Noninvasive gut microbiota sampling via feces collection

We first developed an electroporation protocol to routinely transform DNA into S. alvi (see Methods section), as a faster and simpler technique than the conjugal transfer typically employed for this species [28]. We were able to electroporate the previously described plasmid pBTK570 to wild-type cells [28], resulting in spectinomycin-resistant S. alvi. Next, we fed microbiota-depleted honey bees with an inoculum of engineered cells and allowed for colonization of the gut. We then established a noninvasive protocol to repeatedly collect honey bee feces and analyze its bacterial composition over time (Fig 1). By applying a gentle pressure onto the abdomen of CO₂-stunned bees, a substantial amount of fecal matter could be retrieved (Fig 1A). The volume of feces averaged 4.4 ± 2.4 μl per bee (n = 18), with a 90% success rate of extraction (Fig 1B). To confirm that the collected feces allow recovery of engineered bacteria and that they are an informative proxy for the honey bee gut bacterial content, we looked for the presence of the engineered strain of S. alvi in the feces and guts (i.e., midgut, ileum, and rectum) of inoculated bees (Fig 1C).

Download:

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

Fig 1. A novel method for noninvasive sampling of the honey bee gut.

(a) Photograph of a bee feces extraction. (b) Box plot indicates the median volume of extracted feces from n = 18 bees. (c) Dot plot shows bacterial load of engineered S. alvi found in the feces and in the gut homogenates. Dotted lines between data points indicate matching samples (i.e., feces and gut were sourced from the same bee). Number of bees sampled (n) and the rate of colonization (r) for the experiment are indicated. (d) Feces collection does not significantly decrease bees fitness. Kaplan–Meier graph shows honeybee’s probability of survival over time when subjected to fecal extractions (FE-treated; solid line) compared to control bees (Control; dotted line). Times of feces collection are indicated by arrows. Log-rank test, not significant (ns) at p-value = 0.112 for n = 44 bees. The data underlying this Figure can be found in the S1 Data file, sheets “Fig 1B,” “Fig 1C,” and “Fig 1D”.

https://doi.org/10.1371/journal.pbio.3002523.g001

For this, feces were collected a week postinoculation, directly followed by dissection of the gut tissues of those same individuals. The concentration of engineered S. alvi present in the feces and gut samples was estimated by isolating bacterial colonies onto selective media. Interestingly, genetically modified bacteria were found in the feces, with an average bacterial load of 1.3 10⁻⁵ colony-forming unit (CFU) per μl of feces (Fig 1C). It shows that gut symbionts commonly colonizing the proximal gut can nonetheless be abundantly found in fecal matter. Furthermore, we found a positive correlation between the amount of S. alvi in the feces versus the gut (Pearson correlation coefficient R = 0.3963, p-value < 0.05; S1 Fig). Not only were bacterial loads correlated, but absence of bacteria in feces indicated noncolonized bees (i.e., no bacteria in gut tissues; Fig 1C). Altogether, our data suggest that honey bee feces are a reliable indicator of the level of bacterial gut colonization.

Lastly, we checked that the feces extraction procedure established could be performed without being detrimental to the bees’ health. Results showed that weekly feces collection did not significantly affect the probability of survival of manipulated honey bees compared to the nontreated individuals (Fig 1D). Being now able to noninvasively probe the gut content of honey bees via fecal extraction allows for their longitudinal study as well as for the use of bacterial biosensors.

Functional broad-host range plasmids for honey bee symbionts

To expand the range of molecular tools available to genetically manipulate honey bee gut bacteria, we constructed broad-host range plasmids that can stably replicate in S. alvi. While the RSF1010 replicon had previously been shown to propagate in honey bee gut symbionts [28], we identified in the literature other candidate broad-host range replicons that could also potentially do so, namely, RK2 [34,35], pBBR1 [36,37], pVS1 [38,39], and pTF-FC2 [40,41] (Table 1). They were selected based on the range of species in which they were shown to function and their small sizes. Each origin of replication was cloned in standardized vectors carrying distinct combinations of fluorescent proteins and antibiotic resistance markers (i.e., E2-crimson/specR, GFP/ampR, or E2-crimson/ampR), resulting in a collection of 11 broad-host range plasmids (Figs 2 and S2).

Download:

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

Fig 2. Collection of broad-host range plasmids developed in this study.

(a) Schematic showing the different available combinations of origin of replications and standardized fragments bearing the antibiotic marker (ampR, specR) and fluorescent reporter (crim, gfp). (b) The names of the corresponding plasmids are indicated. Detailed plasmid maps can be found in S2 Fig.

https://doi.org/10.1371/journal.pbio.3002523.g002

Download:

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

Table 1. Main features of the broad-host range replicons tested in S.alvi and B. apis.

https://doi.org/10.1371/journal.pbio.3002523.t001

Upon electroporation of the plasmids in S. alvi, we were able to obtain stable cell lines from all replicons tested, except for RK2, confirming their ability to replicate in this species (Table 1). Additionally, we considered whether our broad-host range plasmids could be employed for other members of the honey bee gut community. As preliminary evidence, we transformed the vectors in Bartonella apis, a proteobacterium from a different class than S. alvi [27], and found that RK2, pBBR1, and pTF-FC2 stably replicate in B. apis (Table 1).

