The Bacterial Metabolite Indole Inhibits Regeneration of the Planarian Flatworm Dugesia japonica.

Planarian flatworms have been used for over a century as models for regeneration. Planarians live in aquatic environments with constant exposure to microbes, but the mechanisms by which bacteria may mediate planarian regeneration are largely unknown. We characterized the microbiome of laboratory populations of the planarian Dugesia japonica and determined how individual bacteria impact D. japonica regeneration. Eight to ten taxa in the phyla Bacteroidetes and Proteobacteria consistently occur across planarian colonies housed in different research laboratories. Individual members of the D. japonica microbiome can delay regeneration including the development of eye spots and blastema formation. The microbial metabolite indole is produced in significant quantities by two bacteria that are consistently found in the D. japonica microbiome and contributes to delays in regeneration. Collectively, these results provide a baseline understanding of the bacteria associated with the planarian D. japonica and demonstrate how metabolite production by host-associated microbes can affect regeneration.

Introduction

The past decade of host microbiome research has rapidly revealed the diversity of microbial communities that live in and on eukaryotic hosts and their potential roles in mediating host biology (McFall-Ngai et al., 2013). Much of this research has focused on contributions of microbes to host metabolism, physiology, and immunology. In a variety of animals, microbial communities have been shown to aid in the provision of nutrients (Sannino et al., 2018, Wong et al., 2014), alter immune system function (Brestoff and Artis, 2013, Lee and Mazmanian, 2010), and shape host behavior (Sampson and Mazmanian, 2015, Vuong et al., 2017).

One aspect of host biology wherein contributions of the microbiome are poorly characterized is regeneration. Regeneration, which is defined as the ability to replace and repair body parts that are severely damaged (e.g., amputation), is a unique feature of certain metazoans (Sánchez Alvarado, 2000, Sánchez Alvarado and Tsonis, 2006). The ability to regenerate tissues or organs spans the kingdom Animalia, with certain species able to completely regenerate entire organ systems and body parts (i.e., planarian worms and Hydra), others possessing the capability to replace entire limbs or organs (i.e., zebrafish and axolotls), and others with limited regenerative abilities (i.e., humans and mice) (Bely and Nyberg, 2010, Brockes and Kumar, 2008, Holstein et al., 2003, Sánchez Alvarado, 2000, Sánchez Alvarado and Tsonis, 2006). Regeneration in multicellular organisms requires the detection of trauma, activation of repair processes, production and differentiation of new cells, and coordinated arrangement of tissues into organ systems (Brockes and Kumar, 2005, Tanaka and Reddien, 2011). All these events require a complex interplay of multiple biophysical and biochemical processes, each of which could potentially interface with bacteria-derived signals.

Host-associated microbes could potentially mediate the outcomes of regeneration through several different processes. First, pathogenic microbes could infect regenerating wound sites and directly inhibit regenerating tissues by killing host cells. Previous work on the microbiology of wounds has documented how microbes can influence wound-healing outcomes (Edwards and Harding, 2004, Eming et al., 2014), but studies on how microbes can impact whole organ or limb regeneration are limited. Second, non-pathogenic microbes could promote or inhibit regeneration by secreting metabolites at wound sites that alter the cellular processes necessary for normal patterning in host tissues. Finally, microbes could promote and accelerate regeneration by inhibiting the growth of microbial species that inhibit regeneration, by secreting metabolites that promote host regeneration, or by inducing host cellular processes that promote tissue repair. Numerous studies have demonstrated that microbes can secrete metabolites that affect host biology (Koppel and Balskus, 2016, Postler and Ghosh, 2017, Sharon et al., 2014), but we are unaware of studies that have considered how microbial metabolites can influence regeneration processes.

Over the past century, planarian flatworms have been used as model organisms to understand mechanisms of metazoan regeneration. Planarian worms harbor specific populations of fully pluripotent cells (neoblasts) that, upon trauma, become highly active and proliferate at the site of injury to form a cluster of new undifferentiated cells called a blastema (Durant et al., 2016, Lobo et al., 2012, Reddien and Sánchez Alvarado, 2004). The blastema cells then differentiate and serve as the basic cellular units for building and replacing tissues, organs, and body parts. The entire process of regeneration is temporally regulated by the expression of a variety of host genes (Durant et al., 2016, Lobo et al., 2012, Reddien and Sánchez Alvarado, 2004). Planarian worms are capable of regenerating entire individuals from small body fragments (Morgan, 1898). These worms live in natural and laboratory aquatic ecosystems wherein wound sites are exposed to microbes.

