Small Noncoding RNA CjNC110 Influences Motility, Autoagglutination, AI-2 Localization, Hydrogen Peroxide Sensitivity, and Chicken Colonization in Campylobacter jejuni

Small noncoding RNAs (ncRNAs) are involved in many important physiological functions in pathogenic microorganisms. Previous studies have identified the presence of noncoding RNAs in the major zoonotic pathogen Campylobacter jejuni; however, few have been functionally characterized to date. CjNC110 is a conserved ncRNA in C. jejuni, located downstream of the luxS gene, which is responsible for the production of the quorum sensing molecule autoinducer-2 (AI-2).

ABSTRACT

Small noncoding RNAs (ncRNAs) are involved in many important physiological functions in pathogenic microorganisms. Previous studies have identified the presence of noncoding RNAs in the major zoonotic pathogen Campylobacter jejuni; however, few have been functionally characterized to date. CjNC110 is a conserved ncRNA in C. jejuni, located downstream of the luxS gene, which is responsible for the production of the quorum sensing molecule autoinducer-2 (AI-2). In this study, we utilized strand specific high-throughput RNAseq to identify potential targets or interactive partners of CjNC110 in a sheep abortion clone of C. jejuni. These data were then utilized to focus further phenotypic evaluation of the role of CjNC110 in motility, autoagglutination, quorum sensing, hydrogen peroxide sensitivity, and chicken colonization in C. jejuni. Inactivation of the CjNC110 ncRNA led to a statistically significant decrease in autoagglutination ability as well as increased motility and hydrogen peroxide sensitivity compared to the wild-type. Extracellular AI-2 detection was decreased in ΔCjNC110; however, intracellular AI-2 accumulation was significantly increased, suggesting a key role of CjNC110 in modulating the transport of AI-2. Notably, ΔCjNC110 also showed a decreased ability to colonize chickens. Complementation of CjNC110 restored all phenotypic changes back to wild-type levels. The collective results of the phenotypic and transcriptomic changes observed in our data provide valuable insights into the pathobiology of C. jejuni sheep abortion clone and strongly suggest that CjNC110 plays an important role in the regulation of energy taxis, flagellar glycosylation, cellular communication via quorum sensing, oxidative stress tolerance, and chicken colonization in this important zoonotic pathogen.

INTRODUCTION

Campylobacter jejuni is a major foodborne and zoonotic pathogen that causes enteritis in humans (1). A sheep abortion (SA) clone, with IA3902 as the prototype strain, has recently emerged as the predominant cause of ovine abortion and as an important pathogen in foodborne outbreaks of human gastroenteritis in the United States (2, 3). C. jejuni clone SA is also widely distributed in the cattle population in the United States. (4). The genome of IA3902 is quite syntenic with that of C. jejuni type strain NCTC 11168, although it is much more virulent than NCTC 11168 in inducing systemic infection and abortion in animals (5). Despite its hypervirulence, C. jejuni IA3902 does not harbor any virulence factors known to be associated with abortion in Campylobacter fetus subsp. fetus (6, 7). Previously, the luxS gene, which mediates autoinducer-2 (AI-2) production, and the genes encoding the capsular polysaccharide have been identified as critical in IA3902 for intestinal colonization and/or translocation across the gut epithelium into the bloodstream (8, 9). Point mutations in the major outer membrane protein encoded by porA of IA3902 have also been shown to be sufficient to cause the abortion phenotype compared to the syntenic nonabortive strain NCTC 11168 (10). The fact that relatively mild changes in genomic sequences have led to significantly enhanced ability to cause disease by C. jejuni IA3902, as described above, suggests that even slight differences in sequence variations or gene regulation may have a major impact on virulence variation among different strains of C. jejuni.

C. jejuni has only three known sigma factors that regulate gene transcription, as follows: σ⁷⁰ (encoded by rpoD), σ⁵⁴ (encoded by rpoN), and σ²⁸ (encoded by fliA) (11). Besides control at the transcriptional level, regulation of gene expression can also occur by posttranscriptional modulation of mRNA translation, stability, and processing, for which small noncoding RNAs are the primary players (12–14). Noncoding RNAs (ncRNAs) can be rapidly produced and serve to regulate multiple different targets within a cell in a variety of ways to coordinate rapid responses to changing environments. Thus, regulation of cellular processes by ncRNAs can provide several advantages to the bacteria compared to the traditional model of protein-mediated regulation (15).

Prior to completion of the transcriptional start site map via high-throughput RNA sequencing (RNAseq) of Helicobacter pylori (16), the Epsilonproteobacteria were thought not to be capable of using small and antisense noncoding RNA as a regulatory mechanism, partly due to a lack of the RNA chaperone protein Hfq (17). Indeed, initial attempts using computational approaches to identify ncRNAs in Campylobacter failed to yield any potential candidates, with only 3 potential loci being identified in Helicobacter (18). Recently, however, clear evidence that C. jejuni also harbors these important regulators has been published, revealing the existence and expression of a wealth of ncRNAs present in strains such as NCTC 11168, 81-176, 81116, RM1221, and IA3902 (19–24). Dugar et al. (21), when comparing the transcriptomes of 4 C. jejuni isolates, observed a large variation in transcriptional start sites as well as expression patterns of both mRNA and ncRNA between strains. This suggests that variations in the existence and expression of ncRNAs even among closely related strains may play a role in differentiating virulence. Despite these advances, identification of the functions of ncRNAs in Campylobacter has been slow to follow, with most lacking both functional characterization and mechanistic investigation.

The first report attempting to elucidate the function of two previously identified noncoding RNAs in C. jejuni suggests that these ncRNAs may play a role in flagellar biosynthesis; however, the authors were unable to demonstrate phenotypic changes following inactivation of either noncoding RNA (25). A later paper, however, did establish the role of an RNA antitoxin (cjrA) as the first noncoding antisense small RNA functionally characterized in Campylobacter to date (26). Just recently, the small RNA pair, CjNC180/CjNC190, has been identified to alter colonization in a newly developed three-dimensional (3D) in vitro model of disease in strain NCTC 11168 (27); the mechanism by which this occurs remains to be fully determined. Thus, beyond simply establishing the existence of noncoding RNA transcripts in Campylobacter, there is a critical need to continue to determine the physiological functions of these potential regulators in this important zoonotic pathogen.

Our previous work has demonstrated the in vivo and in vitro expression of numerous ncRNAs in IA3902 (24), including several that are conserved in other strains of C. jejuni (21). One of the expressed ncRNAs identified in our study, the conserved small RNA CjNC110, is located in the intergenic region immediately downstream of the luxS gene. This ncRNA was of particular interest to our group, as previous work in our lab has already highlighted the importance of the luxS gene in the virulence of C. jejuni IA3902 (8). In Campylobacter, the luxS gene is known to serve two important functions, production of the quorum sensing molecule autoinducer-2 (AI-2) and conversion of S-ribosylhomocysteine to homocysteine in the activated methyl cycle (AMC) (28). Of particular note, following the identification of CjNC110, it was also suggested that different methods of generation of the luxS mutation may have led to polar effects on this downstream ncRNA that may help explain observed differences in phenotypes and gene expression between various studies of luxS mutants in Campylobacter (29).

Based on these observations, in this study we chose to focus on beginning to elucidate the functional role of CjNC110 in the pathobiology of C. jejuni IA3902, as well as investigate the transcriptomic and phenotypic differences between a single mutation of CjNC110 and luxS versus co-mutation of both genomic regions. Using high-throughput RNA sequencing, we successfully demonstrated differences in the transcriptional landscape following mutagenesis of CjNC110 and identified a number of potential mRNA targets for CjNC110 regulation. From these potential targets, we then demonstrated that inactivation of CjNC110 affects several important phenotypes in IA3902, including motility, autoagglutination activity, AI-2 localization, hydrogen peroxide sensitivity, and chicken colonization; our results also demonstrate that these phenotypes consistently differ from those seen with inactivation of luxS. The work presented here provides compelling evidence that expression of the ncRNA CjNC110 is important for the pathobiology of C. jejuni IA3902 and lays the foundation for future work investigating the mechanism of regulation of posttranscriptional gene expression in IA3902 by CjNC110.

