Lethal Consequences of Overcoming Metabolic Restrictions Imposed on a Cooperative Bacterial Population

ABSTRACT Quorum sensing (QS) controls cooperative activities in many Proteobacteria. In some species, QS-dependent specific metabolism contributes to the stability of the cooperation. However, the mechanism by which QS and metabolic networks have coevolved to support stable public good cooperation and maintenance of the cooperative group remains unknown. Here we explored the underlying mechanisms of QS-controlled central metabolism in the evolutionary aspects of cooperation. In Burkholderia glumae, the QS-dependent glyoxylate cycle plays an important role in cooperativity. A bifunctional QS-dependent transcriptional regulator, QsmR, rewired central metabolism to utilize the glyoxylate cycle rather than the tricarboxylic acid cycle. Defects in the glyoxylate cycle caused metabolic imbalance and triggered high expression of the stress-responsive chaperonin GroEL. High-level expression of GroEL in glyoxylate cycle mutants interfered with the biosynthesis of a public resource, oxalate, by physically interrupting the oxalate biosynthetic enzyme ObcA. Under such destabilized cooperativity conditions, spontaneous mutations in the qsmR gene in glyoxylate cycle mutants occurred to relieve metabolic stresses, but these mutants lost QsmR-mediated pleiotropy. Overcoming the metabolic restrictions imposed on the population of cooperators among glyoxylate cycle mutants resulted in the occurrence and selection of spontaneous qsmR mutants despite the loss of other important functions. These results provide insight into how QS bacteria have evolved to maintain stable cooperation via QS-mediated metabolic coordination.

T he recognition of bacteria as social organisms has provided population and evolutionary biological perspectives on bacterial behavior. In Proteobacteria, acylhomoserine lactone (AHL)-mediated quorum sensing (QS) controls social behaviors, including swarming motility, virulence, and biofilm formation (1)(2)(3). These social behaviors require cooperation, but the evolutionary stability of cooperation is precarious, because costly cooperative strategies are vulnerable to social cheating. Several studies have demonstrated that QS-dependent metabolism stabilizes cooperativity. In Pseudomonas aeruginosa, the production of such public resources is positively or negatively regulated in a QS-dependent manner to minimize production costs (3)(4)(5). The integration of metabolic information with QS stabilizes public good cooperation, as cells can cooperate only when they receive the appropriate nutritional resources (6).
Along with the known QS-regulatory mechanisms that support bacterial sociality, we explored whether metabolic evolution to sustain bacterial cooperativity is inherent in the structure of bacterial primary metabolic networks. To address this issue, Burkholderia glumae BGR1 was chosen as a model bacterium because of its dramatic QS-mediated metabolic fluctuations and metabolic plasticity (7,8). In B. glumae, one LuxI-R-type QS system, TofI-R, generates N-octanoyl homoserine lactone (C8-HSL) as a major signaling molecule (9). The TofR and C8-HSL complex regulates expression of the qsmR gene, which encodes an IclR (isocitrate lyase regulator)-type transcriptional regulator (10). QsmR regulates various private and public resources, including oxalate in B. glumae (9)(10)(11). Burkholderia thailandensis E264, a nonpathogenic saprophyte, was used as a model system comparable to B. glumae BGR1. B. thailandensis contains three acyl-homoserine lactone QS circuits. The QS-1 signal (C8-HSL) is synthesized by BtaI1 complexes with BtaR1 to control aggregation, motility, and oxalate production (12). QS is essential for stationary-phase survival in both B. glumae and B. thailandensis (11). When these Burkholderia species use amino acids as a carbon source, deamination results in ammonia release, increasing the extracellular pH (11). Among QS-dependent public goods, oxalate is biosynthesized in the branched tricarboxylic acid (TCA) cycle to protect cells from ammonia-mediated alkaline toxicity during the stationary phase in a QsmR-dependent manner ( Fig. 1) (11). The oxalate biosynthetic enzymes ObcA and ObcB use acetyl-coenzyme A (acetyl-CoA) and oxaloacetate as substrates (13,14).
