A Feed-forward signaling circuit controls bacterial virulence through linking cyclic di-GMP and two mechanistically distinct sRNAs; ArcZ and RsmB

Summary

Dickeya dadantii is a plant pathogen that causes soft rot disease on vegetable and potato crops. To successfully cause infection, this pathogen needs to coordinately modulate the expression of genes encoding several virulence determinants, including plant-cell-wall degrading enzymes (PCWDEs), type III secretion system (T3SS), and flagellar motility. Here, we uncover a novel Feed-forward signaling circuit for controlling virulence. Global RNA chaperone Hfq interacts with an Hfq-dependent sRNA ArcZ and represses the translation of pecT, encoding a LysR-type transcriptional regulator. We demonstrate that the ability of ArcZ to be processed to a 50 nt 3’ end fragment is essential for its regulation of pecT. PecT down-regulates PCWDE and the T3SS by repressing the expression of a global post-transcriptional regulator- (RsmA-) associated sRNA encoding gene rsmB. In addition, we show that the protein levels of two cyclic di-GMP (c-di-GMP) diguanylate cyclases (DGCs), GcpA and GcpL, is repressed by Hfq. Further studies show that both DGCs are essential for the Hfq-mediated post-transcriptional regulation on RsmB. Overall, our report provides new insights into the interplays between ubiquitous signaling transduction systems that were most studied independently and sheds light on multitiered regulatory mechanisms for a precise disease regulation in bacteria.

Introduction

Dickeya dadantii is a plant pathogenic bacterium that causes soft-rot, wilt, and blight diseases on a wide range of economically important crops such as potato, tomato, and chicory (Czajkowski et al., 2011). This bacterium can survive in soil and ground water as an epiphyte or saprophyte. Chemotaxis of D. dadantii is directed toward jasmonic acid produced by wounded plant tissue (Río-Álvarez et al., 2015), and flagellar-based swimming motility enables entry of bacteria into the plant apoplast through natural openings or wounds (Antúnez-Lamas et al., 2009). After penetration, D. dadantii colonizes the apoplast, resulting in a latent infection without provoking disease symptoms (Lebeau et al., 2008). However, when the bacteria encounter favorable conditions, such as high humidity and warm temperatures (30℃), a transition to a pathogenic lifestyle occurs (Reverchon and Nasser, 2013). Soft rot, caused by the disruption of plant cell wall integrity, relies on the secretion of plant-cell-wall degrading enzymes (PCWDE), including pectate lyases (Pels) (Reverchon and Nasser, 2013; Hugouvieux-Cotte-Pattat et al., 2014). Other virulence determinants, such as the type III secretion system (T3SS) (Yang et al., 2002), iron uptake system (Franza and Expert, 2013), and quorum sensing (Nasser et al., 2013), are required for full virulence of D. dadantii.

Small non-coding RNAs (sRNAs), ranging from 50 to 500 nucleotides in length, are important post-transcriptional regulators in bacteria (Gottesman and Storz, 2010). Based on their modes of action and the dependence on the chaperone protein Hfq, sRNAs can be categorized into Hfq-dependent and Hfq-independent sRNAs. Hfq-dependent sRNAs often contain a characteristic stem loop structure at the 3’ end. Nucleotides at the 5’ end can form imperfect basepairing with sequences of the target mRNAs (Vogel and Luisi, 2011). The sRNA-mRNA interaction can result in modulations of mRNA translation, stability, or both (Updegrove et al., 2016). Binding of the RNA chaperone Hfq to the sRNAs enhances the stability of these sRNAs and facilitates the sRNA-mRNA interactions. While Hfq has been well known as a global regulator of many phenotypes, how it exerts these regulatory functions is less clear.

Compared to the Hfq-dependent sRNAs, Hfq-independent sRNAs do not contain the 3’ end Hfq binding structure, and the stability and function of these sRNAs is independent of Hfq. A typical example of this category of sRNA is the CsrB/RsmB sRNA. CsrB/RsmB is a ~350 nt non-coding RNA, which contains 22 potential CsrA/RsmA binding sites that bind to ~9 CsrA/RsmA dimers and antagonize the function of CsrA/RsmA (Babitzke and Romeo, 2007). CsrA, a global translational regulator, is known to facilitate degradation or inhibit translation of target mRNAs. Thus CsrB/RsmB regulates the downstream genes through controlling the amount of active CsrA/RsmA (Liu et al., 1997; Romeo et al., 2013). As these two groups of sRNAs function by unique mechanisms, there is no report suggesting any cross-regulation of these two groups of sRNAs.

In D. dadantii, sRNA RsmB and its counteracting partner RsmA have been characterized as important regulators of the T3SS and PCWDE. RsmA represses the expression of T3SS genes by negatively controlling hrpL (encoding the master regulator of T3SS) at the post-transcriptional level. RsmA also inhibits Pel via an unknown mechanism (Yang et al., 2008). However, the function of Hfq and Hfq-dependent sRNAs in D. dadantii is unknown. Phenotypic studies of hfq deletion mutants in several bacterial species have demonstrated that Hfq is essential for virulence and many other cellular behaviors, such as bacterial growth, stress tolerance, and carbon metabolism (Chao and Vogel, 2010; Zeng et al., 2013; Arce-Rodríguez et al., 2016; Kavita et al., 2018). In this study, we hypothesize that Hfq and Hfq-dependent sRNAs may also contribute to the regulation of bacterial virulence and other virulence related phenotypes (e.g. motility or biofilm formation) in D. dadantii.

In the present study, we performed detailed characterization of the impact of RNA chaperone Hfq and the Hfq-dependent sRNA ArcZ on the expression of virulence and motility genes in D. dadantii. We show that Hfq plays a central role in virulence regulation in D. dadantii by affecting the expression and RNA stability of rsmB, and the intracellular concentration of the bacterial second messenger bis-(3’−5’)-cyclic dimeric guanosine monophosphate (c-di-GMP) (Hengge, 2009; Römling et al., 2013). During this process, the interaction between an Hfq-dependent sRNA, ArcZ, and the mRNA of pecT (encoding a LysR-type transcription regulator) is essential. We also demonstrate that Hfq affects cellular c-di-GMP levels by selectively repressing the production of two GGDEF domain-containing diguanylate cyclases (DGCs), GcpA and GcpL. Findings from this research not only expands our understanding of how the plant pathogenic bacterium D. dadantii utilizes a feed-forward signaling circuit to modulate virulence, but also provides an example of how the RNA chaperone Hfq exerts its global regulation over many phenotypes by controlling other global regulatory systems such as CsrB/RsmB and second messenger c-di-GMP.

Results

Hfq is an essential virulence regulator in D. dadantii

To determine whether the sRNA chaperone Hfq is involved in the virulence regulation of D. dadantii, we generated a deletion mutant of hfq and tested its virulence on Chinese cabbage (Brassica campestris). While the wild-type D. dadantii produced robust lesions and associated maceration on Chinese cabbage leaves, the hfq mutant exhibited an avirulent phenotype (Fig. 1A). To identify the factors regulated by Hfq that affect virulence, we further examined the impacts of hfq deletion on three important virulence determinants: T3SS, PCWDE (Pel production as an example), and swimming motility. Compared to the wild type, expression of three representative T3SS genes hrpA, hrpN, and dspE (encoding a T3SS pilus, a harpin, and a T3 effector, respectively (Wei and Beer, 1995; Chatterjee et al., 2002; Tang et al., 2006)) was reduced by approximately 80% in Δhfq (Fig. 1B). Furthermore, Pel production and swimming motility were reduced by 43% and 53% in Δhfq relative to the wild type (Fig. 1C and D). All observed phenotypes in Δhfq could be complemented by introducing a copy of wild type hfq on a low-copy-number plasmid (Fig. 1B-D). Taken together, these results demonstrate that Hfq is crucial for the pathogenicity of D. dadantii by positively regulating T3SS, PCWDE and swimming motility.

Fig. 1.

