbionts because it separates the anterior digestive regions of the midgut from the posterior symbiotic compartment (Ohbayashi et al., 2015). The flagellar motility of B. insecticola (Ohbayashi et al., 2015) in addition to other as yet unknown molecular features (Itoh et al., 2019) are pivotal for passing through the constricted region and reaching the crypt-bearing symbiotic gut region. After colonization, morphological and physiological alterations occur in B. insecticola ; the cell shape becomes spherical, the DNA content per cell decreases, and the lipopolysaccharide (LPS) of the cell envelope is partially altered. At the transcriptomic level, several metabolic pathways, including assimilation pathways of host metabolic waste, such as sulfate and allantoin, are specifically up-regulated in vivo, indicating that B. insecticola proliferates in the bean bug midgut by recycling the metabolic waste of the host . However, the underlying mechanisms and molecules of the host midgut that induce morphological and physiological alterations in the symbiont currently remain unknown.
Bacterial N-acetylmuramyl-L-alanine amidase encoded by amiC is involved in septal peptidoglycan cleavage during cell division (Heidrich et al., 2001). An amiC (BRPE64_ACDS22630) deletion mutant of B. insecticola (ΔamiC) has a filamentous form composed of chains of unseparated cells because cell division is incomplete. Consequently, the mutant has less motility and infection ability (Lee et al., 2015). The chain form of the ΔamiC mutant is nutrient-dependent; although the mutant forms chains in nutrient-rich YG (yeast extract and glucose) medium, it forms separate cells in minimal medium and its motility and infection ability are restored (Lee et al., 2015).
In the present study, we revealed that morphological alterations were more prominent in the ΔamiC mutant inside the insect midgut, in which the mutant becomes spherical and enlarged, than under in vitro conditions. Furthermore, to clarify the mechanisms underlying morphological changes in the symbiont inside the midgut crypts, we examined the effects of nutrients, stress agents, and antibiotics on the cell morphologies of the symbiont, and found that the antibiotic fosfomycin mimicked the swollen shape in vitro.

Insects and bacterial strains
The bean bug R. pedestris TKS1 inbred line is derived from a pair of wild insects collected from a soybean field in Tsukuba, Ibaraki, Japan in 2007 and has been maintained in the laboratory for more than ten years. Insects were reared in a container at 25°C under a long-day regimen (16 h light, 8 h dark) and fed dry soybean seeds and a cotton pad containing distilled water with 0.05% ascorbic acid. The container was replaced twice a week. In infection experiments, newborn insects were placed in a Petri dish and fed as described above.

Oral administration of symbionts to the bean bug
Three days after hatching, water was removed from the plastic dish containing newly molted 2nd instar nymphs and they were kept overnight without water. This water deprivation stimulated the instant drinking of a bacterial suspension the next day, resulting in the efficient establishment of symbiotic infection. The B. insecticola wild type and ΔamiC mutant were pre-cultured in 3 mL MMGlc medium containing 30 μg mL -1 kanamycin at 27°C and 150 rpm in a rotary incubator; 200 μL of the overnight culture was inoculated into 3 mL MMGlc and incubated at 27°C and 150 rpm until the exponential growth phase. After the confirmation of bacterial motility by microscopic observations, bacterial density was adjusted to 10 7 cells mL -1 by measuring optical density, and the bacterial suspension was provided to insects as their drinking water. These insects were maintained until dissection and further analyses.

Quantitative PCR
To assess the number of B. insecticola symbiont cells colonizing M4 crypts, DNA extraction was performed from dissected M4 crypts infected with the wild type or ΔamiC mutant using the QIAmp DNA Mini kit (Qiagen). A 150-base pair fragment of the dnaA gene was amplified by real-time quantitative PCR using KAPA SYBR Fast qPCR polymerase (KAPA Biosystems) and the primer set BSdnaA-F (5′-AGC GCG AGA TCA GAC GGT CGT CGA T-3′) and BSdnaA-R (5′-TCC GGC AAG TCG CGC ACG CA-3′) (Kikuchi and Fukatsu, 2014). The PCR temperature profile was set to 95°C for 3 min, 40 cycles of 95°C for 3 s, 55°C for 20 s and 72°C for 15 s, and then 95°C for 5 s, 65°C for 1 min, and 97°C for 30 s using the LightCycler ® 480 Real-Time PCR System (Roche Life Science). The number of B. insecticola symbiont cells was calculated based on a standard curve for the dnaA gene with 10, 10 2 , 10 3 , 10 4 , 10 5 , 10 6 , and 10 7 copies per reaction of the target PCR fragment.