We further characterized each replicon, alongside RSF1010, by estimating their copy numbers, protein expression levels and co-compatibility in S. alvi (Fig 3). Data collected by quantitative PCR (qPCR; S3 Fig) in S. alvi showed that the vectors propagate to varying copy numbers, ranging from medium copy (approximately 23 copies per cell for RSF1010) to high copy number (approximately 906 copies per cell for pBBR1; Fig 3A). The fluorescence of individual engineered cells carrying the different replicons was also assessed by flow cytometry (Fig 3B). Except for S. alvi bearing the RSF1010 origin of replication, which intriguingly exhibited the highest fluorescent signal while harboring in principle the lowest amount of plasmid, cells carrying the other replicons produced levels of fluorescence consistent with the determined plasmid copy number (Fig 3A and 3B). These differences translated into a moderate dynamic range, with up to a 4-fold increase in fluorescence between the highest and lowest average green fluorescent protein (GFP) signals (i.e., pTF-FC2 and RSF1010; Fig 3B). What is more, by cotransforming in pairs the different plasmids in S. alvi, we observed that the pTF-FC2 replicon was compatible with all other origins of replication tested and allowed stable dual vector propagation (Fig 3C). The same analyses were also carried out in B. apis and led to largely similar results (S4 Fig). Taken together, characterization of the plasmid collection illustrated its versatility, as it offers a variety of co-compatible expression vectors with distinct properties, thus allowing distribution of genetic circuits and fine-tuning of their output.

Download:

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

Fig 3. Characterization of functional broad-host range replicons in the honey bee gut symbiont S. alvi.

(a) Broad-host range plasmids have different copy numbers in S. alvi. Box plots show median values of plasmid copy numbers obtained by qPCR from 3 independent experiments with 5 biological replicates each (total n = 15). Median copy numbers are indicated above the corresponding box plots. (b) The difference in plasmid copy number results in different protein expression levels in S. alvi. Graph shows mean of GFP fluorescence ± standard deviations of 5 biological replicates. Each replicate represents the average fluorescence of at least 9,000 cells measured by flow cytometry. Plasmids used for panels a and b in S. alvi were pAC08, pAC14, pAC13, and pAC12, carrying the RSF1010, pTF-FC2, pVS1, and pBBR1 origins of replication, respectively. (c) Some replicons are compatible and can be cotransformed in S. alvi. Matrix table indicates compatible (green boxes with check mark) and incompatible (red boxes with cross mark) replicons. Vectors were found compatible upon their successful cotransformation by electroporation in S. alvi. The data underlying this Figure can be found in the S1 Data file, sheets “Fig 3A” and “Fig 3B”.

https://doi.org/10.1371/journal.pbio.3002523.g003

Gut colonization stability over time of engineered S. alvi

Having expanded the set of plasmids to engineer honey bee gut symbionts, we validated the maintenance of such plasmids in vitro, as well as in vivo when engineered cells are in the presence of the natural gut microbial community, and also in absence of antibiotics that may disrupt the natural microbiota (Fig 4). For this, we chose to test the set of vectors providing spectinomycin resistance (crimson/specR), as they prevent the growth of nonresistant satellite colonies typically obtained using ampicillin selection upon CFU estimation. We first performed in vitro experiments where S. alvi bearing the different corresponding broad-host range vectors were grown in liquid cultures with or without antibiotic selection. Upon reaching stationary phase, cells from both conditions were isolated onto nonselective solid media, allowing bacteria with and without plasmid to grow, and the ratio of fluorescently labeled bacterial colonies was measured as a proxy for plasmid maintenance (Fig 4A). Samples across conditions were also plated onto selective media and all bacteria exhibited fluorescence, confirming that observed unlabeled cells growing on nonselective media were the result of plasmid loss rather than mutations that would lead to the loss or inactivation of the fluorophore. Ultimately, the pBRR1, RSF1010, and pTF-FC2 replicons were found stably maintained within the experimental timeframe, as the average percentage of cells carrying the vectors when incubated without selection pressure remained at approximately 89%. On the contrary, the propagation of pVS1 appeared remarkably weak, with less than 3% of the bacterial population still bearing the plasmid at the end of the experiment, and thus even for the control condition comprising cells grown with antibiotics. We hypothesize that this was due to an artifact of the experimental setup where plasmid stability is so low that although all single cells incubated with antibiotics carry the vector when spotted onto nonselective media, the resulting macrocolonies show no fluorescence as bacteria rapidly lose the plasmid throughout cell division. Low maintenance values reported for the “+Ab” condition (i.e., <80%) are thus not accurate in absolute terms but remain an indicator of very feeble plasmid propagation.

Download:

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

Fig 4. Maintenance of broad-host range plasmids and engineered S. alvi over time.