RESULTS

Northern blot analysis validates expression of small ncRNA CjNC110 in IA3902 wild-type and mutant constructs ΔluxS and ΔCjNC110c and absence in ΔCjNC110.

Expression of CjNC110 was previously validated via Northern blot analysis in several strains of C. jejuni, including NCTC 11168, RM1221, 81-176, and 81116, and further genetic analysis suggested CjNC110 to be present in IA3902 as well (21). Therefore, to begin to study the role of CjNC110 in IA3902, insertional deletion of the entire predicted coding sequence was utilized to construct ΔCjNC110 in IA3902, and complementation was achieved via reinsertion of CjNC110 into the 16S and 23S rRNA operon (rrs-rrl) of ΔCjNC110 via homologous recombination using plasmid pRRK, creating ΔCjNC110c (Table 1 lists all strains used in this study). To validate expression of CjNC110 in IA3902, an initial Northern blot using the same probe sequence as Dugar et al. (21) and 15 μg of total RNA extracted at the early stationary phase of growth from cultures collected at the same average optical density (A₆₀₀) was performed to compare IA3902 wild-type, ΔCjNC110, and ΔCjNC110c. The results of this Northern blot analysis clearly demonstrate that small ncRNA CjNC110 is expressed by IA3902 wild-type and ΔCjNC110c but not ΔCjNC110 (Fig. 1), validating expression of CjNC110 in IA3902 and confirming successful mutant and complement constructs. Visual inspection of the Northern blot data from both Dugar et al. and the work presented here also demonstrates the presence of multiple bands of differing sizes in both studies (21). All bands were eliminated in the ΔCjNC110 mutant and restored in the ΔCjNC110c complement, which suggests that processing of the CjNC110 transcript may be occurring; further work is needed to investigate this observation.

TABLE 1.

Bacterial strains utilized in this study

Strain	Descriptiona	Source or reference
Campylobacter jejuni
W7	Wild-type motile variant of NCTC 11168	8
W7 ΔCjNC110	W7 ΔCjNC110::CmR	This study
Sheep abortion (SA) IA3902	Wild-type C. jejuni	3
IA3902 ΔCjNC110	IA3902 ΔCjNC110::CmR	This study
IA3902 ΔluxS	IA3902 ΔluxS::KanR	33
IA3902 ΔCjNC110ΔluxS	IA3902 ΔCjNC110::CmR ΔluxS::KanR	This study
IA3902 ΔCjNC110c	IA3902 ΔCjNC110::CmR CjNC110::KanR	This study
Escherichia coli
DH5α	fhuA2 Δ(argF-lacZ)U169 phoA glnV44 Φ80 Δ(lacZ)M15 gyrA96 recA1 relA1 endA1 thi-1 hsdR17	New England Biolabs, Ipswich, MA
Vibrio harveyi
BB152	AI-1-/AI-2+; luxM::Tn5	74
BB170	AI-2 reporter strain; luxN::Tn5	93

^aKan^R, kanamycin resistance cassette; Cm^R, chloramphenicol resistance cassette.

FIG 1.

Northern blotting demonstrates expression of CjNC110 in wild-type (WT) IA3902 and ΔCjNC110c and absence of expression in ΔCjNC110. Cultures for RNA extraction were collected at the early stationary phase of growth, and Northern blot analysis was conducted using 15 μg of total RNA in each of three separate replicates per strain tested. The arrow indicates the most prominent band, which corresponds to the previously predicted size of CjNC110 in C. jejuni (21).

As previous studies have questioned the method of mutant generation of ΔluxS on CjNC110 expression (29), a second Northern blot comparing wild-type IA3902 to a previously constructed ΔluxS mutant utilized by our lab (8) was performed and confirmed that CjNC110 expression in the ΔluxS mutant is also equivalent to wild-type levels (see Fig. S1 in the supplemental material). Homologous recombination was utilized to transfer the CjNC110 mutation into this ΔluxS mutant to create the double-knockout mutant, ΔCjNC110ΔluxS, with the goal of comparing the transcriptomic and phenotypic differences between mutation of ΔluxS or ΔCjNC110 alone versus in combination (ΔCjNC110ΔluxS).

Differential gene expression analysis of RNAseq data suggests several potential regulatory targets for CjNC110.

To begin to identify potential regulatory targets for CjNC110, RNAseq was utilized to compare gene expression changes between ΔCjNC110, ΔluxS, and ΔCjNC110ΔluxS mutants. Total RNA was isolated from an in vitro growth curve performed in triplicate, and colony counts (CFU/ml) over time were used to select samples representative of exponential (3 h) and early stationary (12 h) phases of growth (Fig. S2A). Analysis of variance (ANOVA) did identify a statistically significant difference in growth between strains (P < 0.05); however, multiple-comparison analysis of individual time points and strains compared to wild-type growth revealed that the only significant difference was a decrease in the A₆₀₀ of ΔCjNC110 at 30 h (Fig. S2B), which represents the point where a decline in the A₆₀₀ was noted to occur in all strains tested. Further examination of the actual CFU/ml over the course of the growth curve revealed that by 30 h, cell death was beginning to occur and colony counts were decreasing. As further growth is assumed to have ceased at this time, there is likely minimal effect of this difference on samples collected during growth prior to 24 h; however, there is possibly biological significance to the difference noted during the late stationary phase, which may warrant further investigation. It should be also be noted that the use of CFU/ml more clearly allowed for delineation of actual growth stage of each strain, as the A₆₀₀ results were noted to significantly lag behind the results of actual colony counts; therefore, the CFU/ml data was utilized to determine selection of time points for RNAseq.

Sequencing of the total RNA libraries prepared from the growth curves was performed on an Illumina HiSeq 2500 instrument in high-output single-read mode with 100 cycles, and Rockhopper was utilized to analyze the resulting data for differential gene expression between the mutants (30). Overall, 24 barcoded libraries were sequenced, yielding over 109 million reads, with close to 100 million high-quality reads aligning to either the genome or pVir plasmid of C. jejuni IA3902 and averaging 4,553,847 reads per library (Tables S1 and S2). The vast majority of reads (average of 83% of total reads) mapped to protein-coding genes of the chromosome, with only 7% of reads mapping to rRNA on average following rRNA depletion with Ribo-Zero (median of 3%). Over half of the libraries contained less than or equal to 3% rRNA reads, which is consistent with the manufacturer’s predicted rRNA removal efficiency. One-third of the libraries did not exhibit efficient rRNA removal (>10% rRNA reads); the reason for this is unclear. Despite this difference in efficiency, all samples provided a more than adequate number of non-rRNA reads and thus were successfully utilized in the analysis. Less than 1% of reads mapped to antisense regions of the annotated genome. Rockhopper also identified a number of putative ncRNAs in the data set; Table S3 includes the curated list of identified ncRNA candidates.