Epifluorescence microscopy observations
Insects were dissected in phosphate-buffered saline (PBS) buffer under a binocular (Leica, S8APO), and M4 crypts were collected. To observe in vivo bacterial cells, M4 crypts were transferred onto a glass microscopy slide with 5 μL of PBS buffer and a cover glass. Bacteria were released from the crypts by exerting pressure on the cover glass. They were observed by epifluorescence microscopy. In vitro cultured bacterial cells were similarly observed.

Confocal laser scanning microscopy observations
In spatiotemporal observations of M4-colonizing ΔamiC mutant cells, symbiotic insects at the early 3rd instar stage (a few hours after molting from the 2nd instar), 3rd instar middle stage (1~2 days after molting from the 2nd instar), and 3rd instar late stage (3~4 days after molting from the 2nd instar) were dissected and the posterior region of the midgut was collected. Tissue was stained by 3.75 μg mL -1 4,6′-diamidino-2-phenylindole (DAPI), washed with PBS, fixed by 1% paraformaldehyde, washed again with PBS, mounted in 20% glycerol, and then observed by confocal laser scanning microscopy using the Leica TCS SP8 instrument.

Quantitative analysis of bacterial cell shapes
Bacterial cells were randomly selected from microscopy images and cell shapes were evaluated using ImageJ software (ver.1.51) (Schneider et al., 2012). An individual cell was selected as a Region Of Interest (ROI) and the lengths of the minor axis and major axis were measured. The aspect ratio as an index of circularity was calculated by dividing the minor axis length by the major axis length. Cell shape was assessed by the aspect ratio shown in Fig. S1. Parameters were measured in dividing cells and chain-forming cells. The number of cells investigated under each condition is shown in each figure (Fig. 2, 4, and S3).