Each broad-host-range vector has a distinct level of maintenance in S. alvi cells, as demonstrated in vitro (a) and in vivo (c). Box plots represent plasmid maintenance median values of n biological replicates. Plasmid maintenance was estimated by measuring the percentage of fluorescent colony forming units (CFUs) plated onto nonselective media of samples obtained from (a) liquid cultures (in vitro) after 3 days of growth with (+Ab) or without (−Ab) spectinomycin supplementation; or (c) from feces (in vivo) collected 7 (orange) and 14 (red) days after gut monocolonization of bees continuously fed sugar water supplemented with (+Ab) or without (−Ab) spectinomycin. Only samples for which we detected bacterial colonies upon plating the fecal material were considered. (b) Schematic outline of the in vivo experiment to estimate engineered S. alvi colonization maintenance. Arrows indicate the 2 times of feces sampling. (d) Bacterial load of engineered S. alvi cells found in feces collected 7 (orange) and 14 (red) days after gut monocolonization of bees continuously fed sugar water supplemented with (+Ab) or without (−Ab) spectinomycin. Concentrations of our different S. alvi strains isolated from the feces of bees co-colonized with the natural gut community and that were fed sugar water without spectinomycin (+Gc) was also analyzed. Those CFUs values were obtained by plating diluted feces onto selective media (i.e., supplemented with spectinomycin). Bees for which we did not detect engineered bacteria were considered noncolonized individuals. The rate of colonized bees (r) and the number of bees sampled (n) are provided. Corresponding median values of colonized bees only are represented by colored horizontal bars with interquartile ranges. Plasmids used for all panels in S. alvi were pAC10, pAC04, pBTK570, and pAC11, carrying the pVS1, pBBR1, RSF1010, and pTF-FC2 origins of replication, respectively. The data underlying this Figure can be found in the S1 Data file, sheets “Fig 4A,” “Fig 4C,” and “Fig 4D”.

https://doi.org/10.1371/journal.pbio.3002523.g004

Additionally, we characterized plasmid maintenance in vivo by monocolonizing 4 groups of microbiota-depleted honey bees with S. alvi bearing the different broad-host range plasmids. Each group of bees was divided into 2 subgroups that were fed either normal (−Ab) or antibiotic-supplemented (+Ab) sugar water. We then carried out a longitudinal analysis of their microbiota by weekly feces sampling over 14 days (Fig 4B). Plasmid maintenance for each replicon was calculated similarly to what had been done for the in vitro samples, using cells isolated from the collected fecal matter (Fig 4C). Consistent with the in vitro data, the pVS1-based vector propagated poorly in vivo, as the entire S. alvi population recovered from nontreated bees was no longer carrying the plasmid a week postinoculation. Interestingly, at the scale of the average life span of caged honey bees (i.e., approximately 3 weeks) [42], the other replicons displayed distinct stability profiles ranging from short to long term, even without the presence of antibiotics. For instance, the pBBR1-based plasmid was lost quickly, with a median of 28% of the bacterial population bearing the plasmid after a week, before dramatically decreasing to 0% after 14 days. The 2 abovementioned vectors that are rapidly lost have the highest copy numbers in S. alvi (Fig 3A), which might indicate that the metabolic burden resulting from their replication significantly impacts their maintenance within cells. Contrastingly, the lower-copy RSF1010- and pTF-FC2-based vectors showed robust propagation over time, with a respective median of 55% and 86% of the bacterial populations still carrying the plasmids after a 2-week period without selection pressure. The pTF-FC2 plasmid is known to encode for a maintenance system, which might participate in its steady propagation in our cells (see Discussion).

Finally, due to the potential fitness cost linked to the replication of heterologous genetic material, we wondered whether utilization of the broad-host range plasmids would affect the ability of the engineered cells to colonize and persist in the honey bee gut over time in the presence of the natural microbiota. To answer this, we coinoculated microbiota-depleted honey bees with modified S. alvi strains harboring the different vectors and the natural gut community (+Gc), which included wild-type S. alvi (Fig 4B). Feces were sampled from these colonized bees and analyzed together with those collected for the measurement of plasmid stability. Fecal samples from the 3 treatment groups (i.e., +Ab, −Ab, and +Gc) were ultimately examined by isolating engineered S. alvi onto selective solid media, allowing us to determine the levels of gut colonization through time (Fig 4D). In the case of S. alvi transformed with the pVS1-based plasmid, they were detected consistently over 14 days at a concentration of approximately 10⁴ cells per μl of feces but solely from bees that were fed antibiotic. This further suggests that low plasmid maintenance values for the “+Ab” condition in Fig 4A are indeed due to an artifact of the experimental setup and that maintaining the pVS1 origin of replication might be detrimental to cells’ fitness and that stable colonization of the gut requires the addition of selection pressure. On the other hand, S. alvi bearing the pBBR1, RSF1010, and pTF-FC2 replicons were able to establish persisting communities that were not constrained to the presence of antibiotics, with the detection of approximately 10³ to 10⁵ engineered cells per μl of feces across the experiment. Bacterial concentrations and success rates of colonization found for each strain were also in line with the varying degrees of plasmid maintenance established in this study. S. alvi carrying the pTF-FC2-based plasmid, for instance, reached median values of 2.7 10⁴ cells per μl of feces after 2 weeks, while the less frequently maintained vector pBBR1 resulted in fecal bacterial populations 18 times smaller. These observations were also consistent with the rates of nonantibiotic-treated honey bees successfully colonized by each strain, as on average 50%, 61%, and 76% of bees were colonized by S. alvi transformed with the pBBR1, RSF1010, and pTF-FC2 replicons, respectively.