Following computational analysis via Rockhopper, a change in gene expression was deemed significant when the q value (false-discovery rate) was <0.05, and a >1.5-fold change in expression was observed. A summary of the differences in numbers of genes with significantly increased and decreased expression in the various strains and time points is given in Table 2, with a listing of specific genes given in Table 3. In the ΔCjNC110 mutant strain, six genes were found to be downregulated and four genes were found to be upregulated compared to the IA3902 wild-type strain during exponential growth (for complete information, see Table S4). In addition, a previously described ncRNA, CjNC140 (21), was found to be upregulated in the mutant condition. During early stationary phase, 16 genes were found to be downregulated and seven genes were found to be upregulated in the mutant strain compared to the wild-type strain. Of the differentially expressed genes, three genes (neuB2, hisF, ptmA) were downregulated in the ΔCjNC110 mutant during both the exponential and stationary phases, and one, CJSA_1261, was found to be upregulated in both conditions. In addition, five separate operons predicted by Rockhopper demonstrated multiple genes within the operon affected by the mutant condition. Analysis of functionality via the Clusters of Orthologous Groups (COG) database revealed that multiple upregulated genes (luxS, cetA, cetB) were present in the “signal transduction mechanisms” category; multiple downregulated genes were included in the “cell wall/membrane biogenesis” (neuB2, ptmB, CJSA_1352) and “posttranslational modification, protein turnover, chaperones” (tpx, CJSA_0687) functional categories. These differentially expressed genes represent potential targets of CjNC110 regulation in IA3902 and thus were used to inform further phenotypic study of the function of this ncRNA. Genes highlighted in bold in Table 3 indicate potential mRNA targets of CjNC110 that were further investigated using phenotypic assays.

TABLE 2.

Summary of differential gene expression results between mutant strains

	Condition
	ΔCjNC110		ΔluxS		ΔCjNC110ΔluxS
	3 h	12 h	3 h	12 h	3 h	12 h
Protein-coding genes
No. downregulated	6	16	1	6	61	32
No. upregulated	4	7	14	6	47	29
Noncoding RNA genes
No. downregulated	0	0	0	0	25b	2
No. upregulated	1a	0	1a	0	1a	0

^aPreviously described ncRNA, CjNC140 (21).

^b17 tRNA genes, 3 known RNA genes (tmRNA, SRP, 6S), and 4 newly predicted ncRNAs.

TABLE 3.

Summary of differential gene expression (>1.5 fold change, q < 0.05) in the IA3902 ΔCjNC110, ΔluxS, and ΔCjNC110ΔluxS mutants as determined by RNAseq and analyzed via Rockhopper

Exponential (3 h)						Stationary (12 h)
ΔCjNC110a	ΔluxSa	ΔCjNC110ΔluxSa				ΔCjNC110a	ΔluxSa	ΔCjNC110ΔluxSa
Genes downregulated
hisF pseA neuB2 ptmA CJSA_1352 rpsN	CJSA_1350	CJSA_0014 CJSA_0041 flgD flgE CJSA_0067 accD trxA panB motB fliN rpsU frdB flaG fliS CJSA_0521 CJSA_0569 pstS hslV flgH flgG2 flgG	mogA aspB CJSA_0788 flgL rpmH CJSA_0920 CJSA_1093 CJSA_1102 dctA luxS CJSA_1182 porA hydD pseB pseC CJSA_1233 neuC2 CJSA_1350 CJSA_1351 flgI CJSA_1388	CJSA_1389 flgK moaE nuoD CJSA_1562 CJSA_1568 CJSA_1577 secY rplO rpsE rplR rplF rpsH rpsN rplE rplX rplN rpmC rplP CJSA_CjSRP1 ssrA	CJSA_t0002 CJSA_t0005 CJSA_t0007 CJSA_t0012 CJSA_t0013 CJSA_t0015 CJSA_t0017 CJSA_t0018 CJSA_t0020 CJSA_t0021 CJSA_t0033 CJSA_t0035 CJSA_t0036 CJSA_t0037 CJSA_t0038 CJSA_t0039 CJSA_t0043 CjNC130	panC CJSA_0559 CJSA_0687 tpx CJSA_0785 ciaB CJSA_1102 petC CJSA_1137 CJSA_1244 hisF neuB2 CJSA_1266 ptmB ptmA CJSAt0002	CJSA_0560 CJSA_0620 CJSA_1349 CJSA_1350 acs leuC	rnhB CJSA_0158 CJSA_0160 CJSA_0284 motB trpD CJSA_0344 CJSA_0372 CJSA_0389 CJSA_0396 sdhA sdhB uxaA CJSA_0560 CJSA_0616 trmD	aroB CJSA_1137 CJSA_1164 pyrH CJSA_1349 CJSA_1350 CJSA_1414 acs CJSA_1461 rnhA CJSA_1560 leuC leuB leuA CJSA_pVir0030
Genes upregulated
CJSA_0008 cetA cetB CJSA_1261 CjNC140	dnaN CJSA_0008 trpF trpB CJSA_0337 CJSA_0370 CJSA_0732 CJSA_1017 CJSA_1131 CJSA_1301 CJSA_1352 CJSA_1449 CJSA_1549 CJSA_pVir0042 CjNC140	dnaN CJSA_0008 gltD folk CJSA_0076 CJSA_0105 CJSA_0110 dgkA pyrC Tal CJSA_0305 trpD	trpF trpB trpA CJSA_0337 CJSA_0370 CJSA_0372 sdhA sdhB CJSA_0490 CJSA_0491 CJSA_0836 Cfa	cetB cetA CJSA_1129 CJSA_1131 CJSA_1145 CJSA_1164 cbpA CJSA_1259 CJSA_1260 CJSA_1301 CJSA_1343 tagF	CJSA_1449 rloH nuoM CJSA_1549 leuC leuB leuA CjNC140 CJSA_pVir0042 CJSA_pVir0044 CJSA_pVir0025	hcrA CJSA_0716 CJSA_0717 CJSA_1107 omp50 luxS CJSA_1261	flgE flgG2 flgG hcrA CJSA_1107 nrfA	dsbI CJSA_0040 CJSA_0041 flgD flgE flgC flgB flgH flgG2 flgG hcrA CJSA_0716 flgS flgL CJSA_0969	CJSA_1107 omp50 CJSA_1180 pseB CJSA_1262 neuB2 neuC2 flgI CJSA_1387 flgK CJSA_1562 p19 CJSA_pVir0042 CJSA_pVir0013

^aAdditional small RNAs predicted by Rockhopper present with differential expression; see Tables S4, S5, and S6 for additional information. Bold indicates a gene for which further phenotypic investigation of the functional significance of the observed differential expression was performed (pseA, neuB2 (ptmC), CJSA_1261 (ptmG), ptmA, ptmB = autoagglutination; cetA, cetB = motility; tpx = hydrogen peroxide sensitivity; and luxS = AI-2 assay).

ΔluxS displays minimal alterations in gene expression via RNAseq, with more significant gene expression changes observed in ΔCjNC110ΔluxS.

Differential gene expression analysis via RNAseq and Rockhopper was also performed for ΔluxS and ΔCjNC110ΔluxS and compared to wild-type IA3902 to assess the difference between a single mutation of luxS and concurrent mutation of both CjNC110 and luxS. In the ΔluxS mutant, a surprisingly small number of genes were noted to be differentially expressed. One gene was found to be downregulated, and 14 genes were found to be upregulated compared to IA3902 wild-type during exponential growth (for summary, see Table 3; for complete information, see Table S5). Similar to the ΔCjNC110 mutant, the previously described ncRNA CjNC140 was also found to be upregulated in the mutant condition during exponential growth only. During early stationary phase, six genes were found to be downregulated and six upregulated in the ΔluxS mutant strain compared to wild-type. Of the differentially expressed genes, only one gene, CJSA_1350, a putative methyltransferase, was found to be downregulated in the ΔluxS mutant during both exponential and stationary phases. Three separate operons predicted by Rockhopper demonstrated multiple genes within the operon affected by the mutant condition (two operons upregulated, one operon downregulated). Analysis of functionality via the COG database revealed that multiple upregulated (trpF, trpB) and downregulated (CJSA_0620, leuC) genes were present in the “amino acid transport and metabolism” functional category; multiple upregulated genes were also included in the “cell motility” (flgE, flgG2, flgG) functional category. Only two genes were differentially expressed in both the ΔCjNC110 and ΔluxS mutants, CJSA_0008 (upregulated during exponential phase) and CJSA_1107 (upregulated during stationary phase); both genes are annotated as hypothetical proteins at this time. Despite the critical role that LuxS plays in the AMC pathway, significant effects on the expression of other genes within the AMC pathway were not observed.