The ΔamiC mutant of B. insecticola forms cell chains in vitro and swollen cells in the bean bug midgut
The B. insecticola wild-type strain was rod-shaped when cultured in vitro (Fig. 1A), whereas in vivo symbiont cells became smaller and some were spherical, which is consistent with our previous findings ( Fig. 1C; Ohbayashi et al., 2019). In contrast, ΔamiC had a markedly different cell shape. The mutant formed long chains of unseparated cells in nutrient-rich YG medium due to incomplete cell division, as shown previously ( Fig. 1B; Lee et al., 2015), whereas in vivo ΔamiC cells became spherical and chain formation was less pronounced (Fig. 1D). Some spherical cells were very swollen (Fig. 1D).
To quantitatively confirm morphological alterations in the ΔamiC mutant, an image analysis of cell shape was performed by measuring the minor axis and major axis lengths of individual cells in epifluorescence microscopy images. The aspect ratio, calculated as the division of the minor axis length by the major axis length, indicates the cellular circulation level, and the lengths of the major axis and minor axis were used as a proxy to estimate bacterial cell size (Fig.  S1A). The aspect ratio varies between 0 and 1. A value close to 1 indicates a spherical cell shape, whereas strongly elongated cells have a value close to 0 (Fig. S1B). The aspect ratio distribution of the wild type and ΔamiC mutant grown in vitro were indistinguishable and had a normal distribution with a mean aspect ratio of 0.5 ( Fig. 2A and B).  However, the aspect ratio of the in vivo symbiont cells of both the wild type and mutant had a bi-modal distribution, which was different from the aspect ratios of cells grown in vitro ( Fig. 2C and D). Wild-type symbiont cells had an aspect ratio peak at 0.6 to 0.7, which was slightly higher than that when grown in culture, and a second peak at 0.9 to 1.0. However, the mutant had a similar first peak to bacteria grown in vitro and a second major peak at 0.9 to 1.0. The proportion of spherical bacterial cells with an aspect ratio larger than 0.8 was 44% for the in vivo ΔamiC mutant, whereas it was only 24% for the in vivo wild type ( Fig. 2C  and D), showing that the alteration from the rod shape to a spherical form was more pronounced in the mutant than in the wild type.
The distribution of the major and minor axis lengths of the wild type and ΔamiC mutant reflected cell size reductions in vivo, as shown previously (Fig. 2E, F, G, and H; Ohbayashi et al., 2019). The wild type had a mean major axis length of 2.6 μm in vitro and only 1.6 μm in vivo, whereas the mean major axis length of the ΔamiC mutant was 2.3 μm in vitro and 1.8 μm in vivo (Fig. 2E, F, G, and H). Moreover, cell size measurements revealed the presence of a small fraction of spherical in vivo ΔamiC mutant cells with axis lengths larger than 2.3 μm and up to 3.5 μm. These very enlarged swollen cells were absent in the wild type (Fig. 2E, F, G, and H). Collectively, these results confirmed microscopic observations (Fig. 1D) and suggested that in vivo bacterial cells had two populations with rod-shaped and spherical cells and that the majority of ΔamiC mutant in vivo cells became spherical or swollen, with some becoming very large. Despite this marked difference in cell morphology, an infection test and the quantification of occupancy by bacteria demonstrated that the ΔamiC mutant colonized M4 crypts, similar to the wild type, in all nymph stages (Fig.  S2).

Fosfomycin induces swollen cells in vitro in the B. insecticola ΔamiC mutant
Many bacteria change their cell morphology based on environmental conditions, such as nutrient availability and bacterial stress (van Teeseling et al., 2017). The mechanisms contributing to the production of swollen ΔamiC mutant cells in the midgut environment of the bean bug remain unclear. Therefore, we attempted to reproduce swollen cells in vitro. We investigated the cellular morphologies of the wild type and ΔamiC mutant cultured under 13 different conditions. These conditions included growth in minimal medium with five different carbon sources (glucose, fructose, mannitol, yeast extract, and maleic acid) and exposure to four cell membrane stresses (fosfomycin, lysozyme, polymyxin B, and SDS), oxidative stress (hydrogen peroxide), and three osmotic stresses (sucrose, glycerol, and NaCl). Among the 13 different culture conditions examined, only fosfomycin exposure induced spherical cells in the B. insecticola wild type and swollen cells in the ΔamiC mutant (Fig. 3). A quantitative analysis (Fig. 4) demonstrated that fosfomycin exposure resulted in the distribution of the cell shape aspect ratio with one sharp peak close at 0.9 to 1.0 in the wild type and ΔamiC mutant, similar to the major population of the in vivo ΔamiC mutant (Fig. 2D and 4C and D).
Moreover, in the fosfomycin-exposed ΔamiC mutant, 25% of spherical cells had axis lengths of 2.3 to 3.9 μm, similar to the largest swollen in vivo cells of the mutant (Fig. 2H  and 4H). Fosfomycin is an inhibitor of the enzyme UDP-Nacetylglucosamine enolpyruvyl transferase (MurA), which is involved in the biosynthesis of the peptidoglycan precursor UDP-N-acetylglucosamine enolpyruvate (Falagas et al., 2016), suggesting the induction of swollen ΔamiC mutant cells by the complete or partial lack of a cell wall. Consistent with this hypothesis, the observation of viable fosfomycin-treated ΔamiC mutant cells was only possible in the presence of 0.5 M osmoprotective sucrose, while the bacterium was uncultivable in non-osmoprotective minimal medium (data not shown).