Despite the importance of planarians as models for regeneration, the microbiome of planarians and impacts of microbes on planarian regeneration are poorly characterized. A recent study characterized the microbiome of one planarian species (Schmidtea mediterranea) and demonstrated that one bacterial member of the S. mediterranea microbiome, a Pseudomonas species, can elicit a pathogenic response in the host that inhibits regeneration (Arnold et al., 2016). These findings provided the first evidence that microbes can mediate planarian regeneration, but they did not identify mechanisms by which bacteria alter the regenerative process in planarians. Furthermore, it is also unclear whether the previously described microbiome diversity is conserved across other planarian species.

In this study, we describe the diversity of the microbes associated with laboratory populations of the planarian worm Dugesia japonica and demonstrate that individual bacteria within the D. japonica microbiome can mediate regeneration outcomes. D. japonica is a close relative to S. mediterranea and is widely used in studies of regeneration (Álvarez-Presas et al., 2008, Gentile et al., 2011), yet very little is known about its microbiome (Morokuma et al., 2017). By using 16S rRNA sequencing of DNA and RNA, culture-based approaches, and whole-genome sequencing, we characterized the diversity of the D. japonica microbiome across time and across multiple laboratories. By manipulating the D. japonica microbiome through inoculation experiments, we also assessed the roles of individual bacterial species in mediating regeneration outcomes. We identified one common microbiome member, a non-pathogenic Aquitalea species (class Betaproteobacteria), that substantially delays regeneration after tail and head amputation. Production of the metabolite indole by Aquitalea can also delay regeneration, providing a mechanistic link between the planarian microbiome and the regeneration processes.

Results

D. japonica Worms Have a Consistent Bacterial Community

To determine the bacterial composition of laboratory-reared D. japonica, we used high-throughput amplicon sequencing of the 16S rRNA gene. We obtained an initial profile of the bacterial diversity of our worms by sampling individual worms at three times over the course of a year (2016–2017). On average, a single worm harbors 42 distinct bacterial operational taxonomic units (OTUs) using a threshold of 97% sequence similarity. Across all worms, eight OTUs accounted for approximately 80% of the bacterial community, with Bacteroidetes and Proteobacteria dominating the communities (Figure 1A, Table S1). On average, Bacteriodetes taxa (i.e., Taibaiella sp. OTU003 and Pedobacter sp. OTU001) accounted for 36% of the sequence data, while Proteobacteria (i.e., Paucibacter sp. OTU011, Rhodoferax sp. OTU019, Comamonadaceae sp. OTU002, Burkholderiales sp. OTU005, and Aquitalea sp. OTU004) accounted for 40%. Across the three sampling times, these taxa were consistently present, but their relative abundance varied across individuals (Figure 1A).

Figure 1

Microbiome Diversity of the Planarian Dugesia japonica

(A) Bacterial community composition of three individual worms was determined at three time points using 16S rRNA gene amplicon sequencing. OTUs >1% relative abundance are shown. See Table S1 for full list of OTUs. OTUs highlighted in bold have cultured representatives.

(B) Spots of bacterial cultures on brain heart infusion agar from some of the most abundant and culturable bacteria in the D. japonica microbiome.

(C) 16S rRNA transcript mining of the D. japonica microbiome. Total RNA extracted from pooled sets of worms were synthesized into cDNA libraries and sequenced with the Illumina HiSeq platform, and the reads were mapped to a custom 16S rRNA database (see Methods). Reference sequences with ≥1.5% of mapped reads are shown. Each column represents a biological replicate of five worms. See Table S4 for the full list of reference sequences.

(D) History of the D. japonica strain GI across four laboratories.

(E) D. japonica microbiome diversity across four laboratories. The first column on the left is the average of Tufts data from (A). Each column represents a single worm.

See also Figures S1 and S2 and Tables S1, S2, S3, S4, and S5.

In an effort to culture bacterial strains representative of taxa observed in our 16S rRNA survey, worm homogenates were plated on several types of media and conditions (see Methods), and unique bacterial strains were isolated based on colony morphology (Figure 1B). Four isolates were in the phyla Bacteroidetes (Chryseobacterium sp. KBW03., Pedobacter sp. KBW01PK, and Pedobacter sp. KBW06S, and Taibaiella sp. KBW10), five were β-Proteobacteria species (Variovorax sp. KBW07, Acidovorax sp. FJL06, Oxalobacteraceae sp. KBW02, Paucibacter sp. KBW04, and Aquitalea sp. FJL05), and one was a γ-Proteobacteria (Pseudomonas sp. KBW05) (Figure S1). Collectively, we were able to culture bacterial strains that represent five of the seven OTUs that make up over 75% of the diversity in our 16S rRNA survey (Table S1). To confirm the identity of isolates and acquire additional information on functional potential, draft genomes were assembled and annotated for each of these dominant bacterial taxa (Table S2).