Pleiotropic effects of hfq and arcZ deletions in D. dadantii. (A and E) Maceration on the leaves of Chinese cabbage was examined in wild-type D. dadantii, Δhfq, D. dadantii harboring empty vector pCL1920, ΔarcZ harboring pCL1920, and ΔarcZ harboring pCL1920-hfq strains. The expression of T3SS regulon genes (B and F), Pel production (C), and swimming motility (D) were examined in D. dadantii and derivatives. Assays were performed as described in the Methods. All results are from one representative experiment. Three independent experiments were conducted and three replicates (five replicates for maceration assay in plant) were used for each experiment. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (P < 0.05 by Student’s t test).

An Hfq-dependent sRNA ArcZ up-regulates T3SS, Pel, and virulence while down-regulating swimming motility

As an RNA chaperone, the regulatory effect of Hfq is usually mediated through Hfq-dependent sRNAs. The Hfq-dependent sRNA ArcZ has previously been identified as a critical virulence regulator in other bacteria (Zeng and Sundin, 2014). To understand if ArcZ is also an Hfq-dependent sRNA virulence regulator in D. dadantii, we first performed Northern blot to assess the expression of arcZ in wild type and Δhfq. Two transcripts, at sizes ~50 nt and ~80 nt, were detected in wild type D. dadantii but were absent in Δhfq (Fig. S1). The lack of arcZ in Δhfq could be restored by complementation (Fig. S1). This result suggests that ArcZ is indeed an Hfq-dependent sRNA, and exists as multiple transcripts of different sizes.

To determine whether ArcZ controls virulence of D. dadantii, we constructed a deletion mutant of arcZ and compared its virulence with the wild type and the complementation strain. Interestingly, similar to the Δhfq, ΔarcZ also showed significant reduction in lesion size on Chinese cabbage leaf (Fig. 1E). This data suggests that ArcZ may work together with Hfq to promote the virulence of D. dadantii in planta.

To determine whether ArcZ controls virulence via known virulence determinants such as T3SS, Pel, or swimming motility, we compared the activities of these virulence factors in wild type and ΔarcZ. Similar to the Δhfq, deletion of arcZ also drastically decreased hrpA expression and Pel production (Fig. 1C and F). Regarding flagellar motility, however, deletion of arcZ resulted in enhanced flagellar motility in contrast to the reduced motility observed in Δhfq (Fig. 1D). Taken together, these results suggest that the up-regulation of T3SS and Pel production by Hfq is mediated by ArcZ, while ArcZ and Hfq exhibit opposing effects on flagellar motility in D. dadantii.

ArcZ and Hfq positively regulate PecT

To elucidate how Hfq and ArcZ regulate the T3SS and Pel, our first aim was to identify gene targets controlled by Hfq and ArcZ. Hfq was recently shown to inhibit the expression of the PecT homolog RovM in Yersinia enterocolitica (Leskinen et al., 2017). Since PecT is a negative regulator of the T3SS and Pel in D. dadantii (Nasser et al., 2005; Hérault et al., 2014), we speculated Hfq also negatively regulates pecT in D. dadantii. To test this hypothesis, we measured the expression levels of pecT using quantitative real-time PCR (qRT-PCR) in wild type and Δhfq. Deletion of hfq significantly increased pecT mRNA levels and complementation of Δhfq restored pecT expression to near-wild-type levels (Fig. 2A). Furthermore, a significant increase in pecT mRNA level was also observed in ΔarcZ (Fig. 2A). These observations suggest Hfq-mediated repression of pecT expression is coordinated through the ArcZ sRNA.

Fig. 2.

ArcZ controls virulence through PecT-RsmB pathway. (A) pecT RNA levels were examined in wild-type D. dadantii harboring pCL1920, Δhfq harboring pCL1920, Δhfq harboring pCL1920-hfq, ΔarcZ harboring pCL1920, and ΔarcZ harboring pCL1920-arcZ strains using quantitative RT-PCR. The data represent expression levels of each gene relative to that in the wild type, which was mathematically designated as 1. rplU was used as an endogenous control for the calculation. hrpA expression level (B), Pel production (C), and maceration on the leaves of Chinese cabbage (D) were examined in wild-type D. dadantii, ΔarcZ, and ΔarcZΔpecT strains. (E) rsmB expression level was measured in wild-type D. dadantii harboring pCL1920, D. dadantii harboring pCL1920-pecT, ΔarcZ harboring pCL1920, ΔarcZ harboring pCL1920-arcZ, and ΔarcZΔpecT harboring pCL1920 strains. (F) Northern blot analysis of rsmB RNA in wild-type D. dadantii, ΔarcZ, ΔarcZΔpecT, D. dadantii harboring pCL1920, and D. dadantii harboring pCL1920-pecT strains. One representative result of three independent experiments, is presented. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (*P < 0.05; **P < 0.01 by Student’s t-test). Different lowercase letters above the bar indicate statistically significant differences between treatments (Fisher’s LSD, P < 0.05).

ArcZ controls T3SS and Pel mainly through PecT whereas Hfq controls T3SS and Pel through PecT and additional pathways

If regulation of the T3SS and Pel by Hfq and ArcZ occurs mainly through PecT, mutation of pecT in the Δhfq or ΔarcZ backgrounds would fully restore the reduced T3SS expression and Pel production demonstrated in these mutants. As expected, ΔarcZΔpecT exhibited a full restoration of the abolished hrpA expression and a 60% restoration on Pel production in ΔarcZ (Fig. 2B and C), suggesting that PecT might be an essential target of ArcZ through which the regulation of T3SS and Pel occurs. In the ΔhfqΔpecT double mutant, however, the reduced T3SS gene expression and Pel production observed in Δhfq were only slightly restored (Fig. 2B and C). A similar result was also observed when assessing the virulence of these mutants on Chinese cabbage (Fig. 2D). These observations suggest that Hfq controls T3SS and Pel through at least two discrete pathways; one mediated by ArcZ, through the regulation of PecT, and a PecT/ArcZ-independent mechanism.

PecT controls T3SS and Pel through an Hfq-independent sRNA RsmB

Although PecT was identified as a regulator of the T3SS and Pel in D. dadantii (Nasser et al., 2005; Hérault et al., 2014), the detailed regulatory pathway is not clear. In Pectobacterium carotovorum, the PecT homolog HexA represses an Hfq-independent sRNA RsmB through an unknown mechanism (Mukherjee et al., 2000). To determine whether RsmB exerts its regulation downstream of the ArcZ-PecT regulatory pathway in D. dadantii, we compared the expression of rsmB in our wild-type strain to that of a wild-type strain overexpressing pecT. Our results showed that the expression levels of RsmB were significantly reduced in the PecT overexpressing strain compared to the wild-type (Fig. 2E and F), suggesting that PecT represses the expression of RsmB. Since ArcZ repressed pecT (Fig. 2A), it is not surprising to observe that the rsmB expression was decreased by 58% in ΔarcZ (Fig. 2E and F). The reduced RsmB RNA level in ΔarcZ could be restored to the wild-type levels through complementation or through double deletion of arcZ and pecT (Fig. 2E and F). Finally, we confirmed that overexpression of RsmB partially restored the expression of hrpA and Pel production in both ΔarcZ and Δhfq strains (Fig. S2). Taken together, our data indicate that PecT represses the expression of rsmB and this pathway is crucial for ArcZ (Hfq)-mediated virulence regulation in D. dadantii.

ArcZ represses pecT at the translational level and in an Hfq-dependent manner

The repression of PecT by ArcZ could be mediated through the regulation of pecT transcription or through post-transcriptional regulation of the pecT mRNA. To understand through which mechanism ArcZ suppressed pecT expression, we first performed a 5’-rapid amplification of cDNA ends (5’-RACE) assay to determine the transcriptional start site of pecT. The 5’-RACE revealed that PecT mRNA contains a long 5’ untranslated region (UTR), initiating −278 nucleotides upstream of the start codon (Fig. 3A). Based on this information, a transcriptional reporter plasmid (pAT-pecT) containing nucleotides –759 to –279 relative to the pecT start codon was constructed using the promoter reporter plasmid pPROBE-AT (Fig. 3B). We also constructed a translational fusion of the full length pecT 5’ UTR plus the first 30 nucleotides of the pecT coding sequence (CDS) in frame fused with the gfp CDS in the translational fusion plasmid pXG20-sf, in which the transcription of pecT-sfgfp was dependent on an inducible promoter LtetO-1 (Corcoran et al., 2012) (Fig. 3B).

