Strains and related plasmids used in this study are listed in Tables 1, 2, respectively. We routinely cultured Escherichia coli in LB medium at 37°C and Bt in LB medium at 28°C and examined their growth curves. The final concentrations of antibiotics added to the medium used in this study were: ampicillin, 100 μg/mL; kanamycin, 50 μg/mL; spectinomycin at 300 μg/mL and 100 μg/mL for Bt and E. coli, respectively. Restriction endonucleases were purchased from Takara Company, and DNA polymerases from CWbio Company. The PCR-related primers involved in this study are listed in Supplementary Table 1. All PCR experiments were carried out using the genomic DNA of the Bt starting strain BMB171 (GenBank accession number NC_014171NC_014171) or BMB171-derived strains as templates (He et al., 2010).

The primers for constructing the above-mentioned knockout mutants are listed in Supplementary Table 1; they were used to amplify the upstream and downstream sequences of the target genes to facilitate double crossover by homologous recombination. We also made appropriate improvements to the methods reported in the literature (Janes and Stibitz, 2006) to obtain a protocol suitable for knocking out genes in Bt (Zheng et al., 2015), and the target mutants were verified by PCR and sequencing (Fu et al., 2018).

All tested strains were inoculated at the same initial concentration and grown under the same culture conditions. After shaking on a shaker for 9 h, 5 μL of the bacterial liquid culture in the mid-logarithmic phase was taken and dropped onto the center of the plate in semi-solid medium (LB containing 0.5% agar), followed by standing for 10–15 min. After the bacterial liquid became solidified, it was transferred to a 28°C incubator for further culturing for 8 h (Fu et al., 2018).

The determination of biofilm amount was appropriately optimized with reference to the method used for Clostridium difficile (Bordeleau et al., 2011): the strains to be tested were cultured with shaking in glass bottles of LB liquid medium for 12 h, and then left to stand for another 36 h. After gently removing the bacterial liquid culture, the glass bottles were washed twice with ultrapure water. Adherent biofilm was then stained with 1% crystal violet for 30 min at room temperature for quantification. After washing off the residual dye with ultrapure water and allowing the glass bottles to dry, the crystal violet-stained biofilm was re-suspended in 2 ml of 96% (v/v) ethanol and the absorbance of purple suspension at 595 nm was measured (Tang et al., 2016).

The spore counting method was taken from that reported in the references (Wang et al., 2016; Li et al., 2020) and adjusted as follows: different strains were cultured in LB medium, and then 1 mL of each sample was taken out and placed in an 80°C water bath for 10 min. After appropriate serial dilutions, 100 μL of each diluent was spread onto LB plates to count the CFU (colony-forming units) per mL. The calculation formula of the spore numbers is as follows: the actual spore numbers = (the single colony numbers on the plate × the dilution factor × the initial volume of the bacterial suspension)/the spreading volume of the bacterial suspension.

The scanning electron microscopy imaging method in this study comes from the literature (Tang et al., 2016). Bt cultured in LB medium for 9 or 27 h were dropped on a microscope slide and placed at 28°C in an incubator for 27 h. The samples were then fixed with 4% glutaraldehyde for 2 h and dehydrated with ethanol solutions of different concentrations. The samples were finally freeze-dried and imaged using a JSM-6390 Scanning electron microscopy (JEOL, Japan).

Tested strains BMB171, Δ3hfq, BMB171-cry, and Δ3hfq-cry were grown under the same conditions, and the bacterial cultures in the late stationary phase or early sporulation phase (48 h) were centrifuged at 10,000 × g, 4°C for 5 min for collection of pellets, which were re-suspend with an appropriate amount of ultrapure water to obtain properly diluted bacterial suspensions before adding it to the artificial feed. The artificial diet formula for breeding cotton bollworm (Helicoverpa armigera) is: 4 g of yeast extract, 7 g of soybean meal, 0.5 g of vitamin C, 1.5 g of agar, 1.5 mL of 36% acetic acid and 2 g of penicillin in 100 mL of artificial feed medium (Tang et al., 2016). The artificial feed was then dispensed into 24-well cell culture plates (Costar, United States) at 1 mL/well, and 100 μL/well of the bacterial suspension to be tested was added to the culture plate. Each group of 72 cotton bollworm larvae was then placed in 24-well culture plates, with the survival rate recorded daily. After 7 days of continuous observation, the body length and body weight of the larvae were measured.