In this article, we report previously unknown functional roles of the glyoxylate cycle in bacterial cooperativity. The glyoxylate cycle allows an organism to bypass the CO 2 -generating steps in the TCA cycle (15). As a result, the glyoxylate cycle is essential for growth when acetate is the sole carbon source, as is the case for Escherichia coli (16). In addition to this known function of the glyoxylate cycle, we hypothesized that this When Burkholderia uses amino acids as a carbon source, ammonia is produced by deamination, which increases the extracellular pH (11). Among QS-dependent public goods, oxalate is biosynthesized in the branched TCA cycle to protect cells from ammonia-mediated alkaline toxicity during the stationary phase of growth (11). QsmR, the QSdependent transcriptional regulator, directly regulates the expression of the obcAB genes (11). The glyoxylate cycle is composed of two specific enzymes: isocitrate lyase and malate synthase. Isocitrate dehydrogenase in the TCA cycle and isocitrate lyase both use isocitrate as a substrate. The solid and dotted lines denote activation and repression by QS, respectively. cycle may play an important role in bacterial cooperativity. This hypothesis was based on previous QS-dependent transcriptome analyses of B. glumae that showed that the expression levels of two genes in the glyoxylate cycle, aceA and glcB (encoding isocitrate lyase and malate synthase, respectively), are QS dependent (11). We found that QsmR is bifunctional and rewires metabolic networks at the branch point of the glyoxylate and TCA cycles. Despite the presence of an intact QS system, a defective glyoxylate cycle generated metabolic stress and induced high expression of the stressresponsive chaperonin GroEL. The excessive levels of GroEL in the glyoxylate cycle mutants reduced the biosynthesis of oxalate by interrupting the oxalate biosynthetic enzyme ObcA. Metabolic imbalance in the glyoxylate cycle mutants exerted selection pressure, leading to spontaneous mutations in the qsmR gene. Such spontaneous mutations in the qsmR gene relieved the metabolic stress caused by the absence of a functional glyoxylate cycle, but they were accompanied by the loss of pleiotropy, including activation of obcAB gene expression, which subsequently disrupted cooperativity. From our results, we suggest that QS has evolved to be integrated into the central metabolism to maintain stable cooperation along with ensuring public good production.

RESULTS
Metabolic rewiring by QS at the branch point of the glyoxylate and TCA cycles. On the basis of our previous QS-dependent transcriptome analysis, we predicted that central metabolism is QS dependent and plays an important role in B. glumae cooperativity. We confirmed that expression of aceA and glcB in the glyoxylate cycle is directly controlled by QsmR ( Fig. 2A; see also Fig. S1A in the supplemental material). Isocitrate lyase activity was significantly higher in the wild-type strain than in the tofI mutant, BGS2, and the qsmR mutant, BGS9 (Fig. 2B). Exogenous addition of 1 M C8-HSL to the BGS2 tofI mutant returned the isocitrate lyase activity to wild-type levels (Fig. 2B). Complementation strains of the qsmR and aceA mutants recovered full isocitrate lyase activity (Fig. 2B)  genes, which encode isocitrate lyase and malate synthase, respectively, was activated by QsmR in B. glumae. (B) Isocitrate lyase activity was higher in the wild-type strain than in the QS mutants of B. glumae. (C) Isocitrate dehydrogenase (encoded by icd1) activity was higher in the qsmR mutant than in the wild-type strain of B. glumae. A mutation in the aceK gene, which encodes isocitrate dehydrogenase kinase/phosphatase, had no effect on isocitrate dehydrogenase activity in B. glumae. Error bars represent the error ranges of experiments performed in triplicate. ns, no significant difference; *, P Ͻ 0.05; **, P Ͻ 0.001.
significantly higher in the wild-type E264 strain than in the bta1 mutant, JBT101, and the qsmR mutant, BT09539 (Fig. S1B). However, expression of the icd gene, which encodes isocitrate dehydrogenase (IDH), was controlled directly by QsmR, but isocitrate dehydrogenase activity was significantly higher in the qsmR mutants than in the wild-type strains of B. glumae and B. thailandensis ( Fig. 2C; see also Fig. S1A and C).
These results indicate that QsmR rewires metabolic networks to utilize the glyoxylate cycle in preference to the TCA cycle at the branch point of these two metabolic cycles in both bacterial species. Role of the QS-dependent glyoxylate cycle in B. glumae virulence. Since defects in the glyoxylate cycle affected cooperativity in B. glumae, we examined whether the QS-dependent glyoxylate cycle plays an important role in B. glumae virulence. Following the inoculation of rice panicles (in the flowing stage) with the wild-type strain and the aceA mutant, we observed a disease index for the aceA mutant that was 30% lower than that for the wild-type strain (Fig. 3). Our results indicated that the glyoxylate cycle-deficient mutant was less virulent than the wild-type strain in rice panicles. Inoculation with the complementation aceA strain recovered the disease index to wild-type levels. These results indicate that the QS-dependent glyoxylate cycle is important for full B. glumae virulence in rice panicles.