Fig. 3.

Impact of ArcZ on PecT expression at transcriptional and post-transcriptional levels. (A) Sequence of the 5’ UTR of pecT transcript. Capital letters indicate the start of the pecT open reading frame. (B) Schematic drawing of the construction of pAT-pecT and p20-pecT plasmids. Nucleotides are numbered from the +1 of the pecT mRNA. (C) pecT promoter activity was determined in wild-type D. dadantii, ΔarcZ, and ΔarcZΔpecT strains. (D) pecT-sfgfp expression was measured in wild-type D. dadantii harboring pBBR1-MCS4, ΔarcZ harboring pBBR1-MCS4, ΔarcZ harboring pBBR1-arcZ, ΔarcZ harboring pBBR1-arcZ^mut1, ΔarcZ harboring pBBR1-arcZ^mut2, ΔarcZ harboring pBBR1-arcZ^mut3, Δhfq harboring pBBR1-MCS4, and Δhfq harboring pBBR1-arcZ strains. Values are a representative of three independent experiments. Three replicates were used in each experiment. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (ns, P > 0.05; P < 0.05 by Student’s t-test). ns, not significant. (E) The predicted base-pairing region of ArcZ and pecT. The substitution mutations are indicated by arrows and nucleotides. (F) Northern blot analysis of the effects of the wild-type and mutated ArcZ sRNAs in ΔarcZ. Wild-type ArcZ was transcribed from pBBR1-arcZ. ArcZ(CAU_91,92,93GCT), ArcZ(GG_81,82UC) and ArcZ(UAC_68,69,70AUG) were transcribed from pBBR1-arcZ^mut1, pBBR1-arcZ^mut2 and pBBR1-arcZ^mut3, respectively. Each lane was loaded with 1 µg total RNA. Representative blot from three independent replicates is shown.

Unexpectedly, we found that deletion of arcZ significantly reduced the transcription of pecT (Fig. 3C). Since PecT represses its own transcription in D. dadantii (Castillo and Reverchon, 1997), we reasoned that this inhibitory effect could be a consequence of PecT auto-regulation caused by PecT overproduction in the arcZ mutant. The transcription of pecT was fully recovered in the ΔarcZΔpecT double mutant, which confirmed this hypothesis (Fig. 3C). As expected, deletion of arcZ resulted in a 2-fold increase in pecT translation (Fig. 3D). This increase in pecT translation was unaffected in ΔarcZΔpecT (data not shown).

ArcZ-mediated target repression requires Hfq in several bacterial species. Since the presence of Hfq was necessary for ArcZ accumulation in D. dadantii (Fig. S1), we speculated that Hfq is essential for ArcZ-mediated pecT repression. Indeed, we observed that the repressive effect on pecT translation by ArcZ overexpression was abrogated in Δhfq (Fig. 3D) supporting the model that ArcZ-mediated translational repression of pecT is Hfq-dependent.

Identification of critical bases on ArcZ essential for the regulation of pecT

The 3’ half of ArcZ is highly conserved in D. dadantii and other enterobacterial species (Papenfort et al., 2009; Mandin and Gottesman, 2010) (Fig. S3). To understand how ArcZ regulates the translation of pecT, we generated mutations of ArcZ (ArcZ^mut1, ArcZ^mut2 and ArcZ^mut3) within the 3’ region (Fig. 3E) and compared their ability to complement the activated pecT translation in ΔarcZ. Expression of wild type ArcZ and ArcZ^mut3 successfully complemented pecT translation in ΔarcZ restoring it to the wild-type level, while mutation of sequences at ArcZ^mut1 and ArcZ^mut2 sites were incapable of restoring translational repression of pecT (Fig. 3D). The impact of these ArcZ mutations on Pel production was also determined, and the results are consistent with the pecT translation assay (Fig. S4). These data suggest that sequences at ArcZ^mut1 and ArcZ^mut2, but not ArcZ^mut3, are essential for the translational repression of pecT by ArcZ.

We hypothesized that the lack of ArcZ^mut1 and ArcZ^mut2 pecT translational repression by may be the result changes to the sRNA’s native secondary structure. To explore this mechanism, we utilized two RNA secondary structure prediction programs (RNAfold and RNAstructure) to compare the folding of the wild-type ArcZ RNA with the mutant constructs. In support of this hypothesis, both programs modeled ArcZ^mut1 and ArcZ^mut2 with structures which diverged from the predicted wild type ArcZ secondary structure (Fig. S5). In agreement with our experimental observations, the ArcZ^mut3 was predicted to maintain similar secondary features to those found in the wild type ArcZ sRNA (Fig. S5). Because secondary structure often affects RNA processing and stability, we also evaluated the integrity of the three ArcZ RNA mutants using Northern blot. Indeed, complementation of ΔarcZ with either wild type ArcZ and ArcZ^mut3 produced a ~50 nt transcript, which was absent in ΔarcZ strains complemented with either ArcZ^mut1 and ArcZ^mut2 (Fig. 3F). According to previous studies (Papenfort et al., 2009; Chao et al., 2017), this ~50 nt ArcZ might be the processed 3’ half of ArcZ and is likely to be essential for the translational repression of pecT.

To confirm this hypothesis, we constructed vector expressing a truncated ArcZ containing 63 nt from the conserved 3’ end (ArcZ⁶³) (Fig. S3) and determined whether this shorter version of ArcZ is able to restore the ArcZ-mediated pecT repression and Pel production. Indeed, ArcZ⁶³ partially restored both ArcZ-mediated phenotypes (Fig. S6). Taken together, these data suggest that sequences at ArcZ^mut1 and ArcZ^mut2 affect the RNA processing and stability of a functional ArcZ transcript at ~50 nt. It also suggests the 50 nt from the 3’ end of ArcZ are essential for the interaction with the PecT mRNA and the downstream regulation.

We also generated site directed mutations on the pecT 5’ UTR according to predicted RNA-RNA interactions between PecT mRNA and ArcZ (Fig. 3E). However, mutation of the predicted interaction sites did not abolish the translational regulation (Fig. S7).

Hfq downregulates the c-di-GMP level to control swimming motility

As our results (Fig. 2B and C) showed that Hfq controls T3SS and Pel through additional pathways independent of PecT, and c-di-GMP negatively regulates T3SS, Pel, and swimming motility in D. dadantii (Yi et al., 2010; Yuan et al., 2015; Yuan et al., 2018), we further assessed whether Hfq controls T3SS and Pel through c-di-GMP signaling. We first compared the intracellular c-di-GMP levels between wild type and Δhfq using ultra performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS-MS). Compared to the wild type, a 2.3-fold increase in intracellular c-di-GMP levels was detected in Δhfq, and this increase can be complemented by introducing a plasmid copy of hfq (Fig. 4A). This suggests that Hfq plays a role in repressing the intracellular concentration of c-di-GMP in D. dadantii.

Fig. 4.