More importantly, the S. alvi strains genetically modified with the broad-host range vectors, except for pVS1, were able to colonize the honey bee gut environment upon competition with the native gut community, albeit to different degrees of success (Fig 4D). Cells bearing the RSF1010 and pTF-FC2 plasmids, for example, were both detected at concentrations of approximately 10⁴ cells per μl of feces across sampling times, but they differed in their average colonization rates, which were 52% and 75%, respectively. Nevertheless, the metabolic burden imposed by the broad-host range plasmids allowed engineered cells to settle as part of the bee microbiota without being outcompeted by their wild-type counterparts. This is of relevance as it confirms their potential to be utilized in situ for biosensing of the gut environment and its microbial communities.

Reprogrammed S. alvi cells act as in situ biosensors for IPTG

Engineering of bacterial biosensors relies on promoters that are specifically induced upon exposure to molecules of interest and that subsequently drive the expression of dedicated reporter genes. To date, the only functional inducible system reported for members of the honey bee gut microbiota is based on the previously described pBTK552 vector, which carries a fluorescent protein whose expression is conditional to the presence of the widely used sugar derivative IPTG [28]. The efficacy of induction was, however, solely demonstrated in vitro. Additionally, upon its transformation in S. alvi, we were unable to maintain integrity of the plasmid in our experimental conditions, as the system showed high genetic instability resulting in frequent deletion of the gfp gene (S5 Fig). Consequently, we built several variants of an IPTG biosensor using our tool kit in different gene arrangements and obtained 3 stable genetic constructs. Two were based on a dual-plasmid system whereby the circuit was distributed between 2 compatible and stably maintained broad-host range plasmids (i.e., RSF1010 and pTF-FC2; S6A Fig). Following the characterization of the different constructs, we selected the pAC17V5 circuit to pursue our work, as it showed the best dynamic range in response to IPTG (S6B Fig). Briefly, the pAC17V5 system consists of a first pTF-FC2-based plasmid bearing the lacI gene constitutively expressed (pAC17V5b) and a second RSF1010-based vector carrying gfp under the control of a constitutive promoter with a Lac operator (lacO) sequence directly downstream (pAC17V5a; Fig 5A). When transformed together in S. alvi, the plasmids enabled cells to respond in vitro to IPTG with a 7-fold increase in GFP expression (Fig 5B). Additionally, we wondered if other species of the honey bee gut microbiota could also be engineered to respond to IPTG. We tested our 3 genetically stable constructs in the 2 γ-Proteobacteria Gilliamella apicola wkB7 and Gilliamella apis ESL0169 from the honey bee, as well as in another strain of S. alvi isolated from stingless bees (S7 Fig). We showed that all 3 bacteria engineered with our one-plasmid system pAC17V3 (S6A Fig) were able to respond to the presence of IPTG in vitro (S7 Fig).

Download:

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

Fig 5. Engineered S. alvi sense IPTG in vivo and report exposure in the gut tissue and feces.

(a) Schematic of the IPTG-inducible dual plasmid system pAC17V5. (b) S. alvi cells engineered with pAC17V5 respond to IPTG exposure in vitro. Box plots represent median values of GFP fluorescence of 5 biological replicates. Each replicate value is based on the average fluorescence of at least 9,000 S. alvi cells measured by flow cytometry, which were grown in liquid with (+) or without (−) IPTG. (c) Schematic outline of the in vivo experiment using reprogrammed S. alvi to sense IPTG in situ. (d) Confocal microscopy image of a bee gut colonized with engineered S. alvi cells bearing pAC17V5. A biofilm of S. alvi within a gut crypt is shown. The bee was fed with sugar water supplemented with 1 mM IPTG. Blue channel shows DAPI staining of host and bacterial cells. Green channel shows GFP fluorescence. (e) S. alvi engineered with pAC17V5 display a dose-response to IPTG exposure in vivo, which can be measured from gut tissue. Graph shows box plots representing median value of GFP fluorescence of S. alvi biofilms imaged from the gut of bees fed sugar water supplemented with either 0, 0.1, or 1 mM IPTG. Five bees were analyzed for each condition, and fluorescence values were averaged from 3 distinct sections of each gut. Tukey HSD test, * significant at q-value < 0.05. (f) Confocal microscopy image of bee feces, collected from an individual fed with sugar water supplemented with 1 mM IPTG and colonized with S. alvi cells bearing pAC17V5. Green channel shows GFP fluorescence. (g) S. alvi engineered with pAC17V5 display a dose-response to IPTG exposure in vivo, which can be measured from the feces. Box plots represent median values of GFP fluorescence of individual S. alvi cells imaged from the feces of bees fed sugar water supplemented with either 0, 0.1, or 1 mM IPTG. At least 8 bees were analyzed for each condition. The cumulated number n of cells analyzed per conditions is indicated. Tukey HSD test, * significant at q-value < 0.05 and *** significant at q-value < 0.001. The data underlying this Figure can be found in the S1 Data file, sheets “Fig 5B,” “Fig 5E,” and “Fig 5G”.