In the ΔCjNC110ΔluxS double-knockout mutant, however, a large increase in both downregulated and upregulated genes compared to the wild-type was observed at both time points (for summary, see Table 3; for complete information, see Table S6). Compared to the IA3902 wild-type, 61 protein coding genes, 17 tRNA genes, 3 known ncRNA genes (transfer-messenger RNA [tmRNA], signal recognition particle [SRP], 6S), and 7 newly predicted ncRNAs were downregulated during exponential phase. In addition, 47 protein-coding genes were upregulated in exponential phase, along with the previously described ncRNA CjNC140, which was also observed to be upregulated in the ΔCjNC110 and ΔluxS single-knockout mutants. During early stationary phase, 32 genes and two newly predicted ncRNAs were found to be downregulated, while 29 genes were upregulated in the mutant strain compared to the wild-type. Of the observed genes with differential expression, many were observed to show altered expression during both exponential and stationary phase compared to the wild-type; however, most were differentially expressed in the opposite direction between the two time points. Only two genes (motB, CJSA_1350) were downregulated in both conditions, while no genes were observed to be upregulated at both time points. Many of the differentially upregulated genes demonstrated in both background mutants, ΔCjNC110 and ΔluxS, were observed in the double-knockout mutant (Fig. 2A and B), while fewer of the downregulated genes were shared (Fig. 2C and D).

FIG 2.

Venn diagram depicts overlap of shared up- and downregulated genes during the exponential and stationary growth phases of all 3 mutant strains. BioVenn was used to compare the lists of known genes identified by Rockhopper as upregulated in the RNAseq data in all 3 mutant strains during both exponential (A) and stationary (B) growth; downregulated genes were also compared for all 3 mutant strains during both exponential (C) and stationary (D) growth.

Of the differentially expressed genes affected by the ΔCjNC110ΔluxS double-knockout mutation, 20 separate operons predicted by Rockhopper demonstrated multiple genes within the operon affected by the mutant condition (8 operons upregulated, 4 operons downregulated, and 8 operons that showed opposing regulation at the different time points). Figure S3 shows the number of genes in each COG category that were differentially expressed in the double-mutant condition during exponential growth. The categories most affected by these mutations were “energy production and conversion,” “amino acid transport and metabolism,” “translation,” “cell wall/membrane biogenesis,” “cell motility,” and “signal transduction mechanisms.” Using KEGG pathways (31), a significant effect was observed in the flagellar assembly pathway in the double-knockout mutant compared to either mutant alone (Fig. S4 and S5). Of particular interest, this affect was also observed to be altered based on growth phase, with decreased expression of σ⁵⁴-associated genes noted during exponential phase and increased expression noted during stationary phase.

Phenotypic evaluation of increased luxS expression unexpectedly reveals alteration of AI-2 transport in ΔCjNC110.

Utilizing the potential targets identified in the RNAseq data for ΔCjNC110, phenotypic testing was next performed to begin to determine the physiologic role of CjNC110 in IA3902. Further investigation of a potential interaction with luxS, which was identified as demonstrating significantly increased expression in ΔCjNC110 during stationary phase, was initially pursued due to the proximity of the two genes. LuxS activity is traditionally measured using the Vibrio harveyi bioluminescence activity assay as an approximation of AI-2 production (32). Using this assay as previously described, AI-2 levels were initially evaluated within the extracellular environment by collecting cell-free supernatant (E-CFS) at various time points during the growth of IA3902 wild-type and mutant strains. To collect these samples, a second set of growth curves independent of those utilized for RNAseq was performed in triplicate to include the ΔCjNC110c (Fig. 3). Similar to the first growth curves utilized for RNAseq, ANOVA again identified a statistically significant difference between the growth of the strains (P < 0.05); however, the only significant difference was a decrease in the A₆₀₀ of ΔCjNC110 after 27 h, similar to the difference seen in the first set of growth curves. This defect was corrected in the ΔCjNC110c complement in the second set of growth curves, indicating that while the biological significance is unknown, this observation is repeatable and related to the loss of CjNC110.

FIG 3.

Shaking growth curve of wild-type IA3902 (WT) and isogenic mutants (mean ± standard error of the mean [SEM]) confirms normal exponential growth. Results of four replicates (A₆₀₀) of a shaking growth curve performed in 250-ml Erlenmeyer flasks under microaerophilic conditions in MH broth. Samples were pulled from this growth curve and used for qRT-PCR and AI-2 bioluminescence assays; for the results of the growth curve from which the RNAseq samples were taken, please see Fig. S2A and B. Significance is denoted by an asterisk (*) above each time point.

As expected based on previous studies (8, 33), the results of the V. harveyi bioluminescence assay demonstrated that both the ΔluxS and ΔCjNC110ΔluxS mutant strains displayed no bioluminescence activity at any point during growth, indicating a complete lack of AI-2 production, which is consistent with the absence of a functional LuxS protein (Fig. 4A). For the ΔCjNC110 mutant, however, bioluminescence was determined to be statistically significantly decreased compared to the wild-type at time points 6, 9, and 12 h (P < 0.05) of the growth curves, occurring during mid to late exponential phase and early stationary phase. This is in distinct contrast to the expected result based on RNAseq analysis, which demonstrated increased luxS mRNA expression by stationary phase. Complementation of ΔCjNC110 completely restored extracellular AI-2 to wild-type levels, strongly suggesting that the observed defect in extracellular AI-2 was a true phenotype warranting further exploration.

FIG 4.

Bioluminescence activity as measured via the Vibrio harveyi bioassay over the course of growth demonstrates (A) decreased extracellular AI-2 levels with (B) a concomitant increase in intracellular AI-2 in ΔCjNC110 compared to wild-type IA3902 (WT) (mean ± SEM). The V. harveyi bioluminescence assay was used initially to detect AI-2 levels in the extracellular CFS to approximate LuxS activity (A). A second experiment was performed to investigate AI-2 levels in the intracellular CFS (B). Each bar represents the average relative light units (RLU) (AI-2 activity level) from three biological replicates consisting of three technical replicates each. Significance is denoted by an asterisk (*) above each time point.

As the RNAseq results (i.e., increased luxS gene expression) did not match the results of the AI-2 assay (i.e.,- decreased extracellular AI-2), reverse transcription-quantitative PCR (qRT-PCR) was performed on samples collected at 12 h from the same growth curve as the AI-2 assay using targets specific for the 3′ region of the luxS gene. Figure 5 demonstrates the expression levels of luxS in the ΔCjNC110 mutant compared to the wild-type and confirms that increased luxS mRNA levels are indeed present during early stationary phase in the mutant, resulting in a fold change increase of 3.20 compared to IA3902 wild-type (P < 0.05). Additionally, ΔCjNC110c was tested to determine if complementation of the mutation returned luxS to wild-type levels; interestingly, ΔCjNC110c also demonstrated a fold change increase of 3.77 compared to IA3902 wild-type (P < 0.05).