Hfq represses GcpA and GcpL to regulate swimming motility. (A) Measurement of intracellular c-di-GMP in wild-type D. dadantii harboring pCL1920, Δhfq harboring pCL1920, Δhfq harboring pCL1920-hfq, Δhfq harboring pML122, Δhfq harboring pML122-egcpB, and Δhfq harboring pML122-ecpC strains using UPLC-MS-MS. Swimming motility was examined in wild-type bacteria harboring pML122, Δhfq harboring pML122, Δhfq harboring pML122-egcpB, and Δhfq harboring pML122-ecpC strains (B) and in wild-type D. dadantii, Δhfq, gcpA^D418AΔhfq, ΔgcpLΔhfq, and gcpA^D418AΔgcpLΔpecT (D) strains. One representative experiment was chosen, and three independent experiments with three replicates were performed. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (P < 0.05 by Student’s t test). Mean values of swimming diameters (panel D) labeled with different lowercase letters are significantly different and those labeled with the same lowercase letter are not significantly different (Fisher’s LSD, P < 0.05). (C) Western blot analysis was performed to detect GcpA-HA in gcpA-HA harboring pCL1920, Δhfq gcpA-HA harboring pCL1920, and Δhfq gcpA-HA harboring pCL1920-hfq strains. GcpL-HA protein levels were determined in gcpL-HA harboring pCL1920, Δhfq gcpL-HA harboring pCL1920, and Δhfq gcpL-HA harboring pCL1920-hfq strains. Three independent experiments were performed; a representative blot is shown.

To explore the molecular mechanism of swimming regulation by Hfq, we tested whether the increased concentration of c-di-GMP found in Δhfq was responsible for its previously observed swimming deficiency (Fig. 1D). Overexpression of ecpC, which encodes a phosphodiesterase (PDE) that controls swimming motility (Yi et al., 2010) both significantly reduced c-di-GMP levels and enhanced the swimming motility by 2-fold in the Δhfq mutant (Fig. 4A and B). To rule out that these phenotypes are specifically caused by EcpC, we also examined the effect of overproduction of another PDE, EGcpB, in Δhfq and observed similar changes to both the intracellular concentration of c-di-GMP and swimming motility (Fig. 4A and B). These results suggest that Hfq regulates swimming motility in a c-di-GMP-dependent manner.

Hfq controls c-di-GMP synthesis by repressing two DGCs GcpA and GcpL

We hypothesized that the increased c-di-GMP observed in Δhfq was due to repression of a DGC or DGCs by Hfq. In D. dadantii, a total of 12 genes are annotated as putative DGCs. Aside from one of these enzymes, GcpA (Yuan et al., 2018), the remaining putative DGCs have unknown functions and regulatory mechanisms. We thus determined if any additional annotated DGCs repress flagellar motility in D. dadantii by examining the swimming behavior of individual deletion mutants. While most of these mutants did not significantly affect flagellar motility, one mutant, ΔgcpL, displayed significantly enhanced swimming motility similar to the ΔgcpA active-site mutant (gcpA^D418A) (Fig. S8). This result suggests that both GcpA and GcpL repress swimming motility in D. dadantii. The contribution of GcpL to the intracellular concentration of c-di-GMP was demonstrated by showing that deletion of gcpL decreased the c-di-GMP concentration while overproduction of gcpL in trans significantly increased the c-di-GMP relative to wild type (Fig. S8).

To determine whether Hfq controls c-di-GMP synthesis through these two DGCs, we compared the protein levels of GcpA and GcpL in the wild type and the Δhfq mutant. Inactivation of hfq resulted in a 1.7-fold increase of GcpA-HA and a 3.7-fold increase of GcpL-HA. Complementation of hfq in trans restored the phenotypes to near wild-type levels (Fig. 4C). In contrast, the protein levels of two major PDEs in D. dadantii (Yi et al., 2010), EcpC-HA and EGcpB-HA, were not significantly altered in the presence or absence of hfq (Fig. S9). Additionally, if GcpA or GcpL are truly required for the Hfq-mediated c-di-GMP regulation, inactivation of these DGCs should enhance the swimming phenotype of Δhfq. As expected, the double mutants of gcpA^D418AΔhfq and ΔgcpLΔhfq both partially, but not fully, restored the reduced swimming motility of Δhfq to the wild-type levels (Fig. 4D). Additionally, a triple mutant (gcpA^D418AΔgcpLΔhfq) lacking both DGC-encoding genes was able to enhance swimming motility to above the wild-type levels (Fig. 4D). Collectively, these results implicate that Hfq positively controls swimming motility by repressing two major DGCs GcpA and GcpL.

By reducing cellular c-di-GMP levels, Hfq positively controls RsmB at the post-transcriptional level

In our recent publication, we demonstrated that cellular c-di-GMP negatively controls RsmB at the post-transcriptional level (Yuan et al., 2018). As Hfq reduces the intracellular concentration of c-di-GMP, we further evaluated whether this activity enhances RsmB RNA abundance at the post-transcriptional level. Compared to the wild type, the expression level of rsmB was significantly lower in Δhfq. This could be explained by the Hfq and ArcZ mediated activation of rsmB through repression of PecT, as mutation of pecT in Δhfq restored rsmB expression level (Fig. 5A). Overexpression of the PDE encoding gene ecpC in either Δhfq or ΔhfqΔpecT backgrounds did not alter the expression of rsmB (Fig. S10), suggesting that Hfq-mediated c-di-GMP regulation does not affect rsmB at the level of transcription. However, using Northern blot, we detected a significant increase of RsmB RNA in ΔhfqΔpecT when cellular c-di-GMP was decreased following overexpression of ecpC (Fig. 5B). These results suggest that by reducing the cellular c-di-GMP level, Hfq positively controls RsmB at the post-transcriptional level. Together with the results earlier, we showed that the regulation of Hfq on RsmB occurs via two mechanisms. First, Hfq interacts with sRNA ArcZ and positively controls the expression of RsmB, through a transcriptional regulator PecT. Second, Hfq also enhances the stability of RsmB sRNA through c-di-GMP. The regulation of the T3SS expression and Pel production are in accordance to the regulation of RsmB (Fig. 5C and D) suggesting that PecT regulation of T3SS and Pel is epistatic to c-di-GMP.

Fig. 5.

Hfq controls T3SS and Pel by linking c-di-GMP and ArcZ-PecT pathways. (A) rsmB expression level was determined in wild-type D. dadantii harboring pCL1920, Δhfq harboring pCL1920, Δhfq harboring pCL1920-hfq, and ΔhfqΔpecT harboring pCL1920 strains. (B) RNA levels of RsmB were measured in wild-type D. dadantii harboring pML122, Δhfq harboring pML122, ΔhfqΔpecT harboring pML122, and ΔhfqΔpecT harboring pML122-ecpC strains. Blot is from one representative experiment. Three independent experiments were performed. hrpA expression level (C) and Pel production (D) were measured in wild-type D. dadantii harboring pML122, Δhfq harboring pML122, ΔhfqΔpecT harboring pML122, and ΔhfqΔpecT harboring pML122-ecpC strains. Three independent experiments were performed with three replicates in each experiment. Values are from one representative experiment. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (*P < 0.05; **P < 0.01 by Student’s t-test). ns, not significant.

Additionally, we explored the contributions of GcpA and GcpL to T3SS and Pel in wild type, the hfq mutant, and the hfq/pecT double mutant. Unlike the gcpA deletion that increases both the hrpA expression and Pel production (Yuan et al., 2018), deletion of gcpL in the wild-type background only slightly increased Pel production (Yuan et al., 2018) and had no effect on the expression of hrpA (data not shown). Deletion of either of gcpA or gcpL in the hfq mutant background did not alter T3SS or Pel (Fig. S11). However, deletion of gcpA or gcpL in ΔhfqΔpecT increased the expression level of hrpA by 1.5-fold and 3.6-fold, and Pel production by 1.4-fold and 2-fold, respectively (Fig. S11).

Discussion

In this study, we demonstrate that the RNA chaperone Hfq regulates virulence determinants by affecting both the activity of a global post-transcriptional regulatory system, RsmA/RsmB, and of c-di-GMP signaling via two mechanistically distinct sRNAs; ArcZ and RsmB.