Defects in the glyoxylate cycle affect cooperative behaviors. To evaluate the biological significance of such metabolic rewiring in B. glumae and B. thailandensis, we investigated whether the glyoxylate cycle plays a role in sustaining cooperativity. Growth, extracellular pH, ammonia accumulation, and oxalate production of the aceA mutants of B. glumae and B. thailandensis were monitored for 7 days in LB medium to identify any differences in known cooperative behaviors. Growth of the aceA mutants was comparable to that of the wild-type strains until the stationary phase was reached. Viable cell counts of the aceA mutants of B. glumae and B. thailandensis were decreased by approximately 10% compared to those of the wild-type population 1 to 3 days after subculture ( Fig. 4; see also Fig. S2A). Extracellular pH of the aceA mutants of B. glumae and B. thailandensis remained at 7.6 to 8.2 throughout the growth period 1 day after subculture, whereas the wild-type strain exhibited slight alkalization followed by acidification and neutralization patterns ( difference was found between the prpB mutant and the wild-type strain (see Fig. S2B).
These results indicate that isocitrate lyase and methylisocitrate lyase have some crossactivities but that methylisocitrate lyase alone does not influence oxalate biosynthesis significantly in B. glumae.

Interruption of ObcA by GroEL.
To identify the molecular mechanisms underlying the significant reduction in oxalate biosynthesis in the B. glumae aceA mutant, we measured the transcriptional and translational levels of obcA in the wild-type strain and aceA mutant. No difference between the wild-type strain and the aceA mutant was observed in the transcriptional or translational levels of obcA ( glumae were decreased by approximately 10% compared to that of wild-type strain, and the extracellular pH of the aceA mutant was maintained at 7.6 to 7.9 throughout growth, whereas the wild-type strain showed acidification and neutralization patterns. Ammonia production by the aceA mutant was slightly lower than that by the wild-type strain, and oxalate production by the aceA mutant was~60% that of wild-type strain. mutant, we performed a transcriptome sequencing (RNAseq) comparison of the wildtype strain and the aceA mutant strain. We found that expression of genes encoding stress-responsive chaperones or chaperonins, such as GroEL, was approximately 30-fold higher in the aceA mutant than in the wild-type strain ( Table 1). The expression level of groEL was significantly higher in the aceA mutant than in the wild-type strain at both the transcriptional and translational levels (see Fig. S4). These results prompted us to hypothesize that highly expressed GroEL in the aceA mutant interacts with ObcA and interferes with its activity. GroEL protein was pulled down using an anti-ObcA antibody in affinity pulldown experiments in both the wild-type strain and aceA mutant (Fig. 5B), which indicated that ObcA and GroEL interact. Interestingly, even though the same amounts of ObcA were pulled down from all samples, the amount of ObcA that interacted with GroEL was higher in the aceA mutant than in the wild-type strain (Fig. 5B). The higher level of GroEL in the aceA mutant relative to the wild-type strain was in accordance with the RNAseq transcriptomic data, which showed a higher level of GroEL in the aceA mutant than in the wild-type strain (see Fig. S4). To demonstrate that the interaction between GroEL and ObcA inhibits the enzymatic activity of ObcA, we performed an ObcA activity assay in vitro. Addition of 500, 1,000, and 1,500 nM GroEL reduced ObcA activity by approximately 4%, 52%, and 68%, respectively, compared to the control without GroEL (Fig. 5C). ObcA activity was not affected by addition of the nonspecific protein bovine serum albumin (BSA) (Fig. 5C). These results suggested that the reduced activity of ObcA in the aceA mutant was due to physical interruption of ObcA by excess GroEL. Metabolic stress in the aceA mutant. Based on the differential expression of multiple genes encoding chaperones and chaperonins in the aceA mutant of B. glumae, we hypothesized that the mutant may experience intracellular physiological stress. Since the aceA mutation blocks the glyoxylate cycle, we assumed that either isocitrate dehydrogenase (IDH) activity is increased to mitigate metabolic stress in the aceA mutant or isocitrate, aconitate, and citrate accumulation is greater in the mutant than in the wild-type strain. IDH activity in the aceA mutant was not significantly different from that of the wild-type strain (Fig. 2C). However, concentrations of isocitrate, aconitate, and citrate were significantly higher in the aceA mutant than in the wild-type strain (Fig. 6A). Consistent with these results, both the transcriptional level of the citA gene, which encodes citrate synthase, and its enzymatic activity were higher in the aceA mutant than in the wild-type strain ( Fig. 6B; Table 1). These results indicate that the aceA mutant experiences physiological stresses due to metabolic imbalance, which is supportive of the concept of elevated expression of stress-responsive chaperones or chaperonins, including GroEL.