Swollen cells of the ΔamiC mutant are induced in the bean bug midgut prior to the molting stage
B. insecticola symbiont cells actively proliferate in M4 crypts with slight fluctuations in symbiont numbers during host development due to an increase in antimicrobial activity in M4 crypts before molting (Kim et al., 2014;Ohbayashi et al., 2019). We examined spatiotemporal variations in the in vivo cell shape of the ΔamiC mutant to establish whether host physiology affects symbiont morphology during host development (Fig. 5A). The ΔamiC mutant formed chains in the whole midgut region at the early 3rd instar stage (Fig. 5B, C, and D); however, many spherical cells were observed in the middle and posterior regions of M4 crypts at the middle 3rd instar stage ( Fig. 5F and G). Most cells became spherical at all regions of M4 crypts at the late 3rd instar stage immediately prior to molting (Fig.  5H, I, and J), and some were very swollen in the anterior and middle regions (Fig. 5H, I, and J). These results suggested that host physiology before molting induced swollen cells in the B. insecticola ΔamiC mutant.

Discussion
The present results revealed that the cell shape of the ΔamiC mutant of B. insecticola became swollen in the midgut crypts of the bean bug (Fig. 1D), and among the various in vitro conditions examined, only fosfomycin exposure induced a similar swollen shape in the ΔamiC mutant (Fig.  3). Moreover, the wild type showed a similar, but milder phenotype ( Fig. 1C and D). These results indicate that the morphological alterations observed in the Burkholderia symbiont in the midgut crypts  were induced by a cell wall-targeting agent.
The in vitro mimicking of swollen cells by fosfomycin strongly suggests that the swollen shape of the ΔamiC mutant was due to the complete or partial lack of a cell wall. Bacteria did not grow in non-osmoprotective minimal medium with fosfomycin. Bacterial lysis may have occurred due to impaired cell wall integrity, resulting from the inhibition of peptidoglycan precursor biosynthesis by fosfomycin combined with the deletion of the amiC gene. In the presence of the osmoprotectant sucrose, lysis was prevented in the fosfomycin-treated ΔamiC mutant and cells became spherical and enlarged. This morphology resembles the shape of cell wall-deficient bacterial forms, called the "L- form" (Allan et al., 2009).
The "L-form" was initially described in Streptobacillus moniliformis ("L" was named in honor of the Lister Institute where the bacterial form was discovered) (Klieneberger, 1935). This pleomorphic bacterium switches its morphology from chains to swollen cells in vitro (Dienes, 1939). L-form bacteria were subsequently shown to be induced by the suppression of cell wall integrity with cell wall-inhibiting agents, such as penicillin and lysozyme (Madoff, 1986). Previous studies demonstrated that the L-form may be induced under osmoprotective conditions by lytic enzymes and/or inhibitors of cell wall biosynthesis, such as lysozyme, penicillin G, ampicillin, and fosfomycin, in diverse bacterial species, including Gram-positive and -negative bacteria (Allan et al., 2009;Errington et al., 2016). In the case of B. insecticola, only fosfomycin induced L-form-like swollen cells (Fig. 3). Fosfomycin targets MurA, which performs the first step in peptidoglycan synthesis (Egan et al., 2020). Peptidoglycan formation is complex, consisting of many biosynthesis and regulatory steps (Table S1). It is thus possible that other molecules inhibiting steps downstream of MurA may cause similar phenotypes to fosfomycin. It currently remains unclear why other lytic enzymes and/or inhibitors of the cell wall, including lysozyme, polymyxin B, and SDS, did not induce the swollen cell shape (Fig. 3). We speculated that Burkholderia species have multiple defense mechanisms against surface-attacking antimicrobial agents, which may maintain the cell shape of the ΔamiC mutant under membrane stress conditions. For example, the modified LPS lipid A structure with positively  charged 4-amino-4-deoxy-arabinose (Ara4N), membraneintegrated hopanoids, and several efflux pump mechanisms contribute to this cell envelope stability (Rhodes and Schweizer, 2016). C-type lysozymes, similar to that used in our assays, are known to hydrolyze the peptidoglycan of Gram-positive and -negative