Hfq and Hfq-dependent sRNAs, such as ArcZ, are widespread in many bacterial organisms. Northern blot assay showed three ArcZ transcripts (full-length ~130 nt, ~80 nt, and ~50 nt) in D. dadantii, with the ~50 nt ArcZ as the most abundant length (Fig. 3F, S1, and S12). Prior to this study, it was unclear whether the smaller fragments of ArcZ are functional regulatory RNAs or rather some intermediates during sRNA turnover. Here, we demonstrated the presence of these smaller fragments is essential for the regulatory function of ArcZ. Mutants that fail to produce these fragments could not exert regulatory functions (ArcZ^mut1, ArcZ^mut2); whereas a 5’ end truncated version of ArcZ (63 nt remaining from the 3’ end) is highly functional. Although ArcZ sRNAs possess a conserved 3’ region among enterobacterial species (Fig. S3), similar phenotypes were not observed in Escherichia coli and Salmonella (Papenfort et al., 2009; Mandin and Gottesman, 2010). This suggests the mechanisms for ArcZ cleavage may be different between Enterobacteriaceae.

We also show that three independent mutations of the 3’ region of ArcZ, ArcZ^mut1, ArcZ^mut2 and ArcZ^mut3, altered the predicted secondary structure of ArcZ. ArcZ^mut1 and ArcZ^mut2, but not ArcZ^mut3, greatly changed their predicted secondary structure, were unable to process a functional ~50 nt ArcZ transcript (Fig. 3F and S5). Interestingly, restoration of the secondary structure but not consensus sequence (ArcZ^mut1* and ArcZ^mut2*) of ArcZ (Fig. S5) could not restore its cleavage in vivo (Fig. S12), suggesting that the sequences of ArcZ^mut1 and ArcZ^mut2 sites are essential for processing. Expression of ArcZ^mut1* partially restored the ArcZ-mediated pecT repression (Fig. S6), implying there might be an ArcZ-independent pathway controlling PecT, which is affected by the presence of ArcZ^mut1*. Indeed, the repression on pecT translation was stronger in Δhfq than that in ΔarcZ (Fig. 3D), suggesting that Hfq is likely to control pecT mRNA through ArcZ-independent pathways rather than an ArcZ-dependent pathway alone. ArcZ sRNAs have been shown to control target gene expression cooperatively or competitively with other Hfq-dependent sRNAs (Mandin and Gottesman, 2010; De Lay and Gottesman, 2012).

We conclude that ArcZ positively controls T3SS and Pel by de-repressing rsmB expression through PecT (Fig. 6). ArcZ often exhibits its function by direct base-pairing with the 5’ UTR of target mRNAs. Our data clearly demonstrate that the 5’ UTR of pecT mRNA is important for the ArcZ-mediated translational repression of pecT repression (Fig. 3), although the direct base-pairing remains unclear. ArcZ negatively contributed to swimming motility in D. dadantii (Fig. 1D), which is consistent with the findings from E. coli and Salmonella (Papenfort et al., 2009; De Lay and Gottesman, 2012). ArcZ base-pairs the 5’ UTR of flhDC mRNA, encoding a flagellar master regulator, in E. coli. However, the consensus ArcZ-binding box described by De Lay et al. (De Lay and Gottesman, 2012) was not present in the upstream of D. dadantii flhDC (Fig. S3), suggesting that despite that the overall impacts of ArcZ on swimming motility are similar between D. dadantii and E. coli, the underlying regulatory mechanisms are different.

Fig. 6.

Model of Hfq-mediated regulatory mechanism of virulence factors in D. dadantii. Hfq and its dependent sRNA ArcZ negatively regulate pecT expression by targeting 5’UTR of pecT mRNA. The LysR-type transcription regulator PecT auto-inhibits its own transcription. It is likely to repress the transcription of rsmB, which is essential for the Hfq-ArcZ-mediated T3SS and Pel regulation. sRNA RsmB binds to the global post-transcriptional regulator RsmA and subsequently titrates RsmA activity. RsmA negatively controls T3SS gene expression and Pel production in D. dadantii. Hfq represses c-di-GMP synthesis by down-regulating two DGCs, GcpA and GcpL. These different c-di-GMP signaling pathways are both required for the swimming regulation by Hfq. Furthermore, Hfq up-regulation RsmB expression at the level of post-transcription relies on its inhibitory effects on both GcpA- and GcpL-dependent c-di-GMP signaling pathways. ⊥represents negative control; →represents positive control. The dotted lines indicate regulatory mechanisms identified in this study.

Convergent and divergent c-di-GMP signaling pathways contribute to Hfq-mediated post-transcriptional regulation on rsmB

hfq deletion increased intracellular c-di-GMP levels in D. dadantii (Fig. 4A). A similar phenotype has been reported in the plague pathogen Yersinia pestis, in which Hfq modulates the abundance of a DGC and a PDE (Bellows et al., 2012). D. dadantii encodes eighteen GGDEF and/or EAL domain encoding genes in the genome (Yuan et al., 2018). After individually inactivating each one of these genes, we found that GcpA, EGcpB and EcpC modulate T3SS, Pel and swimming motility (Yi et al., 2010; Yuan et al., 2018), whereas GcpL seems to be involved only in swimming regulation, suggesting a functional specificity in different c-di-GMP signaling pathways. Surprisingly, Hfq selectively repressed the abundance of two DGCs, GcpA and GcpL, and that both c-di-GMP signaling pathways were required for the positive effects of Hfq on swimming motility, T3SS and Pel (Fig. 4D and S11).

Why does D. dadantii require Hfq to control two DGCs that affect the same biological function? Structurally, GcpA is a predicted cytoplasmic protein that contains two types of N-terminal sensory domains, GAF (cGMP phosphodiesterase, Adenyl cyclase, FhlA domain) and PAS (Per/Arnt/Sim), whereas GcpL contains two N-terminal transmembrane domains and a predicted periplasmic CACHE (calcium channel chemotaxis receptor) domain (Yuan et al., 2018) (Fig. S13). Enzymatically, the GGDEF domain of GcpA contains an allosteric c-di-GMP binding site (inhibition site or I-site) that is known to repress the cyclase activity of DGCs (Christen et al., 2006; Yuan et al., 2018). However, this I-site is not present in the GGDEF domain of GcpL (Yuan et al., 2018). In line with this, in trans overproduction of GcpL increased c-di-GMP level by 50-fold compared with that in wild type (Fig. S8), and this increase is much higher than that of a GcpA overproduction strain (Yuan et al., 2018). A recent study reported that exogenous citrate binds to the CACHE-domain of GcbC, a DGC that controls biofilm formation in P. fluorescens, and stimulates its DGC activity (Giacalone et al., 2018). Similar to GcbC, GcpL in D. dadantii also contains a CACHE-domain at its N-terminus in between two transmembrane domains (Fig. S13). Together, we reveal a convergent and divergent c-di-GMP signaling network controlled by Hfq in D. dadantii, and such signaling models are likely to be common in bacteria. Signals triggering all c-di-GMP metabolism, expression and localizations of each individual c-di-GMP-metabolic enzymes need to be further elucidated.

An Hfq-dependent feed-forward signaling controls bacterial virulence

We propose a unique regulatory mechanism for virulence regulation in D. dadantii (Fig. 6). Hfq regulates rsmB via the ArcZ-PecT pathway and this regulation is likely at the level of transcription; Hfq also post-transcriptionally regulates rsmB via c-di-GMP. Interestingly, we did not observe any crosstalk between these two Hfq-mediated pathways. Our results show that inactivation of either pathway alone could not fully recover the rsmB expression and down-stream T3SS expression and Pel production in ∆hfq (Fig. 5 and S11). We also observed that deletion of arcZ had no effect on the intracellular c-di-GMP concentration (data not shown) and manipulations of c-di-GMP concentrations in Δhfq or ΔarcZ did not alter pecT translation (data not shown). Since RsmB functions as a central hub in the Hfq-mediated virulence regulation in D. dadantii (Fig. 6), we propose a novel feed-forward signaling circuit that allows bacteria to control virulence in a most direct, effective, and comprehensive manner.