To identify potentially active bacteria associated with D. japonica, we extracted total RNA from three pools of five worms and sequenced cDNA libraries (Tables S3 and S4). Across the three libraries, 11 taxa dominated transcription in the worm microbiome, with Bacteroidetes and β-Proteobacteria accounting for 36% and 64%, respectively, of 16S rRNA transcripts for taxa above a 1.5% relative abundance threshold. Many of the bacteria identified in our 16S rRNA amplicon sequencing and culturing efforts were also identified as active in the worm microbiome (Figure 1C). For example, of the three Bacteroidetes species identified in our analysis, Taibaiella sp. was the most transcriptionally active. These results are recapitulated in our 16S rRNA survey wherein Taibaiella is typically observed as a highly abundant species across worm samples.

D. japonica planaria are used across many laboratories as models for regeneration. To determine if the bacterial communities associated with D. japonica are consistent across different laboratory populations, we sampled the microbiome of the same clonal population of D. japonica shared across three additional laboratories. The D. japonica worms in our laboratory at Tufts originated from the Umesono laboratory in Japan (Hyogo University) and were later used to seed and start colonies in subsequent laboratories (Marchant lab at the University of Minnesota; Oviedo lab at the University of California, Merced; Figure 1D). These worms are all of the same genetic background and are known as the “GI strain” or sometimes “Dj-GI.” Worm care practices differ between laboratories (Table S5), and we predicted that this might impact the microbiome composition of D. japonica across the different worm colonies. A total of 10 OTUs from the phylum Bacteroidetes and Proteobacteria dominated the D. japonica microbiome across laboratories (Figure 1E). Nine of these OTUs were also the dominant taxa in our Tufts University worm colony. Across the four laboratories, some taxa ranged widely in relative abundance (Figure S2; Table S1). For example, Aquitalea sp. OTU004, was observed across all laboratory-reared worms, with a high abundance of Aquitalea sp. (56% relative abundance) in the University of Minnesota worms and a low abundance in other laboratories (<1%–10% relative abundance).

Spatial Variation and Stability of the Dugesia japonica Microbiome

To better understand what factors shape the diversity of the D. japonica microbiome, we assessed the spatial structure and stability of bacterial communities associated with D. japonica. For all of this work, we used a culture-based method to determine bacterial community composition because we found that much of the D. japonica microbiome can be cultured and because we are ultimately interested in the dynamics of bacteria that can be easily manipulated in regeneration experiments. We acknowledge that some rare or difficult-to-culture taxa that were observed in 16S rRNA sequencing efforts may be missed with this approach.

Many planarian regeneration experiments use head, trunk, and tail fragments in regeneration assays (Lander and Petersen, 2016, Lobo and Levin, 2015, Petersen and Reddien, 2008, Reddien and Sánchez Alvarado, 2004). To determine whether these various body fragments are colonized by distinct bacterial communities, we divided individual worms (n = 6) into these three fragments (Figure 2A) and plated worm homogenates onto brain heart infusion agar to determine the bacterial community composition. The composition of bacterial communities across the head, trunk, and tail fragments was highly variable across individual worms (Figure 2A), but we did not detect statistically significant differences in bacterial community composition across these three regions (PERMANOVA, F = 2.16, p = 0.08). None of the bacterial taxa were significantly enriched in any of the body fragments. This suggests limited spatial structure at the coarse scale of whole worm bodies and a well-mixed bacterial community across D. japonica.

Figure 2

Spatial Variation and Stability of the Dugesia japonica Microbiome

(A) Bacterial community composition across the head, trunk, and tail fragments of individual worms. Data show relative abundance of culture-based (plating on brain heart infusion agar) bacterial community composition.

(B) Relative abundance of D. japonica-associated bacteria on the external surfaces and within internal components of individual worms. Each point represents the relative abundance for each worm sampled. *Int, bacterium enriched in internal fraction; *Ext, bacterium enriched in external fraction.

(C) Experimental assessment of the ability of exogenous microbes to colonize the D. japonica microbiome. Detected, taxa detected at >1% of the total bacterial community using culture-based approach (plating on brain heart infusion agar); not detected, <1% the total bacterial community.

Planarian worms have distinct external and internal features that may create unique environments for specific bacterial communities. For example, worms have a distinct gastrovascular tract inside their bodies and secrete significant quantities of mucus on their external surfaces (Hayes, 2017, Pedersen, 1963). To better understand whether bacterial taxa are unevenly distributed across external and internal components of the worm, we used a fractionation approach. Externally associated bacteria were detected after gently washing worms in 1X phosphate buffered saline, and internally associated bacteria were detected in the remaining worm homogenate after washing (see Methods for details). The gastrovascular tract cannot be easily dissected from planarian worms, so we were unable to specifically assess the gut microbiome of the worms. The relative abundance of Chryseobacterium sp. was higher in the external fraction than in the internal fraction (external = 65.7% ± 17% relative abundance, internal = 24.7% ± 19% relative abundance) (Figure 2B). Two other bacteria, Polaromonas sp. and Pedobacter sp. KBW01PK, were not detected in the external fraction and were exclusively found in the internal fraction. No significant differences between the external and internal fractions were observed for the other major bacterial taxa isolated from these worms (Pseudomonas sp., Herminiimonas sp., Variovorax sp., Pedobacter sp. KBW06S, and Taibaiella sp.).