https://doi.org/10.1371/journal.pbio.3002523.g005

We then wished to establish whether the genetically reprogrammed S. alvi could also sense and respond to IPTG within the gut environment. For this, we monocolonized microbiota-depleted honey bees with S. alvi bearing the pAC17V5 dual-plasmid system, before splitting them into 3 groups that were fed sugar water supplemented with either 0, 0.1, or 1 mM IPTG. Feces and gut tissues were then collected from the honey bees, and the fluorescence of S. alvi cells present in each sample type was analyzed (Fig 5C). In whole guts, S. alvi was found to form biofilms localized in the intestinal crypts of the ileum (Fig 5D). Subsequent image analysis of those biofilms showed that engineered cells exhibited a dose-response to IPTG, whereby the strength of the measured fluorescent signals rose with increasing IPTG concentrations present in the food of the colonized honey bees (Figs 5E and S8).

Lastly, genetically modified S. alvi could also be monitored from their planktonic form contained in fecal material (Fig 5F). The quantification of cell fluorescence from confocal microscopy pictures of the feces was automated via a Fiji macro developed for this study (S1 Text), which allowed the analysis of thousands of individual bacteria in each condition. The obtained data further indicated that S. alvi could sense different concentrations of IPTG in the gut and respond accordingly with a dose-dependent fluorescent output (Fig 5G). Taken together, the in vivo experiments carried out confirmed that the native honey bee gut symbiont S. alvi can be genetically reprogrammed to act as an in situ biosensor of the intestinal environment. Moreover, the cell’s response informing on the exposure to the detected molecule can be assessed locally, directly in the gut tissues, and noninvasively in honey bee feces.

Bacterial strains and plasmids

The type strains S. alvi wkB2^T (ATTC No. BAA-2449) and B. apis PEB0122^T (DSM No. 29779) were routinely grown in Tryptic Soy Broth (TSB, Bacto BD) and Columbia Broth supplemented with 5% defibrinated sheep blood (CBA, Difco BD), respectively. Cells were incubated at 34°C in a microaerophilic environment inside a 5% CO₂ incubator, with orbital shaking (170 rpm) when appropriate. Additionally, E. coli NEB 5-α (New England Biolabs) was used for cloning, while the diaminopimelic acid (DAP)-auxotroph E. coli JKE201 strain [57] was employed for conjugation with B. apis. Both E. coli strains were cultured in normal atmosphere in Lysogeny Broth (LB) at 37°C with orbital shaking (220 rpm), when needed. Colony selection and plasmid maintenance was achieved by supplementing the appropriate culture media with 30 to 60 μg ml⁻¹ spectinomycin (S. alvi/B. apis—E. coli), 30 to 100 μg ml⁻¹ ampicillin (S. alvi/B. apis—E. coli), and 60 μg ml⁻¹ DAP, when required.

The vector pDR401 was generously given by Dr. Shelly Deane (Stellenbosch University, South Africa). The plasmid pME6012 was kindly provided by Prof. Jan van der Meer (Lausanne University, Switzerland). The vectors pSEVA1213S (Addgene plasmid No. 122095) and pBMTBX-2 (Addgene plasmid No. 26073) were gifts from Dr. Pablo Ivan Nikel and Prof. Ryan Gill, respectively. The plasmids pBTK552 (Addgene plasmid No. 110618), pBTK503 (Addgene plasmid No. 110616), and pBTK570 (Addgene plasmid No. 110615) were gifts from Prof. Jeffrey Barrick.

Construction of broad-host range and inducible vectors

To build the broad-host range plasmids, we cloned the different replicons together with a standard fragment bearing either (a) an ampicillin resistance marker with a GFP driven by the CP25 promoter, (b) a spectinomycin resistance marker with a E2-crimson fluorescent protein driven by the PA3 promoter, or (c) an ampicillin resistance marker with a E2-crimson fluorescent protein driven by the PA3 promoter. Each replicon was thus effectively cloned as 3 versions: GFP/ampR, E2-crimson/specR, or E2-crimson/ampR. The latter was generated to allow for the cotransformation of plasmids in B. apis to assess replicon compatibility, as it was found that plasmids expressing gfp were not genetically stable in this species. The pVS1 replicon, which did not replicate in B. apis, was therefore not cloned with the E2-crimson/ampR fragment.