First, total RNA was isolated from Bt cells grown in LB liquid medium to specific time points using TRIzol reagent (Life Technologies, United States). Transcription of the first cDNA was carried out using the PrimeScript RT kit with gDNA Eraser (Takara Biotechnology, Japan) and used as the DNA template for real-time quantitative PCR (RT-qPCR) as previously described (Zheng et al., 2020) using gapdh (encoding glyceraldehyde 3-phosphate dehydrogenase) gene as the reference (Løvdal and Saha, 2014). Finally, we used the comparative C_t method to calculate changes in gene expression.

BMB171 and Δ3hfq cultured for 48 h were selected as transcriptome sequencing samples, with two biological replicates set for each sample. Total RNA was extracted with TRIzol reagent (Life Technologies, United States), and after appropriate processing, paired-end sequencing was carried out using an Illumina HiSeq TM 2500 sequencer (Illumina, United States). Sequencing was obtained by strand-specific RNA-seq using the Illumina Genome Analyzer IIx sequencing platform (Li et al., 2017). The raw data were deposited in NCBI under the Sequence Read Archive (SRA) with the BioProject number PRJNA811318 and SRA accession number SRP361911. The raw data were collected and filtered by FASTX-Toolkit¹ to get a clean data set that could be further analyzed and processed to clean-reads with an average length of 100 nt. Then, Bowtie2 (version 2.2.3) was used to map the clean-reads of each sample to the unigene of the reference genome with an e threshold set to 0.00001 using the “-N<1>” parameter (Langmead and Salzberg, 2012), allowing clean-reads to be mapped to the genome with only one mismatch and reads matched to rRNA were removed. The R language package was used to calculate the numbers of reads mapped to each gene (Mortazavi et al., 2008). At the same time, the numbers of reads were normalized to Reads Per Kilo bases per Million reads (RPKM), which are the main data for analyzing gene transcription level.

Statistical analysis was performed on the processed data, and differentially expressed genes (DEGs) were filtered out using the MA-plot-based method with Random Sampling model (MARS) approach in the DEGseq package (Wang et al., 2010). The p-value was obtained from the differential gene expression tests. Benjamini’s False Discovery Ratio (FDR) operation was used to determine p-value thresholds in multiple tests and analyses (Benjamini and Hochberg, 1995). Both FDR ≤ 0.001 and the absolute value of a ≥2-fold change [|log2(RPKM-Δ3hfq/RPKM-BMB171) normalized| ≥ 1] were used as thresholds to identify significant DEGs. The R language package was used to draw a volcano plot for the overall distribution of DEGs between the two samples.

We have analyzed the distribution of Hfq in the genus Bacillus and found that Hfq is widely distributed, present in 78.5% of the species in the genus with fully sequenced genomes (Supplementary Table 2). Through Pfam (Mistry et al., 2021) and NCBI’s Conserved Domain Database (CDD) (Yang et al., 2020), we analyzed the copy numbers of hfq genes in some strains of Bc group with fully sequenced genomes and found that hfq is widely distributed in the Bc group, and the copy numbers of hfq genes are greater than 2 (Supplementary Table 3) with relatively conserved upstream and downstream genes (Vilas-Boas et al., 2007). We defined the Hfq present in most bacteria as the main Hfq, while the other Hfq that also occurs recurrently but less frequently as the secondary Hfq. Intriguingly, the upstream gene of the main Hfq coding gene is usually the coding gene of a hypothetical protein of unknown function, while the downstream gene is usually miaA (encoding tRNA Δ²-isopentenylpyrophosphate transferase) (Vrentas et al., 2015). In contrast, the upstream and downstream genes of secondary hfq typically encode sporulation regulatory protein and HAD superfamily hydrolase, respectively (Supplementary Figure 1).