FIG 5.

qRT-PCR confirms that luxS is transcribed at higher levels in the ΔCjNC110 and ΔCjNC110c backgrounds compared to the IA3902 wild-type (WT) (mean ± SEM). C. jejuni cultures were collected at 12 h for RNA conversion to cDNA for qRT-PCR to assess expression of mRNA luxS. Each bar represents the average fold change expression from three biological replicates. Significance is denoted by an asterisk (*) above each strain.

Proteomic analysis using liquid chromatography with tandem mass spectrometry (LC-MS/MS) was then conducted to determine if the relative abundance of the LuxS protein was also increased in ΔCjNC110. Whole-protein-extract LC-MS/MS analysis resulted in a total of 262 identified proteins that were discovered in each biological group. Two-way ANOVA with FDR correction was used to identify statistically significant differences in protein abundance; these statistically significant proteins are listed in Table 4. In both ΔCjNC110 and ΔCjNC110c, the abundance of LuxS was noted to be statistically significantly increased (P < 0.0001). Hierarchical heat maps were also constructed to further investigate protein expression patterns when comparing IA3902 wild-type to ΔCjNC110 and ΔCjNC110c. The top 20 hierarchical clustered proteins are illustrated for both ΔCjNC110 and ΔCjNC110c compared to wild-type IA3902 in Fig. S6; LuxS was also identified via this approach as being differentially expressed in both ΔCjNC110 and ΔCjNC110c. Therefore, the proteomics data confirmed the RNAseq and qRT-PCR observations of significantly increased expression of LuxS when comparing the wild-type to both ΔCjNC110 and ΔCjNC110c.

TABLE 4.

Significant differentially expressed proteins identified from proteomic analysis ΔCjNC110 and ΔCjNC110c versus IA3902 wild-type (WT)

Protein ID	Geometric mean (abundance)a		Fold change vs WT	Q-valueb
	WT	Mutant
ΔCjNC110
CJSA_0067 - iron-sulfur protein	23726566	851708	-27.9	<0.0001
CJSA_0489 - rhodanese-like domain-containing	35962750	2965821	-12.1	0.0001
ThiJ - thiazole monophosphate synthesis protein	1589344	31307392	19.7	<0.0001
FrdA - fumarate reductase flavoprotein subunit	3178688	280959	-11.3	0.0002
LuxS - S-ribosylhomocysteine lyase	3913424	50859008	13.0	<0.0001
ΔCjNC110c
CJSA_0067 - iron-sulfur protein	23726566	1123836	-21.1	<0.0001
CJSA_0489 - rhodanese-like domain-containing	35962750	2767209	-13.0	0.0001
CJSA_0166 - lipoprotein	25429504	2581897	-9.8	0.0006
RplN - 50S ribosomal protein L14	10327588	1048576	-9.8	0.0007
RplU - 50S ribosomal protein L21	15653696	3178688	-4.9	0.0007
RecJ - single-stranded-DNA-specific exonuclease	2408995	19271960	8.0	0.002
AtpF - ATP synthase subunit B	2965821	27254668	9.2	0.001
LuxS - S-ribosylhomocysteine lyase	3913424	54509336	13.9	<0.0001
CJSA_0926 - lipoprotein	1703417	23726566	13.9	0.0001

^aGeometric mean calculated by back-transforming log₂ transformation value.

^bTwo-way ANOVA FDR set at 0.05.

As decreased LuxS abundance was not the cause of the observed decrease in extracellular AI-2, several additional hypotheses were proposed, including (i) decreased activity of the LuxS protein leading to decreased AI-2 production, (ii) increased degradation of AI-2, or (iii) altered AI-2 transport into or out of the extracellular environment. To assess for altered AI-2 transport, the traditional V. harveyi bioluminescence activity assay was modified as described in Materials and Methods to assess for intracellular AI-2 levels using intracellular cell-free supernatant (I-CFS) obtained from the cell pellet of the same growth curve sample as the E-CFS. Results of this assay are shown in Fig. 4B and clearly demonstrate that intracellular AI-2 levels increased to statistically significant (P < 0.05) levels in the ΔCjNC110 mutant compared to IA3902 wild-type at both time points tested. Again, complementation of ΔCjNC110 corrected the phenotypic alteration of increased intracellular AI-2 presence to wild-type. When taken together, these results indicate that the observed decrease in extracellular AI-2 activity is not due to impaired activity of LuxS but, instead, strongly suggests that transport of AI-2 across the cell membrane is altered in the ΔCjNC110 mutant, leading to decreased extracellular and increased intracellular AI-2 accumulation.

Mutation of CjNC110 significantly increases motility and hydrogen peroxide sensitivity while decreasing autoagglutination in IA3902, opposite of the effects seen with mutation of luxS.

To further investigate the role of CjNC110 in IA3902, several additional phenotypes suggested by the RNAseq results were assessed, including motility, hydrogen peroxide sensitivity, and autoagglutination. Loss of cetA and cetB has been previously shown to affect energy taxis and motility (34); therefore, the motility of the mutant strains in semi-solid agar was initially compared to wild-type IA3902 at 30 h post-inoculation. The results of this assay confirmed that all isolates were highly motile, and statistical analysis via one-way ANOVA indicated that there was a significant difference between strains (P < 0.0001) (Fig. 6). Further post hoc analysis via Tukey’s multiple-comparison test did not reach statistical significance (P > 0.05) when comparing ΔCjNC110 to the wild-type despite motility for the ΔCjNC110 strain being observed to be consistently increased above the wild-type phenotype in all biological replicates performed. Results did reach statistical significance (P < 0.05), however, when comparing the wild-type to both the ΔCjNC110ΔluxS and ΔluxS mutants, which displayed statistically significant decreased motility; this result is consistent with previous reports of decreased motility of luxS mutants in the IA3902 background (8). As many of the strains had migrated to the edge of the plate by 30 h, a second set of experiments was performed with measurements taken at 24 h post-inoculation. When measured at 24 h, motility for the ΔCjNC110 strain did demonstrate a statistically significant increase (P < 0.05) compared to all other strains tested, and motility of both the ΔCjNC110ΔluxS and ΔluxS mutants was again noted to be significant decreased (Fig. 6). Following complementation of CjNC110, no significant difference was observed compared to IA3902 wild-type in either experiment, confirming that complementation of CjNC110 quantitatively restored the phenotypic change back to wild-type levels. These results indicate an increased motility of ΔCjNC110, which differs from ΔluxS, which displays decreased motility in IA3902.

FIG 6.

Motility assays confirm that isogenic mutants remain motile, with ΔCjNC110 motility increased and ΔluxS motility decreased compared to the IA3902 wild-type (WT) (mean ± SEM). Swarming motility in semisolid agar was assessed via measurement in mm of the outermost reach of growth at both 30 h and 24 h. Each bar represents the average motility of six biological replicates from six independent studies. Significance is denoted by an asterisk (*) above each bar when comparing the wild-type to other strains at each independent time point.

Based on the potential regulatory interaction identified in our RNAseq data of CjNC110 with tpx, which was previously demonstrated to encode a dedicated hydrogen peroxide reductase in Campylobacter spp., a hydrogen peroxide (H₂O₂) disk inhibition assay was performed as previously described (35). Fig. 7 demonstrates the results of this assay, which indicate that the lack of CjNC110 significantly impacts H₂O₂ sensitivity compared to the wild-type, as increased sensitivity was observed in both CjNC110 mutant backgrounds (ΔCjNC110 and ΔCjNC110ΔluxS) (P < 0.05). Complementation of CjNC110 led to a significantly decreased sensitivity to H₂O₂ compared to the wild-type, indicating that the potential overexpression of CjNC110 in ΔCjNC110c may lead to overcorrection of the phenotypic change (P < 0.05). There was no significant difference (P > 0.05) noted between the wild-type and ΔluxS, which is consistent with previous studies of luxS mutation in other strains of C. jejuni such as NCTC 11168 (36). These results indicate that CjNC110 may play a key role in sensing and responding to oxidative stress in the environment for IA3902 and again represent a differing phenotype between ΔCjNC110 and ΔluxS.