The fragments GFP/ampR, E2-crimson/specR, and E2-crimson/ampR were amplified from the pAC08, pBTK570, and pAC09 vectors, respectively, using the primers AC_09/10. The replicons RK2, pBBR1, pTF-FC2, and pVS1 were obtained from the plasmids pSEVA1213S, pBMTBX-2, pDR401, and pME6012, respectively. The replicons were PCR amplified as 1 or 2 fragments, using the primers AC_16/17 (pSEVA1213S), AC_11/12 (pBMTBX-2), AC_48/49 and AC_50/51 (pDR401), and AC_46/46 with AC_45/47 (pME6012). Similarly, pAC17V5a was constructed by linearizing the pBTK552 plasmid in 2 fragments, using the primers AC_59/AC_114 and AC_60/AC_70, before assembling them back together, effectively removing the lacI gene. The pAC17V5b vector was built by amplifying the lacI gene from pBTK552 using the primers AC_115/AC_116 and cloning it with the pTF-FC2 origin of replication obtained as 2 fragments with the primers AC_36/AC_50 and AC_48/AC_49. Ultimately, all fragments were assembled by Gibson assembly cloning using the NEBuilder HiFi DNA Assembly kit following manufacturer guidelines. Details about cloning primers can be found in S2 Table. The resulting broad-host range and inducible plasmids are listed in S1 Table, and full plasmid maps are depicted in S2 and S6 Figs.

Electroporation

To prepare electrocompetent cells, a single colony of S. alvi was used to inoculate a 5-ml liquid culture of TSB media. Cells were grown to early stationary phase for 2 to 3 days at 34°C with orbital shaking (170 rpm) in a 5% CO₂ incubator. The culture was then diluted 1:50 in 50 ml of fresh TSB and incubated again for 24 hours in identical conditions to allow cells to reach mid-exponential phase (OD ≈ 0.3). Cells were then washed twice with 1:1 (v/v) ice-cold sterile 10% glycerol solution. For this, cells were centrifuged at 2,916g for 10 minutes at 4°C, and the subsequent pellet was resuspended in the glycerol solution. Washed cells were then resuspended in 50 μl of sterile 10% glycerol solution and used for one transformation. Cells were either transformed fresh or stored at −80°C.

Electroporation of cells was performed with an Eppendorf Eporator at 2.5 kV (capacitance 10 μF with resistance 600 Ω) with 0.25 μg of plasmid DNA in 0.2 cm-gap electroporation cuvettes. Cells were resuspended in 1 ml of TSB and allowed to recover for 2 hours at 34°C with orbital shaking (170 rpm) in a 5% CO₂ incubator. Recovered cells were centrifuged (2,916g for 5 minutes at room temperature) and resuspended in 100 μl of fresh TSB before plating onto selective TSA media. Plates were incubated at 34°C with 5% CO₂ atmosphere. Colonies resulting from successfully transformed cells typically appeared after 3 to 4 days, and transformation efficiencies averaged 6.6 10³ ± 1.92 10³ CFU μg⁻¹ of DNA.

Conjugation

Broad-host range plasmids were transferred to B. apis by conjugation. For this, recipients B. apis were obtained by streaking single colonies onto CBA solid media to grow small bacterial lawns after 3 to 4 days of incubation. Cells were gently scrapped off the plates using a sterile plastic loop and resuspended in 500 μl of fresh CBA. The donor strain E. coli JKE201 transformed with the relevant plasmid was grown overnight in 3 ml of LB. The liquid culture was then diluted 1:100 (v/v) into 5 ml of fresh LB. Media used up to this stage was supplemented with DAP and antibiotics, as appropriate. Cells were grown for about 4 hours to mid-exponential phase (OD = 0.4 to 0.6) and then washed 3 times with fresh LB to remove antibiotics. Donor cells were resuspended in 500 μl of LB media without antibiotics.

For cell mating, donor and recipient cells were mixed, pelleted by centrifugation at 2,916g for 5 minutes, resuspended in 50 μl of sterile LB broth, and, finally, aliquoted as 10 μl spots onto solid CBA media supplemented with DAP. Cells were incubated overnight at 34°C in a microaerophilic environment inside a 5% CO₂ incubator to allow mating. To select for exconjugant, cells were scrapped off the plates using a plastic loop and resuspended into 1 ml of sterile CBA. Bacteria were then centrifuged at 2,916g for 5 minutes and homogenized in 100 μl of fresh CBA, to ensure the removal of residual DAP. Finally, cells were plated onto selective CBA plates, which were incubated at 34°C with 5% CO₂ atmosphere. Colonies resulting from successful conjugation typically appeared after 3 to 4 days.