Emergence of spontaneous mutations in the qsmR gene in glyoxylate cycledefective strains. Such continuous intracellular metabolic imbalance may not be endurable by mutants with a nonfunctional glyoxylate cycle. Since we observed morphologically distinct colonies when glyoxylate cycle-defective mutants were cultured in the preliminary experiment, colony morphology was evaluated at various culture time points. We collected samples from batch cultures in LB broth each day (days 0, 1, 2, 3, 4, 5, 6, and 7). The samples were serially diluted, spread on an LB agar  and no survivors were detected (see Fig. S6). To characterize such spontaneously occurring mutants, two mutants that were siblings were subjected to whole-genome resequencing. We found that IS1418, an insertion sequence (IS) 865 bp in length that carries 15-bp inverted repeats with a target duplication of 3 bp (17), was inserted in the qsmR gene in each mutant. On the basis of whole-genome resequencing data, we amplified the qsmR gene region of another 30 mutants. PCR products representing three different mutation types, insertion, small deletion, and large deletion, were observed (Fig. 7C). We selected three insertion mutants (qsmR_IS), four small-deletion mutations (qsmR_SD), and three large-deletion mutants (qsmR_LD) to determine the nature of the mutations by sequencing (Fig. 7C). These spontaneous qsmR mutants exhibited elevated IDH activities due to derepression of icd genes (Fig. 7D), which indicates that instinctive coping mechanisms to relieve metabolic stresses are operational in the glyoxylate cycle mutants. However, these mutants consequently failed to activate oxalate biosynthesis (see Fig. S7A). Thus, due to the toxic alkaline extracellular pH, the spontaneous qsmR mutants did not survive stationary phase after they were isolated, as previously observed in the qsmR mutant, BGS9 (see Fig. S7A) (11). To determine whether the spontaneous qsmR mutants are cheaters, we monitored the population density, pH, oxalate, and ammonia in mixed culture with the wild type. Both the wild type and the spontaneous mutants failed to survive in mixed cultures ( Fig. 7E; see also Fig. S7B). This population collapse was due to the alkaline condition caused by ammonia accumulation by the wild type and the spontaneous mutants and also to the level of oxalate being insufficient to neutralize it in mixed culture ( Fig. 7E; see also Fig. S7B).

DISCUSSION
Positive control of the glyoxylate cycle by QS allowed us to predict that the QS-dependent glyoxylate cycle may have an important role in sustaining cooperativity. In fact, it is known that the glyoxylate cycle is positively controlled by QS in B. cepacia, Yersinia pestis, and P. aeruginosa (18)(19)(20). Such upregulation of the glyoxylate cycle by QS provides advantages for growth on certain carbon resources (19,20). However, the biological significance of positive control of the glyoxylate cycle by QS is unknown. In this study, we explored whether the QS-dependent glyoxylate cycle plays a role in bacterial sociality.
We first showed that QsmR rewires metabolic networks at the branch point of the glyoxylate cycle and TCA cycle in B. glumae. This indicated that QsmR functions as a transcriptional activator and a repressor, which is not unusual because transcriptional regulators belonging to the IclR type are often bifunctional (21). Such metabolic rewiring by bifunctional QsmR raised questions as to what the underlying reasons are and whether it is related to bacterial cooperativity. We chose B. thailandensis E264 as a system comparable to B. glumae, because it also exhibited QmsR-dependent oxalate biosynthesis (11) as well as an 80.53% identity with B. glumae BGR1 at the genome level, as calculated by average nucleotide identity (ANI) based on BLASTϩ (ANIb) parameters (22). Concerns about the possibility that QsmR-mediated metabolic rewiring is a peculiar phenomenon in B. glumae were dismissed since expression of aceA, glcB, and icd genes in B. thailandensis exhibited the same gene regulation patterns as those observed in B. glumae. This phenomenon appears to be widely distributed among the members of the Burkholderia genus. Recent QS-dependent transcriptome analysis in B. thailandensis showed that expression of both the aceA and glcB genes is not controlled by QS (12). However, this may be due to different culture conditions. For QS-dependent transcriptome analysis, the strains of B. thailandensis were grown in LB supplemented with 50 mM MOPS (morpholinepropanesulfonic acid) (12) whereas expression of both the aceA and glcB gene was estimated in B. thailandensis cells grown in LB broth without any buffer in this study.