FIG 7.

ΔCjNC110 displays increased hydrogen peroxide (H₂O₂) sensitivity in vitro compared to wild-type IA3902 (WT) (mean ± SEM). Cell cultures normalized to an A₆₀₀ of 1.0 were inoculated into melted MH agar and then incubated with a 6-mm disk soaked in 3% H₂O₂ placed at the center of the solidified MH agar plate; after 24 h, the zone of sensitivity to H₂O₂ was measured in mm. Each bar represents the average of three biological replicates from four independent studies. Significance is denoted by an asterisk (*) above each bar when comparing the wild-type to other strains.

Differential expression of several genes associated with flagellar modification (pseA, ptmA, ptmB, neuB2, and CJSA_1261) was also observed in ΔCjNC110; these genes have been previously shown to affect the autoagglutination ability of C. jejuni (37–39). Therefore, autoagglutination was also assessed as previously described (40) to determine if the observed gene expression changes in ΔCjNC110 led to a phenotypically observable alteration. Autoagglutination activity was measured at both 25°C and 37°C following 24 h of incubation, with an increase in optical density correlating with decreased autoagglutination ability and a decrease in optical density correlating with increased autoagglutination ability. A statistically significant difference (P < 0.0001) between strains at both 25°C and 37°C was noted based on initial analysis via one-way ANOVA (Fig. 8). Compared to IA3902 wild-type, ΔCjNC110 exhibited statistically significant decreased autoagglutination activity (P < 0.05) at both temperatures, while autoagglutination activity for ΔluxS and ΔCjNC110ΔluxS was noted to be increased at a statistically significant level (P < 0.05), again at both temperatures. Complementation of CjNC110 restored the autoagglutination ability of ΔCjNC110 to wild-type levels. These results indicate that luxS and CjNC110 function to influence autoagglutination activity in opposing directions; interestingly, when mutation of both genes was combined, the results favored the ΔluxS mutation but did return closer to wild-type levels.

FIG 8.

The autoagglutination ability of IA3902 is decreased in ΔCjNC110 and increased in ΔluxS compared to wild-type IA3902 (WT) (mean ± SEM). Autoagglutination as determined by optical density (A₆₀₀) was measured following a 24-h incubation at either 25°C or 37°C. Each bar represents the average autoagglutination of 3 biological replicates consisting of 4 technical replicates during each independent study. An increase in optical density correlates to decreased autoagglutination ability, and a decrease in optical density correlates to increased autoagglutination ability. Significance is denoted by an asterisk (*) above each bar when comparing the wild-type to other strains at each temperature.

Sustained colonization of chickens is impaired in ΔCjNC110 and absent in mutants ΔCjNC110ΔluxS and ΔluxS but is restored to wild-type colonization levels in ΔCjNC110c.

The ability to colonize the host is the critical first step in the establishment of infection by Campylobacter spp. Previous studies have demonstrated a loss of colonization ability of ΔluxS in IA3902 (8); therefore, a chicken colonization model was utilized to determine if a similar phenotype was present in either ΔCjNC110 or ΔCjNC110ΔluxS. Three-day old chicks were orally inoculated with approximately 200 μl of 10⁷ CFU per strain (wild-type, ΔCjNC110, ΔluxS, ΔCjNC110ΔluxS, and ΔCjNC110c). Once weekly for 3 weeks following inoculation, 6 chicks were humanely euthanized and cecal contents collected for analysis of colonization levels via plating of serial dilutions to determine colony counts. For ΔCjNC110, while colonization was still present at all time points, a statistically significant decrease in colonization was noted when compared to the wild-type on days post-inoculation (DPI) 5, 12, and 19 (P < 0.05); complementation of the ΔCjNC110 mutant restored colonization to wild-type levels at all time points analyzed (Fig. 9). Similar to previous studies, while some colonization, albeit significantly decreased, was noted for ΔluxS at DPI 5, no colonization was observed by DPI 12 and 19 (P < 0.05). For the ΔCjNC110ΔluxS mutant, normal colonization levels compared to wild-type were observed on DPI 5; however, by DPI 12 and 19, no colonization of ΔCjNC110ΔluxS was detectable (P < 0.05). These results indicate that the presence of CjNC110 is necessary for optimal colonization, while luxS is required for establishing sustained colonization of IA3902 in chickens.

FIG 9.

Sustained colonization in chickens is impaired in ΔCjNC110 and absent in ΔluxS and ΔCjNC110ΔluxS mutants compared to the wild-type (mean ± SEM). Results of a single study investigating the colonization ability of mutant strains compared to wild-type (WT) IA3902 are reported as log₁₀ CFU/ml of ceca content for each sampling day (DPI 5 [week 1], DPI 12 [week 2], and DPI 19 [week 3]). Each bar represents the average colonization levels for each biological group with weekly technical replicates using a minimum of 6 birds. Significance is denoted by an asterisk (*) above each time point.

DISCUSSION

In other species of bacteria that harbor known RNA chaperones such as Hfq, methods of identification of cognate mRNA binding partners of ncRNAs are available via several straightforward approaches (recently reviewed in reference 41). As Campylobacter spp. lack all known RNA chaperones, identification of targets via these methods is not feasible in these pathogens; previous attempts at computational predictions have also failed (18). However, recent work in Campylobacter has demonstrated that RNAseq can be useful for analyzing transcriptomic differences between wild-type and mutant strains following inactivation of protein-coding genes (42, 43). Indeed, over the past several years, RNAseq has become the method of choice to analyze the transcriptomes of bacteria under various conditions, as it allows evaluation of the entire transcriptome rather than only previously annotated regions as with older technologies such as microarrays (44–46). In theory, RNAseq should also be able to be utilized for the same purpose of discovering the global effects of inactivation of ncRNAs. Small ncRNAs have been shown to stabilize mRNA transcripts by multiple mechanisms (47), which should lead to increased levels of transcript availability and identification of significant differences using RNAseq. Some interactions with ncRNAs have also been shown to increase transcript turnover by targeting transcripts for RNase degradation or exposing RNase cleavage sites (47), which should lead to decreased levels of transcripts available for identification via RNAseq. Thus, assessment of differential gene expression via RNAseq should be useful for uncovering potential interactions of ncRNAs with mRNA targets when that interaction directly leads to altered levels of mRNA transcripts in the cell; target interactions that do not lead to direct changes in mRNA transcript levels but instead affect the translation efficiency of the target mRNA or are the result of direct protein-ncRNA interactions cannot be determined via this approach. While our differential gene expression results do not guarantee a direct mRNA-ncRNA interaction or provide a mechanism for that interaction, the results of the work presented here clearly indicate that RNAseq can be a useful tool for investigating the role of ncRNAs in post-transcriptional regulation of gene expression in bacteria and can successfully be used to inform the design of phenotypic studies to begin to elucidate the function of ncRNAs in pathogenic bacteria that lack RNA chaperones.