Our survey of D. japonica across laboratories from a common origin demonstrated that some bacterial taxa were unique to specific worm colonies. This suggests the potential for D. japonica worms to pick up and/or lose bacterial species in response to changing environments. To experimentally assess plasticity of the D. japonica microbiome, we tested whether exogenous microbes that are not normally associated with D. japonica can establish within the D. japonica microbial community. We placed worms in an inoculum of exogenous microbes that were previously isolated from various aquatic systems, including a laboratory population of axolotls (Ambystoma mexicanum), a laboratory population of the South African clawed frog (Xenopus laevis), and a freshwater pond in a garden in Boston. We also tested a common laboratory strain of E. coli. After removing the worms from this inoculum and placing them in sterile water, we destructively sampled worms at different time points over the course of 15 days (Figure 2C).

Some exogenous microbes were detected at 1 and 4 days after being removed from the inoculum, whereas at 15 days we no longer detected any of the exogenous microbes (Figure 2C). These results indicate that D. japonica does not readily associate with environmental bacteria and that exogenous microbes may have difficulty colonizing an established microbiome of D. japonica. This observation is consistent with previous reports that D. japonica is not readily colonized after exposure to pathogenic bacteria (Abnave et al., 2014). Some endogenous microbes, including Chryseobacterium and Taibaiella, were not detected in worms wherein exogenous microbes were added. They may have been present in very low abundances but were below the detection limit of our culture-based method or may have been displaced by the addition of exogenous microbes. This work did not specifically address how exogenous microbes may affect the endogenous bacterial community of D. japonica, but the role of microbe-microbe interactions in shaping the D. japonica microbiome is a promising area of future investigation. We also acknowledge that these experiments have only tested just a few easy-to-isolate bacteria. The unique bacterial species detected in D. japonica microbiomes from different laboratory populations may be present due to rare invasion events of bacterial taxa not examined here.

Bacteria Associated with D. japonica can Delay Regeneration

Our results described above provide a baseline understanding of the microbes that inhabit this planarian species and their dynamics over space and time in laboratory cultures. To determine whether any of the identified bacteria impact D. japonica regeneration, we conducted a regeneration screen wherein the head and tail of fully regenerated worms were aseptically amputated and the remaining trunk fragment submerged into pure cultures of bacteria (Figure 3A). We were unable to generate completely germ-free planarian worms due to difficulties in maintaining worms in antibiotics for long periods of time (Miller et al., 1955). We instead used a short-term antibiotic treatment and bacterial inoculation approach to manipulate the D. japonica microbiome. Before the screen, fully regenerated worms were subjected to an antibiotic cocktail, which significantly depletes the bacterial load without affecting host regeneration (Figure S3). The bacterial densities used in our screen (Table S6) are similar to what were used in a previous study of regeneration in a different planarian species (Arnold et al., 2016). Over the course of two weeks, regenerative progress was monitored and compared to that of fragments regenerating in sterile water.

Figure 3

A Regeneration Screen of the Dugesia japonica Microbiome Demonstrates Bacterial-Induced Delays in Regeneration

(A) Overview of the experimental design of the regeneration screen.

(B) Representative images of the regenerative progress of trunk fragments treated in bacterial inocula or sterile worm water over the course of 12 days. Worm images were cropped to remove background shadows using the Magic Eraser Tool in Adobe Photoshop. Uncropped images from other bacterial treatments can be found in Figure S4.

(C) Frequency of trunk fragments with eye spots in different bacterial treatments. Many treatments reached 100% at day 4, so they are not visible (are under the Water line). Inset shows examples of worms with and without eyes 4 days post-amputation. Error bars represent ± standard deviation.

(D) Length-to-width ratios of worms at 7 and 12 days post-amputation (DPA). Red dots represent data for individual worms, and boxplot shows mean and standard deviation. Welch's t test determined statistical difference in length-to-width ratios when comparing water controls and bacterial treatments at 7 DPA (F(10, 45) = 2.68, p =.011) and 12 DPA (F(10, 42) = 12.0, p < 0.001). Bonferroni corrected post-hoc analysis identified bacterial treatments statistically different from the respective water control at 7 DPA (lowercase letters indicate statistical differences; p < 0.005) and 12 DPA (uppercase letters indicate statistical differences; p < 0.001). Error bars represent ± standard deviation.