Hfq and ArcZ were required for optimal bacterial growth in D. dadantii (Fig. S14), raising the possibility that the regulation of virulence factors by Hfq and ArcZ is growth-dependent. However, inactivation of c-di-GMP or ArcZ-PecT-dependent pathway was unable to restore the bacterial growth in Δhfq (Fig. S14), suggesting that this growth defect is not due to the elevated c-di-GMP levels or PecT overproduction. More importantly, these results demonstrate that bacterial growth is likely independent of the swimming, T3SS and Pel phenotypes caused by hfq deletion.

In summary, Hfq, Rsm, and c-di-GMP signaling pathways are universal bacterial signal transduction systems that have been traditionally studied individually. Recent studies have focused on the interplays between these different pathways. An Hfq-dependent sRNA has been shown to bind RsmA to regulate biofilm formation in E. coli (Jørgensen et al., 2013). RsmA has been shown to down-regulate c-di-GMP signaling by repressing DGCs in Pseudomonas aeruginosa and Pseudomonas putida (Moscoso et al., 2014; Huertas-Rosales et al., 2017). Our previously data showed that c-di-GMP post-transcriptionally inhibits RsmB in D. dadantii (Yuan et al., 2018). To our knowledge, this is the first report demonstrating regulatory interactions of all three pathways (Hfq, Rsm, and c-di-GMP) in controlling bacterial virulence and cellular behaviors. Our findings imply a complex regulatory network of multiple signal transduction systems in D. dadantii. This multilevel regulation tightly coordinates the expression of virulence factors to ensure precise control of disease processes, highlighting the importance of the crosstalk between post-transcriptional regulators and sRNAs, transcriptional factors, and bacterial second messengers.

Experimental Procedures

Bacterial strains, plasmids, primers, and media

The bacterial strains and plasmids used in this study are listed in Table S1. E. coli strains were grown in lysogeny broth (LB) medium (10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl) at 37°C. D. dadantii strains were grown in LB medium, mannitol-glutamic acid (MG) medium (10 g/L mannitol, 2 g/L glutamic acid, 0.5 g/L potassium phosphate monobasic, 0.2 g/L NaCl, and 0.2 g/L MgSO₄) or M9 minimal medium (MM) at 28°C. Antibiotics were added to the media at the following concentrations: ampicillin (100 μg/ml), gentamicin (10 μg/ml), kanamycin (50 μg/ml), chloramphenicol (20 μg/ml), and spectinomycin (100 μg/ml). The D. dadantii 3937 genome sequence was retrieved from a systematic annotation package for community analysis of genomes (ASAP) (https://asap.ahabs.wisc.edu/asap/home.php). Primers used for cloning and qPCR are listed in Table S2.

Mutant construction and complementation

The hfq and arcZ deletion mutants were constructed by allelic exchange mutagenesis (Yang et al., 2002). Briefly, upstream and downstream fragments flanking each target gene were amplified by polymerase chain reaction (PCR) using specific primers (Table S2). The kanamycin (Km) cassette was amplified from pKD4 plasmid (Datsenko and Wanner, 2000) and was cloned between two flanking regions using a three-way cross-over PCR. The PCR construct was purified and digested by XhoI and NotI restriction enzymes, followed by insertion into a suicide plasmid pWM91 digested by the same enzymes. The resulting plasmid was transformed into E. coli strain S17–1 λ-pir, which was then conjugated with D. dadantii 3937. Recombinants were plated on 10% sucrose plate supplemented with Km for selection of strains with chromosomal deletions. Sucrose resistant cells due to the loss of SacB-mediated toxicity were plated on ampicillin plate, and the ampicillin sensitive cells were confirmed by PCR using outside primers. Mutations were further confirmed by sequencing.

To remove the Km cassette from the constructed hfq and arcZ deletion mutants, the pFLP2 plasmid encoding the FLP (flipase) recombinase enzyme was transferred into the hfq::Km or arcZ::Km strain by conjugation using E. coli S17–1 λ-pir. The Km cassette was removed by flipase-mediated excision targeting two FLP recombinase target (FRT) sites flanking the Km cassette. Unmarked mutant cells that were sensitive to Km and sucrose were confirmed using outside primers and sequencing. To generate double mutants, gcpA^D418A, gcpL, and pecT were allelic exchanged in an hfq unmarked mutant strain, and pecT was allelic exchanged in an arcZ unmarked mutant strain. The gcpA^D418AΔgcpLΔhfq triple mutant was constructed by allelic exchanging gcpA^D418A in an hfq gcpL unmarked mutant strain.

To generate complemented or overexpression strains, the promoter and open reading frame (ORF) regions of target genes were amplified and cloned into the plasmids pCL1920, pML122 or pBBR1-MCS4 (Table S1). The resulting plasmids were confirmed by sequencing and were electroporated into D. dadantii cells.

Wild-type and mutated ArcZ sRNAs were expressed from pBBR1-MCS4 plasmid. Site-directed mutagenesis was used to generate mutations on wild-type ArcZ. In brief, arcZ fragment was PCR amplified and inserted into a cloning vector pGEM-T Easy. The resulting plasmid was used as a template to generate pGEM-T Easy-arcZ^mut1, pGEM-T Easy-arcZ^mut2, and pGEM-T Easy-arcZ^mut3 plasmids according to the manufacturer’s instructions of the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent, Santa Clara, CA, USA). pGEM-T Easy-arcZ^mut1* and pGEM-T Easy-arcZ^mut2* were constructed using pGEM-T Easy-arcZ^mut1 and pGEM-T Easy-arcZ^mut2 as template, respectively. arcZ fragment and substitutions were retrieved by digestion using XbaI and HindIII restriction enzymes and cloned into pBBR1-MCS4 vector. All constructs were confirmed by sequencing.

Swimming motility assay

Swimming motility assay was performed as described (Antúnez-Lamas et al., 2009). Briefly, 10 µl of overnight bacterial cultures (OD₆₀₀=1.0) were inoculated onto the center of MG plates containing 0.2% agar. The inoculated plates were incubated at 28°C for 16 h. The diameter of the radial growth was measured.

Transcriptional GFP reporter plasmid construction and flow cytometry assay

To generate the transcriptional reporter plasmid pAT-pecT, the promoter region of pecT gene (Fig. 3a) were PCR amplified and cloned into the promoter probe vector pPROBE-AT, which contains a ribosomal binding site upstream of the gfp gene (Miller et al., 2000; Leveau and Lindow, 2001). The reporter plasmids pAT-hrpA, pAT-hrpN, pAT-dspE and pAT-rsmB were constructed following the same procedure (Yang et al., 2007; Li et al., 2015). Bacterial cells containing reporter plasmid pAT-pecT or pAT-rsmB were grown in LB medium and cells containing reporter plasmid pAT-hrpA, pAT-hrpN or pAT-dspE were first grown LB medium and then inoculated 1:100 into MM medium at 28°C for 16 h. Cells were collected by centrifugation, washed with 1× phosphate-buffered saline (PBS) buffer (8.0 g of NaCl, 0.2 g of KCl, 1.44 g of Na₂HPO₄, and 0.24 g of KH₂PO₄ per litter; pH 7.2 to 7.4), and resuspended in PBS at 1 × 10⁶ colony-forming unit (CFU)/ml. The GFP intensities of D. dadantii carrying reporter plasmids were measured by a four-color flow cytometry (FACSCalibur; BD Biosciences, San Jose, CA, USA). GFP fluorescence intensity was measured using the 488-nm laser and the FL1 channel on the gated population. For each sample, 10,000 cells were analyzed, and the mean fluorescence intensity was calculated using CellQuest Pro software (BD Biosciences, San Jose, CA, USA).

Pel activity assay

Extracellular Pel activity was measured by spectrometry as previously described with few modifications (Matsumoto et al., 2003). In brief, cells were cultured in LB medium supplemented with 0.1% polygalacturonic acid (PGA) at 28°C for 16 h. The OD₆₀₀ values of bacterial cultures were measured and normalized to the same value. 1 ml bacterial cultures were centrifuged at 13, 000 g for 2 min, and the supernatant was collected. 10 μl of supernatant was added to 990 μl of the reaction buffer (0.05% PGA, 0.1 M Tris-HCl [pH 8.5], and 0.1 mM CaCl₂, prewarmed to 30°C). Pel activity was monitored at A₂₃₀ for 3 min and calculated based on one unit of Pel activity being equal to an increase of 1 × 10^–3 OD₂₃₀ in 1 min.