One particular area of interest from the onset of our work was the potential for an interaction between CjNC110 and luxS. Our results related to a potential interaction between luxS and CjNC110 were surprising in that despite increased LuxS expression, extracellular AI-2 levels were decreased while intracellular AI-2 increased; this ultimately suggested that CjNC110 plays a key role in the transport of AI-2 between the intra- and extracellular environment of IA3902. In Escherichia coli, the YdgG (TqsA) protein, part of the AI-2 exporter superfamily, has been described to export AI-2 molecules (48); additional in vitro studies assessing the membrane permeability of AI-2 also demonstrate that it is hydrophilic with low affinity toward lipids, which strongly suggests the need for a transport system (49). Previous BLASTP analysis has shown that potential homologues of the YdgG protein exist within the Campylobacterales family, but none directly within the C. jejuni species (50). Within our proteomics data, several periplasmic proteins and lipoproteins which could be located in the cytoplasmic membrane were identified to exhibit decreased expression in ΔCjNC110 that was corrected to wild-type levels in ΔCjNC110c. The function of many of these proteins has yet to be elucidated in Campylobacter; therefore, future work assessing whether these proteins may be regulated by CjNC110 and could serve as an AI-2 exporter in C. jejuni is warranted. In other species of bacteria, such as E. coli, uptake of AI-2 from the extracellular environment has also been shown to occur and is mediated by an active ABC-type transport system (51). Comparative genomics has again failed to identify a homologous AI-2 uptake system in C. jejuni; however, a number of ABC-type transporters whose function remains to be elucidated do exist in Campylobacter spp. (28). Several studies evaluating the concentration of AI-2 over time following addition of exogenous AI-2 to cell culture have attempted to determine whether or not AI-2 is able to be internalized in C. jejuni with mixed results (52, 53). Thus, further research into an AI-2 uptake system in C. jejuni and the role that CjNC110 may play in regulation of that uptake may also be warranted.

Based on the AI-2 results, the increased expression of LuxS in ΔCjNC110 and failure of complementation of luxS gene expression and LuxS protein abundance in ΔCjNC110c was unexpected, particularly as all other phenotypes tested, including AI-2 detection intra- and extracellularly, fully complemented to wild-type levels. The reason for the increased luxS expression and LuxS protein abundance in ΔCjNC110 and ΔCjNC110c is currently unknown, but one reasonable hypothesis is that this increased expression may be related to additional unknown effects of CjNC110 on the AMC leading to feedback regulation of luxS expression. While regulation of the AMC in other Proteobacteria has been studied in detail, the regulatory pathways of the AMC are currently unknown in Campylobacter, as this genus lacks all of the known transcription factors and riboswitches typically involved regulation (54). As CjNC110 is located immediately downstream of luxS, en bloc removal of the CjNC110 coding sequence could also have led to disruption of normal termination of the luxS transcript, thus altering the natural mRNA stability; this effect would have been maintained in the complement as well. Further investigation to determine the location of a terminator via the WebGesTer database failed to identify a predicted terminator sequence for luxS (http://pallab.serc.iisc.ernet.in/gester/) (55). Therefore, additional work is needed to determine which hypothesis might explain the increase of LuxS in both ΔCjNC110 and ΔCjNC110c. Regardless of the cause, our data clearly demonstrate that the increased luxS expression and LuxS protein abundance in ΔCjNC110 is unrelated to the observed changes in extracellular and intracellular AI-2 concentrations which fully complemented back to wild-type levels in ΔCjNC110c.

Several other important phenotypes were also affected by mutation of CjNC110, including motility, autoagglutination, hydrogen peroxide sensitivity, and chicken colonization. Motility is considered critical for the in vivo virulence of C. jejuni and requires a functional flagellar apparatus (56, 57); however, this alone is not sufficient for normal motility to be present (58). In our RNAseq data, the cetAB operon (Campylobacter energy taxis proteins A and B), which is known to mediate the energy taxis response in Campylobacter (34), was statistically significantly upregulated in the ΔCjNC110 mutant compared to wild-type during the exponential phase of growth; expression of cetAB was increased during stationary phase as well; however, did not reach the level of statistical significance. Defects in both cetA and cetB have been shown to lead to altered motility phenotypes, particularly in response to migration toward critical factors in Campylobacter metabolism, such as sodium pyruvate and fumarate (58). Thus, the increase in cetAB correlates well with the statistically significant increase in motility found between the ΔCjNC110 mutant and the wild-type when motility was assessed at 24 h. Investigation of the expression of cetA and cetB has shown that levels of the gene products are unaffected by mutation of sigma factors σ⁵⁴ or σ²⁸, indicating that transcription of the cetAB operon is likely controlled by σ⁷⁰ or another yet unknown transcription factor (34). Based on this information, it seems plausible that CjNC110 might normally act as a repressor of the CetAB energy taxis system. Our proteomics data failed to detect the CetA and CetB proteins; therefore, we could not validate that increased transcriptional expression of cetA and cetB resulted in increased protein expression levels. Many of the proteins identified in our study exhibited extremely high expression levels, which, without fractioning, prevented proteins of lesser abundance from being detected. Whole-cell proteomic analysis with fractioning and time course analysis would be required to reveal the entire proteomic network regulated by CjNC110 and validate expression patterns of proteins of lesser abundance than the 262 proteins that were identified in our study. Further work to determine if a direct interaction between CjNC110 and the cetAB operon exists and leads to increased expression of the CetA and CetB proteins is warranted.

No statistically significant differences in the expression of genes associated with the flagellar apparatus were noted in the transcriptome of ΔCjNC110; this may help to explain why an increase in energy taxis would allow for an increase in observed motility with the assumption that normally functioning flagella are present. The presence of normal flagella has also been associated with autoagglutination ability and is considered necessary but not sufficient for this important trait (58). Previous studies in C. jejuni demonstrated that in addition to the presence of normal flagella, interactions between modifications on adjacent flagellar filaments, particularly those provided by O-linked protein glycosylation, are required for normal autoagglutination ability (59). In the present study, the ΔCjNC110 mutant exhibited decreased autoagglutination compared to the wild-type at both 25°C and 37°C. Several genes that have been associated with flagellar glycosylation in other strains of C. jejuni were downregulated in our RNAseq data in the ΔCjNC110 mutant compared to the wild-type during both exponential and stationary growth phases. The O-linked protein glycosylation system which is responsible for flagellar glycosylation has been extensively studied in only one strain of C. jejuni (81-176) and has not been studied to date in IA3902 (38); this system is one of the most diverse genomic loci in Campylobacter spp. and also represents one of the key areas of variation between IA3902 and NCTC 11168 (5). Further genetic analysis of the flagellar modification region of the IA3902 genome reveals the presence of genes associated with both of the previously identified pseudaminic acid and legionaminic acid flagellar glycosylation pathways known to exist in Campylobacter spp.; this analysis and the subsequent gene expression data comparison of ΔCjNC110 to the wild-type is presented in Table S7. The pseudaminic acid pathways (encoded by the pseA-I genes, which are fully annotated in IA3902) lead to flagellar glycosylation with either pseudaminic acid (PseAc) (Pse5Ac7Ac), which is the major glycan found on Campylobacter flagella, or derivatives such as PseAm (Pse5Am7Ac) which has an acetamidino substitution (60). In our RNAseq data set, pseA was significantly decreased in expression during exponential growth and also decreased without reaching significance during stationary phase. Mutation of pseA has previously been shown to lead to flagellar modification with only PseAc, and not PseAm (38, 60); pseA has also been shown to be required for optimal adherence and invasion of human epithelial cells and virulence in a ferret model (59).