The starting strain of this study, Bt BMB171, has three copies of hfq gene: BMB171_RS08230 and BMB171_RS18400 are located on the chromosome, while BMB171_RS27540 is located on the only large plasmid pBMB171 (He et al., 2010). We then defined them as hfq1, hfq2, and hfq3 according to their locus numbers, in which, hfq2 is the main hfq gene and hfq1 is the secondary hfq gene. By comparing the amino acid sequences of these three Hfq proteins with that of Bacillus subtilis (Bs) (Someya et al., 2012), we found that their secondary structures all exhibited high similarity and contained two highly conserved core functional motifs, Sm1 and Sm2 (Figure 1A). Next, the transcript levels of the three hfq genes were measured by RT-qPCR at different growth phases, such as at 9 h (logarithmic growth phase), 27 h (early stationary phase), 48 h (late stationary phase or early sporulation phase), and 72 h (final sporulation phase) (Supplementary Figure 2), which showed that hfq1 and hfq2 genes located on the chromosome were well expressed in all four growth phases. Among them, the expression level of the main gene hfq2 was higher than that of hfq1, and the hfq2 transcript was the highest in the logarithmic phase. However, transcription of hfq3 on plasmid pBMB171 was very low in all four growth periods (Figure 1B).

To investigate the regulatory function of Hfq in Bt, we first constructed a series of hfq deletion mutants in combination in BMB171 using a homing endonuclease I-SceI-mediated Markerless gene knockout method, including three hfq single gene mutants (Δhfq1, Δhfq2, and Δhfq3), three hfq double knockout mutants (Δhfq1Δhfq2, Δhfq1Δhfq3, and Δhfq2Δhfq3), and one hfq triple knockout mutant (Δhfq1Δhfq2Δhfq3). We carried out various experiments to determine the phenotypes of these seven mutants, such as their motility (Supplementary Figure 3) and biofilm formation (Supplementary Figure 4), and obtained three mutants with a clear “cumulative effect” in phenotype, named Δ1hfq (Δhfq2), Δ2hfq (Δhfq1Δhfq2), and Δ3hfq (Δhfq1Δhfq2Δhfq3). We then carried out the corresponding motility assays for BMB171 and its mutants Δ1hfq, Δ2hfq, and Δ3hfq. As can be seen in Figure 2A, the motilities of the Δ1hfq, Δ2hfq, and Δ3hfq strains were gradually decreased as the deletion copy numbers of hfq increased (Figure 2B). Since bacterial motility is primarily regulated by flagella, the expression levels of genes involved in flagellar synthesis must be critical for bacterial motility (Mukherjee and Kearns, 2014; Ward et al., 2019). To this end, we measured the transcript levels of flagella-related genes (Supplementary Figure 5) in BMB171 and Δ3hfq and found that the expression levels of fliC and flgE associated with flagellar hooks and flagellar filaments were significantly reduced in Δ3hfq (Figure 2C). Therefore, it is thought that in Bt, Hfq may affect bacterial motility by regulating the transcription and expression of flagella-related genes.

When Bt is applied as an insecticide, it must stay on the plant surfaces to produce insecticidal activity. Improving the ability of bacterial biofilm formation will thus enhance their adhesion to the plant surfaces and ensure the insecticidal performance of Bt (Jha et al., 2021). Therefore, we determined the amount of biofilm formation in Δ1hfq, Δ2hfq, and Δ3hfq. As can be seen in Figure 3A, from BMB171, Δ1hfq, Δ2hfq, to Δ3hfq, the amount of biofilm formation showed a gradient rise with the increase of hfq deletion copy, and the quantitative experimental results also confirmed that this phenotype exhibited a “cumulative effect” (Figure 3B). We also quantified the transcript levels of genes involved in biofilm formation (Supplementary Figure 6) in BMB171 and Δ3hfq using RT-qPCR (Figure 3C), and found that the transcript levels of kinB3, spo0B, and spo0A genes that promote biofilm formation were significantly increased. On the other hand, the transcript level of abrB1, encoding a transcription factor that has been shown to inhibit Bs biofilm formation (Weng et al., 2013; Mielich-Süss and Lopez, 2015), was significantly decreased, which resulted in a significant increase in the amount of biofilm of Δ3hfq. We therefore believe that Hfq plays an important role in regulating the biofilm formation of Bt.