See also Figures S3 and S4 and Table S6.

Bacterial treatments displayed a broad range of effects on host regeneration (Figure 3B). For example, Acidovorax sp. FJL06, Oxalobacteraceae sp. KBW02, Taibaiella sp. KBW10, and Variovorax sp. KBW07 had no major effect on D. japonica regeneration (Figure 3B). All other bacterial inocula caused delays in host eye development. For example, certain bacterial treatments (Aquitalea sp. FJL05, Chryseobacterium sp. KBW03, Pedobacter sp. KBW06S, and Pseudomonas sp. KBW05) caused a delay in eye development for a proportion of individuals in a population, with all regenerating fragments acquiring eyes 9 days post-amputation (Figure 3B). A similar observation was made for Pedobacter sp. KBW01PK and Paucibacter sp. KBW04, except that a subset of individuals subjected to these bacterial inocula remained without eye spots at the conclusion of the study (Figure 3B).

To better understand how bacterial treatments impacted eye spot development, we quantified the presence of eye spots on the anterior end (i.e., head) of regenerating fragments from the regeneration screen described above. Up to three days post-amputation, eye spots were not visible for any treatment condition, including the water control (Figures 3B, 3C, and S4). By 5 days post-amputation, eye spots were observed for all or at least a proportion of regenerating fragments in all treatment conditions (Figure 3C). The most striking difference in eye development was observed at 4 days post-amputation, wherein the average number of worm fragments with eyes was significantly lower, compared with the control treatment, for the following inocula: Aquitalea sp. FJL05, Pedobacter sp. KBW01PK, Chryseobacterium sp. KBW03, and Paucibacter sp. KBW04 (Kruskal-Wallis test with Mann-Whitney U post-hoc tests, p < 0.001).

In addition to quantifying eye development trends, we also tracked the overall shape of the regenerating fragments at 7 and 12 days post-amputation. During typical regeneration of trunk fragments, sites of amputated tissue contract, initiating healing and remodeling processes (Beane et al., 2013, Oviedo et al., 2003, Reddien and Sánchez Alvarado, 2004, Tettamanti et al., 2008). Over the course of a few days, a cluster of self-renewing cells form a blastema, which governs the process of regenerating new tissues that will eventually become a new head and tail, on the respective anterior and posterior end of the trunk fragment (Aboobaker, 2011, Agata and Watanabe, 1999). By 7 days post-amputation, fragments have constructed a new head and tail that continue to develop, while the entire planarian body undergoes remodeling of tissues throughout the entire organism, with the ultimate goal of acquiring morphological body plan that is identical to a fully regenerated worm (Beane et al., 2013). Fragments in early stages of regeneration are shorter in body length and wider in body width, whereas fully regenerated worms have a target morphology that is longer and slimmer. Therefore, body length-to-width ratio is a useful metric to track the normal process of tissue remodeling in planarians (Emmons-Bell et al., 2015, Scimone et al., 2017).

Our initial analysis of body dimensions focused on comparing the length-to-width ratio of control and bacteria-treated worm fragments at 7 days post-amputation. This time point was selected because it is the earliest time point at which nearly all regenerating fragments in our study had eyes and regenerated head and tail structures. Aquitalea sp. FJL05, Pseudomonas sp. KBW05, and Pedobacter sp. KBW01PK inocula significantly altered the length-to-width ratio, producing regenerating worm fragments with a lower ratio (i.e., worm fragments possessed a shorter and wider morphology), indicating slower remodeling of host tissues (Figure 3D). This is consistent with the observed delay in eye development for worm fragments treated in Aquitalea sp. FJL05 and Pedobacter sp. KBW01PK inocula (Figure 3C). Conversely, no statistically significant effect on host length-to-width ratio was observed for Chryseobacterium sp. KBW03 or Paucibacter sp. KBW04 treatments, even though we observed delays in eye development (Figure 3C).

Analysis of length-to-width ratios at 12 days post-amputation revealed shifts from the 7 days post-amputation dataset with Pedobacter sp. KBW01PK, Pedobacter sp. KBW06S, Pseudomonas sp. KBW05, Taibaiella sp. KBW10, and Variovorax sp. KBW07 displaying statistically lower ratios than worm fragments that regenerated in water (Figure 3D). Out of the five bacterial inocula identified to negatively impact the length-to-width ratio of regenerating fragments at 12 days post-amputation, Pedobacter sp. KBW01PK and Pseudomonas sp. KBW05 were observed to cause delays in eye development and/or remodeling of worms 7 days post-amputation (Figures 3B, 3C, and S4). Conversely, Pedobacter sp. KBW06S, Taibaiella sp. KBW10, and Variovorax sp. KBW07 were only observed to delay remodeling of regenerating worms at 12 days post-amputation. Although the reported delays in host regeneration could be partially due to bacterial infection, there were no direct indications that worms harbored a systemic infection. In all treatment conditions, regenerating fragments appeared healthy with no observed lesions in tissues, a common sign of infection (Arnold et al., 2016).