Flow cytometry

Cell fluorescence measurements were performed using a LSR Fortessa Flow Cytometer (BD) instrument and the BD FACSDiva acquisition software (version 9.0, BD). Collected data were processed using the analysis software FlowJo (version 10.8.1). Single colonies of S. alvi and B. apis bearing the different broad-host range plasmids were inoculated into liquid broth and incubated as described above until cells reached late exponential phase. Samples were then diluted 1:10 in sterile PBS prior to flow cytometry analysis. Fluorescent cells were detected by a double discrimination gating strategy, as follows: The cell population was first gated on a FSC-H/SSC-H plot, and fluorescent cells were then discriminated on a FITC-H (ex. 488 nm–em. 494 nm)/SSC-H or AlexaFluor700-H (ex. 633/40 nm–em. 696 nm)/SSC-H plot, for GFP and E2-crimson analysis, respectively (S9 Fig). The reported average fluorescence signal for each sample was calculated based on the measured mean fluorescence values of at least 9,000 cells.

Plasmid copy number measurements by qPCR

Genomic DNA and plasmids were extracted using the Fast Pure Bacteria DNA isolation mini kit (Vazyme) and the QIAprep Spin Miniprep Kit (Qiagen), respectively. DNA concentrations were measured with a Qubit device (Qubit 3.0) and a Qubit dsDNA BR Assay kit (Invitrogen). Standard curves were generated based on known concentrations of serial diluted S. alvi and B. apis genomic DNA, as well as a miniprep of the pBTK570 vector (S3 Fig). qPCR primers used to detect the broad-host range plasmids were AC_24/25, which amplify 100 bp from the E2-crimson gene. Primers for S. alvi and B. apis genomic DNA detection were AC_28/29 and AC_30/31, respectively. Details about the qPCR primers can be found in S2 Table. Each reaction was performed in triplicate, and no-template controls were added to every run to confirm that the reaction mixtures were not contaminated. The generation of specific PCR products by the qPCR primers was confirmed by melting curve analysis and via observation of amplicons on an agarose gel (S3 Fig). Generated standard curves were used to calculate the qPCR efficiency E (values comprised between 90% and 110%) and linearity R² of amplification (S3 Fig).

Plasmid copy number was determined from bacterial lysates. For this, single colonies of S. alvi and B. apis bearing the different broad-host range plasmids were inoculated into liquid broth and incubated as described above until cells reached late exponential phase. Of those cultures, 1 ml was placed into sterile 1.5 ml Eppendorf tubes, boiled for 10 minutes at 100°C in a dry block Thermomixer (Eppendorf), and stored at −20°C until sample processing. Obtained lysates were thawed on ice, homogenized by brief vortexing, and 1 μl of sample was used as template for qPCR reactions, which were performed in a total volume of 10 μl consisting of 5 μl of 2X SYBR Select Master mix (Thermo Fisher), 3.6 μl of MiliQ water, and 0.2 μl of each of the appropriate 10 μM primers. Amplifications were carried out with a QuantStudio 5 real-time PCR system (Thermo Fisher), with the following conditions: 2 minutes at 50°C, 2 minutes at 95°C, followed by 40 cycles of 15 seconds at 95°C, and 1 minute at 60°C. Copy number per cell for each plasmid was calculated by dividing quantities of plasmids with the amount of genomic DNA for each sample, which were determined from the generated standard curves. Copies of genomic DNA were normalized to the number of 16S rRNA loci of the corresponding bacterial genome (i.e., 4 and 2 copies for S. alvi and B. apis, respectively). Reported copy number values were obtained from 3 independent experiments of 5 biological replicate each.

Honey bee rearing and gut colonization

Microbiota-depleted bees Apis mellifera carnica were sourced from outdoor colonies kept at the University of Lausanne (VD, Switzerland), as previously described [58]. In summary, newly emerged honey bees were obtained by transferring mature pupae (i.e., pigmented eyes and light grey cuticula) from brood frames to sterilized plastic boxes. Pupae were kept for 3 days at 35°C with 75% humidity and emerging adult bees had access to a source of sterile 1:1 (w/v) sucrose water. A day prior, the hindguts of 2 newly emerged bees per pupae box were dissected and homogenized in 1 ml sterile 1X PBS to check for contaminated bees. For this, gut homogenates were plated onto NA, CBA, and MRSA solid media and incubated in normal atmosphere, microaerophilic and anaerobic environment, respectively. Pupae boxes for which bacterial growth was observed on the plates of the corresponding tested bees were discarded.

Gut monocolonization of the microbiota-depleted bees was carried out by individually feeding them 5 μl of solutions of S. alvi cells at an OD₆₀₀ of 0.1 resuspended in 1:1 (v/v) 1X PBS:sucrose water, representing an average of 1.9 10⁴ ± 1.89 10³ bacteria per inoculum. For colonization by S. alvi with the natural microbiota, gut homogenate stocks were obtained by combining in equal volumes the homogenized hindguts of 5 hive bees. Solutions used to feed the bees for colonization were then prepared by mixing S. alvi cells to a final OD₆₀₀ of 0.1 with 1:10 (v/v) diluted gut homogenate stock in 1:1 (v/v) 1X PBS:sucrose water. Colonized bees were kept at 32°C with 75% humidity in sterile cup cages with added sterilized pollen and sucrose solution, the latter supplemented with 30 μg ml⁻¹ spectinomycin, when required. Honey bees monocolonized with the engineered S. alvi strain bearing the IPTG-inducible pAC17V5 plasmids were similarly fed with sterilized pollen and sucrose solution, but the latter was supplemented with 30 μg ml⁻¹ spectinomycin, 30 μg ml⁻¹ ampicillin, and 0.1 or 1 mM IPTG, as appropriate. Supplemented sucrose solutions (i.e., spectinomycin and IPTG) were replaced with freshly prepared tubes every 3 days throughout the experiments to ensure constant concentrations of the corresponding molecules.