We have used phase-contrast microscope (Figure 4A) and scanning electron microscope (Supplementary Figure 7) to observe the cell morphology of BMB171 and Δ3hfq strains, respectively, and found that the deletion of hfq did not affect the morphology of bacteria. Subsequently, we determined the growth curves of BMB171 and Δ3hfq in LB medium (Figure 4B), and there seems to be no significant difference in the growth rate and growth cycle between the two neither. In nutrient-rich LB medium, the entire growth cycle of BMB171 was approximately 72 h, and the time point to enter the late stationary phase or early sporulation phase was approximately 48 h. We observed sporulation of BMB171 and Δ3hfq cultured for 40, 44, 48, and 72 h, respectively, and found: (a) At 40 h, neither BMB171 nor Δ3hfq had sporulation; (b) At 44 h, BMB171 started to sporulate, but Δ3hfq had no obvious sporulation; (c) At 48 h, the spore numbers formed in BMB171 were significantly more than that in Δ3hfq; (d) By 72 h, the spore numbers formed in BMB171 were similar to that in Δ3hfq (Figure 4A). We counted the spore numbers in BMB171 and Δ3hfq at 48 and 72 h, respectively. As can be seen that at 48 h, the spore numbers in BMB171 were nearly three times higher than that in Δ3hfq with a significant difference; however, by 72 h, they were not significantly different (Figure 4C). Combined with these experimental results, we believe that the deletion of hfq in Bt delays the sporulation, but does not affect the cell morphology or growth rate of Bt.

To explore the effect of Hfq on the insecticidal activity of Bt, we carried out insecticidal activity assays using larvae of the cotton bollworm Helicoverpa armigera that belongs to the family Noctuidae, order Lepidoptera, and is widely distributed in cotton and vegetable growing regions around the world (Lone et al., 2018). The ICP formed by Bt is the main component of its insecticidal activity against Lepidoptera insects (Baum and Malvar, 1995; Wang et al., 2013c). Since the starting strain BMB171 in this study is a mutant that does not produce any ICPs (He et al., 2010), we transferred the ICP encoding gene cry1Ac10 into BMB171 and Δ3hfq, respectively (Wu et al., 2000; Deng et al., 2014), to construct recombinant strains BMB171-cry and Δ3hfq-cry capable of producing ICP. Although both BMB171 and Δ3hfq do not produce ICPs, there are still some toxins, such as enterotoxins, hemolysins, phospholipases, and proteases that affect the normal growth of the host (Bravo et al., 2011). We fed cotton bollworm larvae with a diet containing suspensions of bacteria BMB171, Δ3hfq, BMB171-cry and Δ3hfq-cry individually and observed their larval survival rates, body weights and body lengths 7 days later.

Compared with those fed with suspensions containing BMB171 and BMB171-cry, H. armigera larvae fed with Δ3hfq and Δ3hfq-cry suspensions resulted in significantly reduced survival rates (Figure 5A), decreased body weights (Figures 5B,C), shorter body length (Figures 5B,D), and substantially increased growth inhibition (Figure 5B). These results indicate that Hfq plays a rather important role in regulating the insecticidal activity of Bt. In addition, it can also be seen from Figure 5 that the inhibitory effect of strains BMB171-cry and Δ3hfq-cry on cotton bollworm larvae is even more obvious, which is mainly due to the expression of the Cry1Ac10 ICP.

As a bacterial RNA chaperone, Hfq may also play a key role in bacterial post-transcriptional regulation. To comprehensively investigate the global regulatory role of Hfq on Bt, we carried out RNA-seq on BMB171 and Δ3hfq. Since Δ3hfq has the characteristics of delayed sporulation and enhanced insecticidal activity compared with BMB171 when grown in LB medium for 48 h, samples at this time point were also chosen for RNA-seq to analyze the similarities and differences in post-transcriptional regulation of BMB171 and Δ3hfq. The raw transcriptome data were processed correspondingly and used to product a volcano plot (Supplementary Figure 8). Comparing the two obtained data sets, we found that the mutant Δ3hfq strain had a total of 1730 DEGs (32.5% of the total gene numbers), of which 1414 genes were significantly downregulated and 316 genes significantly upregulated.