Indole Production by Aquitalea Delays Regeneration of D. japonica

Our results mentioned above demonstrate that bacterial taxa in the D. japonica microbiome can delay regeneration. We next sought to identify potential molecular mechanisms underlying this bacterial control of host regeneration. During our efforts to culture microbes from the D. japonica microbiome, we noticed that certain bacterial inocula exuded strong fecal and barnyard odors that were reminiscent of indole-producing bacteria from other microbial systems studied in our laboratory (Kastman et al., 2016, Wolfe et al., 2014, Wolfe and Dutton, 2015). This olfactory observation piqued our interest because indole has been shown to affect the growth of eukaryotic hosts (Lee et al., 2015), and recent studies of human gut epithelial cells have demonstrated that indole can alter the activity of ion channels (Chimerel et al., 2012), key cellular components that play roles in planarian regeneration and patterning (Beane et al., 2013, Beane et al., 2011, Nogi et al., 2009). Given the potential for indole to impact planarian regeneration, we determined if bacteria in the D. japonica microbiome could produce indole and whether bacteria-produced indole can help explain the observed delays in D. japonica regeneration.

Bacteria produce indole through degradation of tryptophan via the enzyme tryptophanase (TnaA) (Lee and Lee, 2010, Snell, 1975). To confirm whether bacteria endogenous to the worm microbiome are capable of producing indole, we queried annotations of draft genomes and used BLASTp searches with the E. coli tnaA sequence for tnaA homologs. Genomes of Aquitalea sp. FJL05 and Chryseobacterium sp. KBW03 contained tnaA homologs with 44%–49% amino acid identity to the reference E. coli tnaA gene (Figure S5). To confirm that these bacteria produce indole, each isolate was grown in tryptophan broth and indole concentrations in broth supernatants were measured using the hydroxylamine-based indole assay, a colorimetric assay with high specificity for indole (Darkoh et al., 2015). After 2 days of incubation, Aquitalea sp. FJL05 and Chryseobacterium sp. KBW03 produced on average 200 and 40 μM of indole, respectively (Figure 4A). Escherichia coli strain BW25113, a bacterium known to produce indole (Lee and Lee, 2010), was used as a positive control and produced on average 550 μM indole in tryptophan broth. An isogenic ΔtnaA strain from the Keio collection (JW3686-7) was used as a negative control and did not produce detectable levels of indole. In addition, two bacterial species endogenous to the worm microbiome but lacking the tnaA gene (Oxalobacteraceae sp. KBW02 and Variovorax sp. KBW07) did not produce indole.

Figure 4

Indole as a Mechanism Underlying Bacterial Delays in Regeneration

(A) Production of indole in tryptophan broth by bacteria isolated from the Dugesia japonica microbiome and by control (BW25311) and tnaA mutant (JW3686-7) E. coli strains. Indole was quantified colorimetrically using the HIA assay. Bars represent means and error bars represent ± standard deviation; trp ase+ means the tnaA gene was detected in the genome of the indicated bacterium; bdl, below detection limit of HIA assay.

(B) Impacts of pure indole on D. japonica regeneration from 1–7 days post-amputation (DPA). White scale bar for each worm, 0.44 mm. Red rectangles indicate formation of a blastema at the anterior and posterior ends of the regenerating trunk fragment. Red arrow indicates eye spots. Blue arrow indicates lack of blastema formation.

(C and D) Indole production in (C) water control and in (D) Aquitalea inoculum with and without supplemental tryptophan. Lines represent means and error bars represent ± standard deviation.

(E) Regeneration phenotypes observed in water control and Aquitalea inoculum with and without supplemental tryptophan. Data for Oxalobacteraceae sp. KBW02 and E. coli strains can be found in Figure S6. White scale bar for each worm, 0.44 mm.

See also Figure S5.

To determine whether indole can inhibit worm regeneration, trunk fragments were submerged in sterile water with or without 100 μM pure indole. This concentration of indole was used because higher concentrations were toxic, lower concentrations exhibited weak phenotypes, and 100 μM was within the range produced by bacteria in this system when cultured in a tryptophan broth. Control worm fragments developed visible blastemas by 3 days post-amputation, whereas worms exposed to indole failed to develop typical blastemas at either wound site (Figure 4B). By 5 days post-amputation, eye spots were visible at the anterior end of control worms, whereas worm fragments regenerating in indole still lacked an anterior blastema (and, consequently, lacked eye spots). We removed worm fragments from the indole treatment after 7 days, and they were able to fully regenerate over the course of 14 days in plain water. These results suggest that indole can significantly delay D. japonica regeneration without any obvious signs of tissue degeneration or necrosis, similar to the delay that is observed with live bacteria (Aquitalea sp. FJL05 and Chryseobacterium sp. KBW03).