bacteria. However, since peptidoglycan in Gram-negative bacteria is shielded by the outer membrane, the enzyme may not have had access to the peptidoglycan substrate of B. insecticola in our assay. The symbiotic organ of R. pedesteris expresses genes encoding ctype lysozymes as well as a bacterial type lysozyme (Futahashi et al., 2013). These enzymes may contribute to cell morphology modifications in the symbiotic organ, but may require the assistance of other factors in the gut that permeabilize the outer membrane. The absence of these factors may then explain the lack of a response by B. insecticola in our lysozyme assay. On the other hand, B. insecticola may also have a resistance mechanism against lysozyme activity as an adaptation for the colonization of lysozyme-containing midgut crypts.
Spatiotemporal observations of the ΔamiC mutant in midgut crypts revealed that symbiont cells became swollen, particularly before molting (Fig. 5). Kim et al. (2014) reported that a temporary decrease in the symbiont population occurred before molting, when the expression of genes encoding antimicrobial proteins and peptides, such as c-type lysozyme and riptocin, was up-regulated. In addition, these antimicrobial peptides produced in M4 crypts showed mark-edly stronger antimicrobial activity against in vivo symbiont cells than in vitro grown cells Ohbayashi et al., 2019). The cell envelope structure of B. insecticola is altered in vivo, as revealed by the lack of the O-antigen polysaccharide on LPS and by the formation of blebs on the bacterial membrane Ohbayashi et al., 2019). These stresses attacking the cell surface together with the deletion of the amiC gene may induce the swollen cells in M4 crypts before molting.
Although the present results suggest that a peptidoglycantargeting antimicrobial agent is secreted in M4 crypts that induces the swollen cell shape, the nature of this agent remains unclear. In the case of legume-rhizobium symbiosis, NCR peptides in nodules block cell division in rhizobia, which induces cell elongation in bacteroids (Mergaert, 2018(Mergaert, , 2020. Strikingly, a rhizobium mutant in a gene encoding a peptidoglycan-modifying DD-carboxypeptidase enzyme forms bacteroids that have a strongly enlarged and spherical shape, similar to the crypt bacteria produced by the ΔamiC mutant described here. This mutant only forms these abnormally swollen bacteroids in nodules that produce NCR peptides, suggesting that these peptides are the direct cause of the bacterial growth defect of the mutant (Gully et al., 2016). Bacteriocytes in the pea aphid produce bacteriocyte-specific cysteine-rich (BCR) peptides (Shigenobu and Stern, 2013). Although their in vivo functions currently remain unknown, some BCR peptides exhibited in vitro antimicrobial activity, induced cell elongation, and increased membrane permeability and DNA content in E. coli, which is phylogenetically close to Buchnera symbionts (Uchi et al., 2019). These BCR peptides may affect the swollen and spherical morphology of Buchnera cells in bacteriocytes. In the bean bug, more than 90 species of crypt-specific cysteine-rich peptides (CCRs) are secreted in M4 crypts (Futahashi et al., 2013), some of which exhibit antimicrobial activity in vivo against the cells of B. insecticola . Although there is no sequence similarity between NCR, BCR, and CCR peptides, these peptides are all small and rich in cysteine, resembling the typical antimicrobial peptides of innate immunity, such as defensins (Mergaert et al., 2003;Futahashi et al., 2013;Shigenobu and Stern, 2013;Montiel et al., 2017). This type of cysteine-rich peptide may be the result of convergent evolution in the different symbiotic systems in which they have similar functions. Future studies are needed to establish whether CCRs are involved in the morphological alterations of the ΔamiC mutant in midgut crypts by exposing the ΔamiC mutant to CCR peptides in vitro or by analyzing the morphology of the ΔamiC mutant in crypts after the suppression of CCR expression by reverse genetics in the host insect.
In conclusion, a detailed analysis of the symbiotic phenotype of the ΔamiC mutant enabled us to highlight the existence of a novel and partially characterized molecular mechanism of symbiont control by R. pedestris that may be widespread in host-microbe symbioses.