Determination of the intracellular c-di-GMP concentration

Intracellular c-di-GMP concentrations were measured using ultra performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS-MS) as previously described (Massie et al., 2012). Briefly, overnight bacterial cultures were inoculated 1:1000 into 50 ml LB medium in a flask and incubated at 28°C until the OD₆₀₀ of bacterial culture reached about 0.8, corresponding to mid- to late-exponential growth. Cell number was determined by plate counting. Bacterial cells were centrifuged in 50 ml polystyrene centrifuge tubes for 30 min at 1500 g, and the supernatant was removed. The pellet was then re-suspended in 1.5 ml extraction buffer (40% acetonitrile and 40% methanol in 0.1 N formic acid), and was left at −20°C for 30 min. After centrifugation for 5 min at 21,000 g to pellet insoluble debris, supernatant containing c-di-GMP was collected and dried by speed-vac. c-di-GMP was resuspended in 100 μl of HPLC grade water, filtered through a Titan syringe filter [polyvinylidene difluoride (PVDF), 0.45 μm, 4 mm] and analyzed by UPLC-MS-MS. Intracellular concentration of c-di-GMP was calculated at 3 × 10⁷ CFU of cells.

Western blot analysis

To measure the protein levels of GcpA-HA, GcpL-HA, EGcpB-HA, and EcpC-HA, wild-type target genes were first allelic exchanged to a haemagglutinin (HA)-tagged version. In brief, a primer set, for example gcpA-A-XhoI-HA and gcpA-B (Table S2), was used to generate a DNA fragment containing the derivative 3’ CDS and downstream fragment of target gene, in which a HA sequence was cloned immediately upstream of the stop codon. The fragment was then cloned upstream of the Km cassette, followed by its downstream fragment, using three-way cross-over PCR. The construct was inserted into pWM91, and the resulting plasmid was transferred into D. dadantii by conjugation using E. coli strain S17–1 λ-pir. The above described allelic exchange mutagenesis was performed to replace wild-type target gene with HA-tagged version. Substitution was confirmed by sequencing.

D. dadantii cells containing the HA tag were grown in LB broth at 28°C for 12 h. Cells were then collected by centrifugation, resuspended in PBS buffer, and lysed by sonication. The protein in crude lysates was quantified using the Bradford protein assay (Bio-Rad, Hercules, CA, USA). Samples were boiled and loaded onto 12% sodium dodecyl sulfate polyacrylamide (SDS-PAGE) gels. Proteins were separated at 120 V constant voltage for 60 minutes and transferred onto a polyvinylidene fluoride membrane (Millipore, Bedford, MA, USA) using a semidry blot machine (Bio-Rad, Hercules, CA, USA) at 5.5 mA/cm² for 60 minutes. Blots were washed with PBS containing 0.1% Tween-20 and probed with an anti-HA antibody (ThermoFisher Scientific, Waltham, MA, USA). Anti-RNA polymerase monoclonal antibody (Neoclone, Madison, WI, USA) was used as a control. The resulting blots were incubated for 1 min in enhanced chemiluminescence reagent (GE Healthcare, Chicago, IL, USA) and detected using O-MAT X-ray film.

Northern blot analysis

To measure the RNA levels of rsmB and arcZ in D. dadantii strains, LB cultures for 16 h were harvested and total RNA was isolated using TRI reagent (Sigma-Aldrich, St Louis, MO, USA). The residual DNA was removed with a Turbo DNA-free DNase kit (Invitrogen, Carlsbad, CA, USA). RNA samples were loaded onto 15% TBE(Tris-borate-EDTA)-Urea gels (Bio-Rad, Hercules, CA, USA). Northern blot analysis was performed using biotin-labeled probe and a biotin detection system (BrightStar Psolaren-Biotin and Bright Star BioDetect, Ambion, Carlsbad, CA, USA). 16S rRNA was used as an internal control.

qRT-PCR analysis

The mRNA level of pecT was measured by qRT-PCR. In brief, D. dadantii cells cultured in LB broth were harvested and total RNAs were extracted using PureLink RNA Mini Kit (Ambion, Carlsbad, CA, USA) according to the manufacturer’s instruction. DNA was removed using On-column DNase treatment (Invitrogen, Carlsbad, CA, USA). cDNA was synthesized using the QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany). The cDNA levels of different samples were quantified by real-time PCR using a PowerUp SYBR Green Master Mix (Life Technologies, Carlsbad, CA, USA). The relative levels of gene expression were determined by using the 2^–∆∆CT method (Livak and Schmittgen, 2001), with the rplU gene used as the internal control (Mah et al., 2003). Three technical replicates were used for each target gene.

Transcriptional start site determination

To determine the transcriptional start site of pecT mRNA, hfq deletion strain was cultured and RNA was extracted as described for qRT-PCR. The transcriptional start site of pecT was determined using the 5’ RACE System for Rapid Amplification of cDNA Ends, Version 2.0 kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instruction using the primers listed in Table S2.

Determination of sRNA-mRNA interactions in vivo

To verify the putative base-pairing between ArcZ and pecT mRNA, a two-plasmid system described by Corcoran et al (Corcoran et al., 2012) to monitor translation efficiency of mRNA-sfGFP fusions was performed with few modifications. ArcZ sRNA and derivatives were expressed from pBBR1-MCS4 plasmid as above described. pecT-sfgfp and pecT^mut1*-sfgfp transcripts expressed from pXG20-sf plasmid were constructed using similar strategies. The entire 5’ UTR (278 nucleotides) plus the first 30 nucleotides of the pecT CDS was digested by BseRI and NheI restriction enzymes and inserted into pXG20-sf, which has been digested by BsgI and NheI restriction enzymes. The resulting constructs were confirmed by sequencing and transferred into D. dadantii cells by electroporation. Cells containing two plasmids were grown in LB medium at 28°C for 16 h. The translation efficiency of mRNA-sfGFP fusions was measured by detecting sfGFP intensity using flow cytometry.

Sequence alignments and RNA secondary structure prediction

The nucleic acid sequence of D. dadantii arcZ and putative 5’ UTR of flhDC mRNA were aligned with other selected homologs using T-Coffee (Notredame et al., 2000), and edited manually. Sequences for alignment analysis were retrieved from ASAP and the National Center for Biotechnology Information (NCBI) Genbank database. Alignments were performed using the European Bioinformatics Institute (EMBL-EBI) webserver (http://www.ebi.ac.uk).

The putative ArcZ secondary structure was predicted using two independent software, RNAfold (http://rna.tbi.univie.ac.at/) and RNAstructure (https://rna.urmc.rochester.edu/index.html), respectively.

Virulence assay

The local leaf maceration assay was performed using Chinese cabbage (B. campestris) leaves as previously described (Yuan et al., 2015). In brief, 10 µl of bacterial suspension at 10⁷ CFU ml⁻¹ were inoculated into the wounds on the leaves punched with a sterile pipette tip. Five leaves were used for each strain. Inoculated Chinese cabbage leaves were kept in growth chamber at 28°C with 100% relative humidity for 16 h before pictures were taken. To evaluate disease symptoms, APS ASSESS 1.0 software (Image Analysis Software for Plant Disease Quantification) was used to determine the leaf maceration areas.

Statistical analysis

Means and standard deviations of experimental results were calculated using Excel and the statistical analysis was performed using a two-tailed student’s t-test (Microsoft, Redmond, WA) or Fisher’s Least Significant Difference (LSD) test using DPS data processing system (http://www.dpsw.cn/dps_eng).

Supplementary Material

Supp TableS1

Table. S1. Strains and plasmids used in this study.