Since methylisocitrate lyase encoded by the prpB gene has cross-activity with isocitrate lyase in B. pseudomallei and Mycobacterium tuberculosis (23, 24), we expected a similar phenomenon in B. glumae. Cross-activity of methylisocitrate lyase and isocit-rate lyase was detectable; however, the level of the cross-activity was not high enough to substitute isocitrate lyase with methylisocitrate lyase in the aceA mutant. Thus, the activity of methylisocitrate lyase has a negligible effect on oxalate biosynthesis in the wild-type strain.
Similarly to E. coli, the glyoxylate cycle in B. glumae is essential for growth on acetate (see Fig. S8 in the supplemental material). The glyoxylate cycle is important for the survival of pathogens inside the host and provides a physiological benefit to certain pathogenic microorganisms (19,20,24). A similar role of the glyoxylate cycle was observed in B. glumae since the aceA mutant was less virulent than the wild-type strain in rice panicles (Fig. 3). In addition to these biological roles of glyoxylate cycle, we were interested in the biological meaning of QS dependency in the glyoxylate cycle and in QsmR-mediated metabolic rewiring.
It was interesting that the glyoxylate cycle affects biosynthesis of a public resource, oxalate. We then explored why and how the glyoxylate cycle affects oxalate biosynthesis in B. glumae. Our systematic analyses of TCA cycle intermediates and the transcriptome of glyoxylate cycle-defective mutants led us to find that blocking of the glyoxylate cycle leads to stressful conditions and triggers high-level expression of GroEL. Considering that wild-type cells have normal metabolic conditions in the presence of normal concentrations of citrate, aconitate, and isocitrate, it was clear that the aceA mutant experiences stressful metabolic conditions due to significantly high concentrations of these metabolites. Elevated expression of stress-responsive chaperones strongly supported the concept that physiological conditions in the glyoxylate cycle defectives are indeed stressful. The idea of interaction of the oxalate biosynthetic enzyme ObcA with GroEL was supported by the results of affinity pulldown experiments in both the wild-type strain and aceA mutant. In general, GroEL is a chaperonin that facilitates the proper folding of proteins, but it has been reported that an excess of certain chaperones exerts a negative effect on protein production (25). Our current results do not provide evidence that GroEL is required for full activity of ObcA. Rather, the in vitro ObcA activity assay suggested that excess GroEL hampered ObcA activity, resulting in reduced oxalate production (Fig. 5C). These findings indicated that the metabolic flow through the glyoxylate cycle is essential for producing oxalate as a public good in a QS-dependent manner in B. glumae.
In glyoxylate cycle-defective mutants, metabolic stresses were exerted as a selection pressure on the metabolic networks, resulting in the occurrence of spontaneous mutations in the qsmR gene to derepress expression of icd genes. These spontaneous mutations in the qsmR gene had a direct effect on the biosynthesis of public goods (see Fig. S7A), which is quite different from nonsocial adaptation, in which an individual's fitness would increase by derepressing transcriptional repressor PsdR, for example, in P. aeruginosa (26). Spontaneous mutations in the qsmR gene were not observed in the wild-type B. glumae strain under the same growth conditions as those of the aceA mutant. Occasionally, QS enhancement of extracellular stress responses, such as the oxidative stress response, acts as a counterselective force for the appearance and survival of QS cheaters. However, intracellular metabolic stress acts as a pressure on individual cells to develop genomic mutations to relieve such stress, even though these mutations lost all of their attributes under the control of the QS-dependent QsmR. Therefore, intracellular metabolic homeostasis is an important issue in Burkholderia cooperative cells. It might be assumed that spontaneous qsmR mutants in the glyoxylate cycle-defective B. glumae strains are cheaters. However, these mutants are not the cheaters, because three types (IS, SD, and LD) of spontaneous qsmR mutant did not benefit from mixed culture with the wild-type oxalate producer ( Fig. 7E; see also Fig. S7B). Mutations that inactivate qsmR were the fastest means of rescuing the metabolic stress but did not benefit individual cells.