We next tested whether bacteria-derived indole could also inhibit worm regeneration. Using the same worm regeneration screen approach described above, we tracked the regenerative progress of trunk fragments in the presence of a bacterial inoculum, both with and without the addition of tryptophan, as exogenous tryptophan concentrations control the production of indole in bacteria containing the tnaA gene (Li and Young, 2013). As a negative control for the addition of both bacteria and tryptophan, trunk fragments were monitored for regeneration in sterile water with and without tryptophan. As a negative bacterial control, we utilized two bacteria that do not produce indole: Oxalobacteraceae sp. KBW02 and the ΔtnaA E. coli strain JW3686-7 (Figure 4A). Because Aquitalea sp. FJL05 was observed to be robust in the production of indole and strongly affected worm regeneration, all downstream analyses utilize Aquitalea sp. FJL05 inoculum. We used a lower concentration of bacterial inoculum in these experiments (OD600 of 0.3 instead of 2.7 used in the initial bacterial screen above) because preliminary experiments indicated that the addition of tryptophan with higher inoculum densities produced toxic levels of indole killing all worm fragments within 2 days.

Regeneration was significantly delayed in Aquitalea sp. FJL05 inoculum supplemented with tryptophan, compared with both the water controls and the Aquitalea inoculum alone. By 4 days post-amputation, trunk fragments treated in Aquitalea + tryptophan lacked developing tissue near sites of amputation and thus did not develop eye spots, as opposed to worms treated with water or with the Aquitalea inoculum alone (Figures 4C–4E). In this experiment, all worms treated with the Aquitalea inoculum developed eye spots by 4 days post-amputation, which is in contrast to what we observed in the experiment described in Figure 3B. This seemingly contradictory result can be attributed to the lower concentration of Aquitalea used in tryptophan supplementation experiments. Trunk fragments regenerating in the Aquitalea + tryptophan treatment were able to successfully regenerate once transferred from the inoculum to sterile worm water (Figure 4E). Trunk fragments regenerating in the E. coli strain BW25113-positive control inoculum died, likely because the high concentrations of indole produced with the supplemental tryptophan were toxic. The observed regenerative delay reported for the Aquitalea + tryptophan inoculum was not observed in the negative bacterial controls (i.e., Oxalobacteraceae sp. KBW02 or ΔtnaA E. coli JW3686-7), where, regardless of tryptophan concentrations, indole concentrations were below the level of detection (Figure S6).

Discussion

Despite the long use of planaria as models for regeneration, the microbiomes of these animals are poorly characterized. In this study, we describe the diversity of bacteria associated with D. japonica and demonstrate that individual bacterial isolates from the D. japonica microbiome can mediate regeneration outcomes. High-throughput sequencing and culturing demonstrated that laboratory-reared D. japonica worms host a set of bacteria that are consistently observed across individuals and over time, although the relative abundance of individual taxa can vary across individuals. The tractability of the planarian microbiome was demonstrated by our ability to culture many of the dominant bacteria in our 16S rRNA gene amplicon survey. The composition of an established planarian microbiome shows a degree of stability and resistance, as transient exogenous bacteria are unable to colonize and persist after introduction. The consistency of the D. japonica microbiome and resistance to invasion suggests that there are host- or bacteria-derived factors that determine the bacterial community associated with planarians. Indeed, previous studies have identified host immune genes and pathways that eliminate pathogenic bacteria, preventing them from colonizing planarians (Abnave et al., 2014, Pang et al., 2017, Tsoumtsa et al., 2017).

Across laboratories, the microbiome composition of laboratory-reared D. japonica from the same genetic background were generally similar, with enrichment of certain taxa driving differences between samples. For example, Aquitalea sp. was observed in nearly all samples across universities, but was especially high in worms from the University of Minnesota. Colony maintenance practices and rearing conditions (i.e., food and water sources, colony temperature, etc.) differ between laboratories, likely contributing to the observed difference in microbiome composition of laboratory-reared worms. Work in other animal systems has demonstrated that host diet, antibiotic use, and temperature can alter the composition of host-associated microbial communities (Ramsby et al., 2018, Spor et al., 2011, Reese and Dunn, 2018). Several taxa identified in this study (e.g., Rhodoferax sp., Pedobacter sp., Chryseobacterium sp., Pseudomonas sp., Acidovorax sp., and Oxalobacteraceae sp.) have previously been reported in the S. mediterranea microbiome, a close relative of D. japonica (Arnold et al., 2016). These observations suggest that certain bacterial taxa are shared across flatworms in the Dugesiidae family.