Starting from the transcriptome data, we first analyzed the transcription of sporulation-related genes in BMB171 and Δ3hfq. As can be seen from Figures 4A,C, deletion of Hfq delayed BMB171 sporulation, so we focused our analysis on the transcription of genes associated with early sporulation phase. At this stage, the sigma factors SigH, SigF, and SigE are sequentially activated (Kroos et al., 1999; Wang et al., 2013b) to regulate the transcription of other sporulation-related genes. Combined with DBTBS database², we analyzed and identified related genes regulated by SigH, SigF, and SigE in BMB171. At the same time, we also combined transcriptome data to analyze the expression changes of these related genes at the transcriptional level (Supplementary Table 4) and found that: (a) Comparing the transcript levels of 66 genes regulated by SigH in BMB171, 33 genes were downregulated in Δ3hfq, while the transcript levels of the remaining genes exhibited no significant change; (b) 76 genes were regulated by SigF and located in the prospore, of which 43 genes were downregulated in Δ3hfq, and the other 33 genes exhibited no significant changes; (c) 186 genes located in the vegetative cells were regulated by SigE, and in Δ3hfq, 84 genes were downregulated, 94 remained unchanged, and 8 genes were upregulated (Figure 6A).

RNA-sequencing results showed that more than 50% of the sporulation-related genes in Δ3hfq were downregulated, which was consistent with the spore counting results of BMB171 and Δ3hfq (Figure 4C), further confirming that deletion of hfq gene directly caused delayed sporulation. These data indicate that Hfq is an important regulator of early sporulation phase.

In addition, the results of RNA-seq also showed that genes related to the synthesis of virulence factor enterotoxin, such as nheA (encoding non-hemolytic enterotoxin lytic component L2), nheB (encoding non-hemolytic enterotoxin lytic component L2) (Lindbäck et al., 2004), entFM (encoding enterotoxin), entD1, entD2 (encoding enterotoxin/cell-wall binding protein), and hblC (encoding hemolysin BL lytic component L2) (Ngamwongsatit et al., 2008) were all significantly upregulated in Δ3hfq with a log2 fold change of 1.59, 5.82, 1.71, 1.61, 1.1, and 1.5, respectively (Figure 6B). This result well explained that the insecticidal activity of Δ3hfq was higher than that of BMB171, and the insecticidal activity of Δ3hfq-cry was higher than BMB171-cry (Figure 5), and also confirmed that Hfq was an important regulator of bacterial virulence in the Bc group.

Other than validating and explaining the mechanism by which Hfq affects phenotypes such as sporulation and insecticidal activity, we also mined RNA-seq data through pathway enrichment analysis of KEGG and identified the main physiological metabolic pathways and signal transduction pathways involved in DEGs. We used the KEGG Orthology-Based Annotation System (KOBAS) 3.0 Program³ to enrich the known DEGs for the KEGG pathway, with the obtained results shown in the enrichment bubble analysis graph (Figure 6C). In this figure, four pathways, namely, the amino acid metabolism, xenobiotics biodegradation and metabolism, biosynthesis of other secondary metabolites, and transcription were significantly enriched. In addition, enriched pathways with p-values less than 0.05 included lipid metabolism and cell growth and death (Supplementary Table 5 lists all the DEGs enriched in these metabolic pathways). As can be seen from Supplementary Table 5, most of these DEGs were downregulated. Therefore, we hypothesized that Hfq inhibits these metabolic pathways. This result implied that Hfq is mainly involved in the regulation of amino acid metabolism, xenobiotics biodegradation and metabolism, biosynthesis of other secondary metabolites, and transcription.

This study analyzed the copy numbers of hfq genes in Gram-positive Bc group bacteria (Supplementary Table 3), and found that they all contained two or more hfq genes, which is different from most Gram-negative bacteria such as E. coli (Mohanty et al., 2004) and P. aeruginosa (Sonnleitner et al., 2003) that contains only a single hfq gene. In Bt BMB171, there are a total of three hfq genes, and our results showed that the main gene is hfq2, and the deletion of which significantly affected bacterial motility (Supplementary Figure 3) and biofilm formation (Supplementary Figure 4). Furthermore, hfq1 played a relatively smaller role, and hfq3 was generally not transcribed (Figure 1B). When we complemented the hfq2 gene back to the Δ1hfq, Δ2hfq, and Δ3hfq mutants, respectively, we found that complementation of the hfq2 gene did not fully restore the reduced motility in Δ2hfq and Δ3hfq as revealed in the motility phenotype experiments, suggesting that hfq1 and hfq3 in BMB171 also exhibited some other functions (Supplementary Figure 9). We hypothesize from this data that the three hfq genes in BMB171 are all valid copies and can function physiologically under certain specific conditions, for example, in a relatively harsh living environment. If Bt lost its main hfq gene due to external stress, the remaining copies may function to survive external pressure to sustain themselves. This is not only related to the degree of bacterial evolution, but also reflects that Hfq is very important for the entire Bc group bacteria to respond to different external environments.