Supp TableS2

Table. S2. Primers used in this study.

Supp figS1-14

Fig. S1. Accumulation of ArcZ sRNA. (A) Northern blot analysis of ArcZ sRNA in wild-type D. dadantii harboring pCL1920, Δhfq harboring pCL1920 and Δhfq harboring pCL1920-hfq strains. Each lane was loaded with 1 µg total RNA. (B) ArcZ sRNA was determined in wild-type bacteria grown at different time points and each lane was loaded with 2.5 µg total RNA. Time points (hours) and OD₆₀₀ values were indicated. The figure represents results from one experiment which was repeated two additional times that demonstrated similar results.

Fig. S2. Overexpression of rsmB increases hrpA expression level and Pel production in Δhfq and ΔarcZ. hrpA expression level (A and C) and Pel production (B and D) of wild-type D. dadantii harboring pCL1920, Δhfq harboring pCL1920, ΔarcZ harboring pCL1920, Δhfq harboring pCL1920-rsmB, and ΔarcZ harboring pCL1920-rsmB strains. Values are a representative of three independent experiments. Three replicates were used in each experiment. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (P < 0.05 by Student’s t test).

Fig. S3. Alignments of enterobacterial arcZ (A) and putative flhDC mRNA 5’UTR (B) sequences. D. dadantii: Dickeya dadantii 3937; D. zeae: Dickeya zeae Ech1591; E. amylovora: Erwinia amylovora ATCC 49946; LT2: Salmonella typhimurium LT2; K12: Escherichia coli K12; C. koseri: Citrobacter koseri ATCC BAA-895; S. flexneri: Shigella flexneri 301; Enterobacter: Enterobacter sp. 638; P. carotovorum: Pectobacterium sp. PC1; S. boydii: Shigella boydii Sb227. The Conserved ArcZ-binding box in the 5’ UTR of flhDC mRNA is shaded in grey. The 63 nt of ArcZ⁶³ is indicated. Asterisks mean identical residues in all sequences in the alignment.

Fig. S4. Impact of ArcZ^mut1, ArcZ^mut2, and ArcZ^mut3 on Pel production in ΔarcZ. Pel production of wild-type D. dadantii and derivative strains were examined. The figure represents results from one experiment which includes three technical replicates. Error bars indicate standard errors of the means. ns represents non-significant (P < 0.05 by Student’s t test).

Fig. S5. Putative secondary structures for ArcZ and ArcZ derivatives. Putative sRNA secondary structures were determined by combining the results of two independent RNA secondary structure prediction programs, RNAfold and RNAstructure.

Fig. S6. Impact of ArcZ^mut1, ArcZ^mut2, and ArcZ⁶³ on pecT translation and Pel production in ΔarcZ. The expression of p20-pecT (A) and Pel production (B) were measured in wild-type D. dadantii harboring pBBR1-MCS4, ΔarcZ harboring pBBR1-MCS4, ΔarcZ harboring pBBR1-arcZ, ΔarcZ harboring pBBR1-arcZ⁶³, ΔarcZ harboring pBBR1-arcZ^mut1*, and ΔarcZ harboring pBBR1-arcZ^mut1* strains. Three independent experiments were performed with three replicates in each experiment. Values are from one representative experiment. Error bars indicate standard errors of the means. Mean values labeled with different lowercase letters are significantly different and those labeled with the same lowercase letter are not significantly different (Fisher’s LSD, P < 0.05).

Fig. S7. Impact of ArcZ and ArcZ^mut1 on pecT^mut1* translation in ΔarcZ. The expression of p20-pecT^mut1* (pecT^mut1*-sfgfp) was measured in ΔarcZ harboring pBBR1-MCS4, ΔarcZ harboring pBBR1-arcZ, and ΔarcZ harboring pBBR1-arcZ^mut1 strains. Three independent experiments were performed with three replicates in each experiment. One representative experiment was chosen, and three independent experiments were performed. Assays were performed as described in Methods. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (P < 0.05 by Student’s t test).

Fig. S8. Two DGCs GcpA and GcpL repress swimming motility. The swimming motility (A) was determined in wild-type D. dadantii and 12 GGDEF-domain-encoding-gene deletion mutant strains. (B) Intracellular c-di-GMP levels were measured in wild-type D. dadantii harboring pML122, ΔgcpL harboring pML122, and D. dadantii harboring pML122-gcpL strains using UPLC-MS-MS. One representative experiment was chosen, and three independent experiments were performed. Error bars indicate standard errors of the means. Asterisks indicate statistically significant differences of the means (P < 0.05 by Student’s t test).

Fig. S9. Impact of Hfq on the abundances of EGcpB-HA and EcpC-HA. Western blot analysis of EGcpB-HA and EcpC-HA proteins in the presence or absence of hfq in the chromosomal egcpB-HA and ecpC-HA strains, respectively. Two independent experiments were performed. Values are from one representative experiment.

Fig. S10. C-di-GMP has no impact on the expression of rsmB. rsmB expression level was measured in wild-type D. dadantii harboring pML122, Δhfq harboring pML122, Δhfq harboring pML122-ecpC, ΔhfqΔpecT harboring pML122, and ΔhfqΔpecT harboring pML122-ecpC strains. Assays were performed as described in Experimental procedures. Three independent experiments with three replicates were performed. One representative experiment was chosen. ns, not significant (ns > 0.05 by Student’s t test)

Fig. S11. Hfq represses both GcpA and GcpL to regulate T3SS and Pel. hrpA expression level (A) and Pel production (B) were measured in wild-type D. dadantii, Δhfq, gcpA^D418AΔhfq, ΔgcpLΔhfq, ΔhfqΔpecT, gcpA^D418AΔhfqΔpecT, and ΔgcpLΔhfqΔpecT strains. Values are a representative of three independent experiments. Three replicates were used in each experiment. Error bars indicate standard errors of the means. ns represents non-significant. Asterisks indicate statistically significant differences of the means (*P < 0.05; **P < 0.01; ***P < 0.001 by Student’s t test). ns, not significant.

Fig. S12. Sequence rather secondary structure is essential for ArcZ cleavage. Northern blot analysis detecting ArcZ was performed. The first lane was loaded with 2.5 µg total RNA from wild-type D. dadantii. All other lanes were loaded with 1 µg total RNA from Δhfq and Δhfq derived strains. One representative blot of two independent experiments was shown.

Fig. S13. Domain structure for GcpA and GcpL. Protein domains were predicted using the Simplified Modular Architecture Research Tool (SMART).

Fig. S14. Growth of wild-type D. dadantii and mutant bacteria. Growth curves were determined by measuring the values of OD₆₀₀ (optical density at 600 nm). Bacterial cells were cultured in lysogeny broth (LB) medium at 28℃. The experiments were repeated twice independently with similar results. Three replicates were used for each experiment. Error bars indicate standard errors of the means.

Originality-Significance Statement.

Bacterial pathogens often encounter different environmental and host conditions and developed a sophisticated regulatory network to modulate their virulence. Dickeya dadantii is a plant pathogenic bacterium that not only can infect a wide range of plant hosts and organs, but also can survive in non-host conditions such as soil, ground water, and vectored in insects. To effectively adapt to these conditions and successfully cause diseases under conducive host conditions, D. dadantii developed a sophisticated regulatory network to modulate its virulence, which often requires the cooperative regulation between multiple signal transduction systems. However, these universal systems, such as regulatory small RNAs and c-di-GMP were most studied independently, and few report suggest that they can regulate each other. Our paper not only identified Hfq as an important virulence regulator, but also demonstrated its central role in controlling multiple regulatory pathways including c-di-GMP and the Csr/Rsm system. Novel targets of Hfq and Hfq regulatory small RNA ArcZ were identified, and the regulatory pathways of important virulence determinants were elucidated. Our report provides new insights into multitiered regulatory mechanisms for a precise disease control in bacteria and facilitates the discovery of virulence inhibitors, such as small molecules, that target these regulators.