Bacterial load and plasmid maintenance assessment via fecal sampling

To sample feces from honey bees, colonized bees in cup cages were first stunned using CO₂ and immobilized at 4°C on ice. Gentle pressure was applied by hand on the abdomens of asleep bees using a slight motion, starting from the base of the abdomen towards the stinger, until defecation occurred into the cap of sterile 1.5 ml Eppendorf tubes. Pressure applied on the abdomen of the bees was relieved 2 seconds after defecation, which resulted in about 4 μl of feces on average (Fig 1B). Samples were kept on ice until the end of the sampling procedure.

To assess bacterial load from the obtained feces, fecal matter was either directly serially diluted 1:10 (v/v) in sterile 1X PBS if liquid enough for pipetting or first resuspended in 4 μl of sterile 1X PBS prior to the serial dilution if too difficult to pipet. Diluted feces were then plated on TSA solid media supplemented with or without 60 μg ml⁻¹ spectinomycin, as appropriate. CFUs of the engineered S. alvi cells were then counted after 3 to 4 days incubation. For the assessment of plasmid maintenance, fluorescent CFUs were discriminated against nonfluorescent colonies and counted from images of the plates taken by EPI fluorescence with a Fusion FX (Vilber Lourmat) apparatus (high sensitivity, 8 aperture, F-740 filter). A median value of 69.5 colonies were counted and analyzed to obtain the plasmid maintenance percentage for each sample.

Honey bee survival assays

To measure the impact of fecal extraction on honey bee’s health, we reared 2 cohorts of honey bees, with one subjected to weekly feces extraction (as described above) and a second control group that was not manipulated. To account for cage effect, each cohort comprised 2 separate cages of at least 10 honey bees. For 22 days, survival of bees was monitored daily, during which occasions dead individuals were also removed.

Gut and feces microscopy analysis

To prepare samples for microscopy, the hindguts of honey bees were dissected a week after inoculation with the engineered S. alvi and immediately transferred to 4% paraformaldehyde in PBS and fixed overnight at 4°C with rotation. Fixed samples were then washed 3 times for 30 minutes in PBS. Hindguts were permeabilized and stained overnight in the dark at 4°C with rotation, with 5 μg ml⁻¹ 4,6-diamidino-2-phenylindole (DAPI; ex. 359 nm–em. 457 nm) in 1% Triton X-100 in PBS. Excess dye was then washed from fixed samples with PBS, and the ileum of stained guts was dissected out, mounted in PBS, and overlaid with a coverslip (number #1.5). The laser scanning confocal microscopy was performed using an inverted Zeiss LSM 980 Airyscan 2 microscope (Carl Zeiss AG, Jena, Germany). Images were acquired with a 512 × 512 pixel format size (134.6873 × 134.6873 μm and depth of 1.8400 μm), an averaging of 4 images and a 63× oil objective.

For intensity measurement, all ileums were imaged with the same laser intensity and master gain on the detector. Three distinct sections per ileum were imaged. File names were randomized for blind intensity quantification. In Fiji, each image was split into DAPI and GFP channels, and the freeform tool was used to draw 3 densely colonized areas of each ileum section, based on the DAPI channel. Host cells were excluded from the area of interest. Selected areas were then copied to the corresponding images of the GFP channel. The intensity measurement used was the mean intensity per unit area of each specified region with the background subtracted. Normalized intensities of the 3 selected areas per image were averaged, resulting in a single value per ileum section. Finally, GFP intensities reported per gut were mean values of the averaged signals obtained from the corresponding 3 ileum sections imaged. To ensure differences in GFP fluorescence were due to IPTG rather than analytical biases, similar quantification of DAPI signal was performed on the same selected image areas, which showed no significant difference in DAPI signal between the different IPTG treatments (S8 Fig).

To quantify the GFP fluorescence of S. alvi from fecal samples, feces were first collected from honey bees colonized with the engineered bacteria a week after inoculation and immediately diluted 1:10 in sterile PBS. Before being imaged with the same microscopy settings described above for the honey bee guts, 5 to 10 μl of diluted feces was overlaid with a coverslip. Seven distinct areas per feces sample were imaged. Fluorescence quantification of cells was automated via a Fiji macro developed for this study (see Code availability section). Macro was run using the software Fiji (ImageJ2, version 2.9.0). Briefly, cells were counted and segmented based on a GFP fluorescence threshold, and the GFP signals of all cellular areas were averaged per image. Finally, the mean GFP intensities reported per bee (i.e., per fecal sample) were mean values of the averaged signals obtained from the corresponding 7 sample areas imaged.