Our results demonstrated that Bt strains lacking hfq exhibited multiple phenotypic changes, including decreased motility, increased biofilm formation, delayed sporulation, and enhanced insecticidal activity. Through the KEGG Pathway enrichment analysis of DEGs, we found that in Bt, Hfq could participate in the regulation of multiple metabolic pathways, among which amino acid metabolism, xenobiotics biodegradation and metabolism, biosynthesis of other secondary metabolites, and transcription were significantly enriched (Figure 6C), indicating that Hfq can participate in the regulation of these four metabolic pathways in bacteria. In addition, two metabolic pathways, lipid metabolism and cell growth and apoptosis, may also be regulated by Hfq. The above results fully demonstrated that Hfq can play a pivotal role of comprehensively regulating the physiological activities of Bt. Hfq had also been shown to regulate bacterial motility (Schachterle et al., 2019), biofilm formation (Irie et al., 2020), and virulence in some Gram-negative bacteria (Pearl Mizrahi et al., 2021), suggesting that the regulation of certain physiological activities in bacteria by Hfq is universal. However, in Bs, deletion of Hfq was found to affect relatively fewer physiological functions, such as bacterial survival (Rochat et al., 2015), motility, chemotaxis (Jagtap et al., 2016) and cellulolysis etc. (Yang et al., 2021); besides, the deletion of Hfq exhibits no overall effect on the bacterial transcriptome, but only on certain distinct regulons, such as the ResD-ResE signal transduction system required for aerobic and anaerobic respiration, the GerE regulon encoding late spore coat genes and the ComK regulon regulating competence and DNA uptake (Hämmerle et al., 2014). These results suggested that some functions of Hfq are specific and Hfq may regulate different physiological activities in different bacteria even in the same genus.

The above discussion shows that Hfq is universal and specific in regulating bacterial physiological activities. Although we already have a rich understanding about the function of Hfq serving as an RNA chaperone protein, how Hfq regulates these physiological activities by interacting with RNA molecules remains unknown, that is, the molecular mechanism of how Hfq exerts its regulatory role has not yet been fully elucidated. Therefore, we aim to build a complete Hfq regulatory network in Bt in future studies.

This study showed that the insecticidal activity of Δ3hfq against cotton bollworm was significantly enhanced than that of BMB171, indicating that Hfq played a regulatory role on the insecticidal activity of Bt. For many bacterial pathogens, virulence is a very complex and integrative phenotype. In other words, the virulence system is the result of the joint action of many related phenotypes (Raymond et al., 2010; Bravo et al., 2011; Visaggio et al., 2015). There may be indirect or direct relationship between bacterial virulence and other phenotypes, such as cell motility, biofilm formation and sporulation (Josenhans and Suerbaum, 2002; Gallego-Hernandez et al., 2020). In this study, we found that when hfq was deleted, the expression of many enterotoxin genes in bacteria, such as nheA, nheB, entFM, entD1, entD2, and hblC, was significantly increased at the transcriptional level (Figure 6B), which to some extent enhanced bacterial virulence (de Maagd et al., 2003). In addition, the deletion of hfq also led to a significant increase in the amount of bacterial biofilm formation (Figure 3B), and the ability to form biofilms is also considered to be a key virulence factor in a variety of bacteria (Ramírez-Larrota and Eckhard, 2022). Environmental adaptability can help bacteria resist antibiotics and the host’s immune system, which facilitates bacterial survival and infection in the host, ultimately leading to host death (Al Safadi et al., 2012; Giaouris et al., 2015). Based on the above analysis, we believe that the increased insecticidal activity of the Δ3hfq strain may be the result of the combined effect of the increased expression level of enterotoxins and the enhanced biofilm formation ability.

Although how Hfq participates in the regulatory pathways of Bt has not been fully elucidated, our hfq deletion mutant Δ3hfq will provide new research materials for the application of Bt in bio-pesticides, and the effect of Hfq on Bt virulence is also important for increasing our understanding of the Bt insecticidal activity.