Previous work describing microbial impacts on metazoan tissue regeneration and remodeling has largely focused on just a few microbial taxa. A recent study demonstrated that three bacterial species endogenous to S. mediterranea (e.g., Chryseobacterium sp., Pseudomonas sp., and Vogesella sp.) can impede host regeneration (Arnold et al., 2016). In the study, the authors identify specific immune genes in the host (e.g., TAK1/MMK/p38) that are sufficient for eliciting regenerative responses in the presence of Pseudomonas sp. Our work builds on this previous study of planarian regeneration by screening a larger diversity of bacterial isolates on a different planarian species. We identified bacteria endogenous to the planarian microbiome that can alter regeneration dynamics, ranging from no effect to significant delays in host regeneration. For example, Acidovorax sp. FJL06 and Oxalobacteraceae sp. KBW02 showed no effect, whereas other bacterial species (e.g., Aquitalea sp. FJL05, Pedobacter sp. KBW01PK, Pedobacter sp. KBW06S, Psuedomonas sp. KBW05, Pacibacter sp. KBW04, Variovorax sp. KBW07, Chryseobacterium sp. KBW03, and Taibaiella sp. KBW10) were observed to delay and/or alter the host morphology at certain stages of regeneration. These results not only identify specific bacteria of the planarian microbiome that alter host regenerative processes but also demonstrate that bacteria can affect different hallmark features of planarian regeneration (i.e., eye spot development and tissue remodeling). The precise molecular mechanisms that govern atypical regenerative outcomes in the presence of certain bacteria are unclear, but future work will utilize the findings reported in this study to elucidate microbe-host interactions during host regeneration.

Our experiments focusing on indole provide a potential mechanism for microbial mediation of planarian regeneration. Two bacterial species endogenous to D. japonica are capable of producing indole (Aquitalea sp. FJL05 and Chryseobacterium sp. KBW03). Worms regenerating in pure indole suffer substantial inhibition of tissue regeneration. This observation is recapitulated with worms regenerating in an Aquitalea sp. inoculum supplemented with tryptophan. In the absence of an indole-producing bacterium or the metabolite tryptophan, worm regeneration proceeds normally. Multiple studies have provided evidence that indole and indole derivatives inhibit cell proliferation in various human cells lines (Frydoonfar et al., 2002, Ping et al., 2011, Telang et al., 1997, Wu et al., 2005). Our work suggests that the established antiproliferative properties of indole and indole derivatives extend to planarian cells, resulting in the inhibition of tissue regeneration that we observed in worms that were exposed to indole. Collectively, our results suggest that bacteria-derived indole can inhibit planarian tissue regeneration, an observation that supports the hypothesis that bacteria play a role in planarian regeneration beyond infection.

Limitations of the Study

We acknowledge that the bacterial enrichment approach that we used in this study has limitations in terms of relevance to natural planarian populations and bacterial community complexity. The bacterial densities used in our study and in Arnold et al. (2016) are higher than what worms may experience in natural conditions. However, the high-density screens are a useful first approach to discover bacteria that have the potential to shape outcomes of regeneration and the possible mechanisms underlying these host-microbe interactions. These coarse-scale observations serve as platform for future fine-scale analyses at the molecular level. The high-density enrichment approach also provides opportunities to develop endogenous bacteria to deliver specific metabolites or genetic constructs in a manner similar to how E. coli is used for RNAi in planarians (Newmark et al., 2003). Our enrichment approach used single species of bacteria so that we could enrich for individual bacterial species in each treatment. Combinations of bacterial species that make up the D. japonica may produce different metabolites compared with when grown alone, and future work using multispecies inocula may identify additional bacterial metabolites that mediate regeneration.

Another question that remains to be fully answered is whether planarians experience high enough levels of indole to impact regeneration in either a laboratory or the natural environment. We induced high indole concentrations (>200 μM) in the presence of Aquitalea by providing supplemental tryptophan in the worm water. Much higher indole concentrations have been detected in other microbial systems such as the human gut (Darkoh et al., 2015), but it is not known what concentrations planarians naturally experience. Many planarian worms are fed calf or chicken liver when maintained in the laboratory. Liver contains significant amounts of tryptophan (Kuiken et al., 1947) and may provide sufficient levels to stimulate indole production. We are unaware of studies that have quantified indole concentrations in natural aquatic habitats where planarian worms are found.

High concentrations of indole may not be necessary to impact regenerating tissue as localized and lower concentrations of indole may affect regenerative processes at the cellular level. We are currently investigating whether lower concentrations of indole can be detected within planarian tissue to better understand the fine-scale spatial dynamics of indole production in this system. Future studies aim to link these metabolite profiles with the spatial distribution of Aquitalea and localized molecular processes that may impact planarian regeneration.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.