In B. glumae, the QS-dependent glyoxylate cycle functions to support bacterial cooperativity, as well as to provide a metabolic bypass. Our results reveal important physiological roles of the glyoxylate cycle in cooperative bacteria and suggest that the glyoxylate cycle may represent a target for the development of chemical agents to control bacterial social behaviors, such as the virulence of pathogens. This may also be a good example of overcoming metabolic restrictions caused by blocking a critical metabolic flow in a population of cooperators.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Bacterial strains and plasmids used in this study are listed in Table S1 in the supplemental material. Strains of B. glumae and B. thailandensis were grown in Luria-Bertani (LB) broth (Affymetrix, Santa Clara, CA) (0.1% tryptone, 0.5% yeast extract, and 0.5% NaCl [all wt/vol]) at 37°C.
Quantitative reverse transcription-PCR (qRT-PCR). Total RNAs from B. glumae BGR1, BGS2 (BGR1 tofI::⍀), BGS9 (BGR1 qsmR::⍀), and S9NC5 (BGR1 qsmR::⍀/qsmR), grown in LB medium at 37°C for 10 h after subculture, were extracted using RNeasy minikits (Qiagen, Venlo, Netherlands), as described by the manufacturer. Total RNA was treated with RNase-free DNase I (Ambion, Waltham, MA) to remove genomic DNA. Total RNA (1 g) was subjected to reverse transcription into cDNA using Moloney murine leukemia virus (MMLV) reverse transcriptase (Promega, Madison, WI) and incubation for 1 h at 42°C. Primer pairs used for qRT-PCR are listed in Table S2. The 16S rRNA gene served as the positive control. Transcriptional levels were determined using SsoFast Eva Green SuperMix (Bio-Rad, Hercules, CA) and a CFX96 Real-Time PCR system (Bio-Rad). The thermal cycling parameters were as follows: 95°C for 30 s, followed by 40 cycles of 95°C for 5 s and 60°C for 5 s. All PCRs were performed in triplicate, and all data were normalized to the expression levels of the 16S rRNA gene using Bio-Rad CFX Manager software.
Isocitrate lyase activity assay. Isocitrate lyase activity was measured as described previously (27). Briefly, isocitrate was converted to succinate and glyoxylate by isocitrate lyase. During a reaction with phenylhydrazine, glyoxylate formed phenylhydrazine-glyoxylate, and the absorbance of this compound was measured at 324 nm every 60 s for 5 min. Enzymatic activity (in units per milliliter) was calculated as follows: ΔA 324nm /min sample Ϫ ΔA 324nm /min blank)/millimolar extinction coefficient of phenylhydrazine-glyoxylate at 324 nm.
Isocitrate dehydrogenase activity assay. Isocitrate dehydrogenase activity was measured using an isocitrate dehydrogenase activity assay kit (BioVision, Milpitas, CA) according to the manufacturer's instructions. We measured the absorbance at 450 nm every 1 min in a microplate reader (PerkinElmer, Waltham, MA) and used only the data obtained within the linear range to calculate isocitrate dehydrogenase activity using the NADH standard curve, as described by the manufacturer.
Measurement of secreted oxalate and ammonia. Secreted oxalate and ammonia levels were measured as described previously (11).
Overexpression and purification of GroEL. The groEL gene of B. glumae was amplified using the primers listed in Table S2. The amplified product was cloned into NdeI and HindIII restriction sites of a pET28b expression vector (Merck, Darmstadt, Germany), resulting in pET28b-groEL (see Table S1).
EZ-Tn5 mutagenesis and marker exchange. pICL1 and pPEP1 were mutagenized using a EZ-Tn5 ϽDHFR-1Ͼ insertion kit (Epicentre, Madison, WI) and marker exchange, as described previously (10). All constructs were confirmed by Southern blot analysis. The EZ-Tn5 insertion sites were determined by DNA sequencing according to the manufacturer's protocols.
Statistical analysis. Statistical analyses were performed using SPSS version 22 (IBM Corp., Armonk, NY). Comparisons were performed by one-way analysis of variance (ANOVA), followed by Tukey's multiple-comparison test (set at 5%). In Fig. S3A in the supplemental material, the data represented by the 2 bars were compared by unpaired t test.
Accession number(s). The NCBI GEO accession number for the RNAseq data series of BGR1, BICL39, and BICL39C is GSE70582.