The legionaminic acid (LegAm) flagellar glycosylation pathway, encoded by the ptmA-H genes, has been less extensively studied in C. jejuni, in part because it is not present in 81-176; in fact, much of the work that has been done in Campylobacter spp. related to this pathway has been in Campylobacter coli (37, 39). NCTC 11168 does contain homologs to all of the ptm genes (Cj1324 to Cj1332), albeit with different gene nomenclature (leg or neu) often used to describe the genes, and mutation of these genes in NCTC 11168 has been shown to decrease autoagglutination ability as well as reduce but not eliminate the ability to colonize chickens (61). While not currently annotated as such, further analysis has identified that in IA3902, gene locus CJSA_1261-1268 represents homologs to all of the ptm genes present in NCTC 11168 (Table S7). Based on these data, it is reasonable to suggest that IA3902 also harbors an active LegAm flagellar glycosylation pathway. Several genes related to legionaminic acid flagellar glycosylation were identified in our RNAseq data set as being differentially expressed in the CjNC110 mutant during both stages of growth, which may help explain altered autoagglutination levels of ΔCjNC110. In particular, ptmA and neuB2 (homologous to ptmC) were both observed to be statistically significantly decreased at both exponential and stationary phases, while ptmB was statistically significantly decreased at stationary phase and decreased but not reaching statistical significance during exponential phase. Mutation of the ptmA-G genes in C. coli results in the loss of all LegAm derivatives and a conversion to glycosylation with PseAc only; these mutants still retain a normal flagellar apparatus and motility (37, 39, 62). In our RNAseq data, CJSA_1261 (homologous to ptmG) was the sole flagellar modification gene that demonstrated a statistically significant increase in expression during both growth phases; ptmG has been shown to be involved in conversion of LegAm to various derivatives (Leg5AM7Ac and Leg5AmNMe7Ac) (37). Alterations in ptmG levels in the CjNC110 mutant are of particular interest, as it has been recently shown to be also regulated by another small noncoding RNA pair in C. jejuni, CjNC180/CjNC190 in NCTC 11168 (27). While CjNC180/CjNC190 were not detected in this data set by Rockhopper, these ncRNAs are present in IA3902, and CjNC180 has previously been demonstrated to be significantly upregulated in the in vivo host gallbladder environment in IA3902 (24). When taken together, these data suggest that CjNC110 may play a key role in modulation of flagellar glycosylation in IA3902, thus directly affecting autoagglutination ability. Other potential causes for decreased autoagglutination ability include alteration of the flagellar filament. While not identified via any other approach, the FlaA protein was demonstrated to be decreased via hierarchal heat mapping in our proteomics data in the CjNC110 mutant with a return to wild-type levels in the complement (Fig. S6); if this change corresponds to an alteration of the flagellar structure, it could in theory also explain the observed decrease in autoagglutination ability. Thus, further work, including determination of the normal flagellar glycosylation of IA3902 and assessment for changes in both the flagellar filament structure and glycosylation of the ΔCjNC110 mutant compared to the IA3902 wild-type, is warranted.

The RNAseq data also demonstrated that in the ΔCjNC110 background, tpx was significantly downregulated during the exponential phase of growth, prompting phenotypic assessment of hydrogen peroxide sensitivity. Thiol peroxidase (Tpx) was previously shown to function as a hydrogen peroxide reductase in NCTC 11168; however, in that strain, Tpx-defective mutants were not shown to be sufficient to cause alteration of hydrogen peroxide sensitivity via disc diffusion and required a second mutation of the Bcp gene to observe increased sensitivity (35). IA3902 also harbors a Bcp homolog that is 96% homologous to NCTC 11168; this gene did not demonstrate altered expression via RNAseq analysis, and functional evaluation of this gene in IA3902 has not been performed to date. While our results clearly indicate that the loss of CjNC110 increases hydrogen peroxide sensitivity, and that the expression of tpx in the mutant decreased, it is not possible to definitively state that the observed phenotype is due solely to an interaction of CjNC110 with tpx alone. Further analysis of potential alterations of CjNC110 with additional known oxidative stress genes in the RNAseq data did indicate that sodB, which was previously identified to contribute to hydrogen peroxide sensitivity in C. jejuni (63), demonstrated a q value that reached the threshold for significance (q = 0.004); however, the fold change just missed the cutoff for significance at –1.4. Thus, further investigation for the potential interaction of CjNC110 with both tpx and additional genes that control the oxidative stress response in C. jejuni, such as sodB, is also warranted.

When taken together, the results of the in vitro phenotypic studies, which identified several key phenotypes affected by the loss of CjNC110, strongly suggested that mutation of CjNC110 may interfere with normal colonization of IA3902. Indeed, our results show that mutation of ΔCjNC110 led to a significantly decreased although still present colonization ability in IA3902 compared to the wild-type, indicating that this ncRNA has the potential to play a key role in modulating colonization by C. jejuni. The exact mechanism for the decreased colonization ability remains unknown but could be due to any combination of factors identified in the phenotypic screening, including increased sensitivity to reactive oxygen species (ROS), decreased autoagglutination ability, or altered quorum sensing ability; further work to elucidate how these factors may interact to affect in vivo colonization is thus warranted. The loss of the ability of ΔluxS to sustain colonization in chickens, as demonstrated in this study, has been previously described in IA3902, but the same phenotype was not observed in the closely related NCTC 11168 (8); the reason for this difference is currently unknown but has been speculated to be related to strain-specific differences in the metabolism of methionine or S-adenosylmethionine (SAM) recycling, which are also affected by the luxS mutation (28). Additional work in our lab has shown that the luxS mutation in IA3902 does disrupt the AMC, which functions to recycle SAM and produce methionine (64); however, DNA methylation in IA3902 was not affected by the luxS mutation (65). Comparative genomics reveals that NCTC 11168 does harbor an additional genomic system for methionine biosynthesis in the form of metAB which is not present in IA3902 and may represent an alternative pathway for synthesis of this required amino acid and an explanation for the observed difference in colonization ability. Further work is ongoing in our lab to determine if differing methionine metabolism may explain the difference in colonization ability of luxS mutants between strains, as well as to assess if mutation of CjNC110 in NCTC 11168 also leads to a defect in colonization, or if similar to luxS, the phenotype differs from IA3902.

After attempting to compare various strains and methods of mutation of the luxS gene and finding wide variation in phenotypical and transcriptional changes, Adler et al. (29) suggested that some of the differences reported between luxS mutants of C. jejuni in various studies may be due to unknown polar affects caused by alteration of expression of CjNC110 based on the mutation strategy used for the luxS mutation. Our Northern blot analysis confirmed that CjNC110 was still expressed at the same level as the wild-type strain in our luxS mutant construct. In addition, our phenotypic data clearly demonstrate very different phenotypes between ΔCjNC110 and ΔluxS in IA3902 for all phenotypes tested, with variation in the phenotype of ΔCjNC110ΔluxS based on the dominant phenotype of the two mutants. Thus, based on the work presented here, for our luxS mutant constructs, we can definitively state that polar effects on CjNC110 are not present.

Initial analysis of transcriptome changes in the ΔluxS mutant in our study revealed very few genes identified as differentially expressed, and the list of genes identified did not overlap with the list of differentially expressed genes generated primarily via microarray in various other strains (66–68). One explanation for this discrepancy may be differences between strains of C. jejuni or culture conditions compared to these previous studies. However, in distinct contrast to the lack of similar findings to previous gene expression studies exhibited by the ΔluxS mutant, when the ΔCjNC110 and ΔluxS mutations were combined into ΔCjNC110ΔluxS, many of the previously noted transcriptional changes attributed to mutation of the luxS gene alone in a different strain of C. jejuni, 81-176, which also encodes and expresses CjNC110, became apparent (67). Of the 57 genes listed as differentially expressed by the luxS mutant under various conditions in He et al. (67), 23 were also found to be differentially expressed in our RNAseq data (any mutant construct). Of the 23 overlapping genes, no genes were found only in the ΔluxS mutant, and only the flagellar genes flgE, flgG2, and flgG and a putative helix-turn-helix protein CJSA_1449 were differentially expressed in both the ΔluxS and ΔCjNC110ΔluxS mutants. In contrast, three genes (tpx, ptmB, ptmA) were differentially expressed only in the ΔCjNC110 mutant. The He et al. study (67) also reported increased sensitivity to H₂O₂ in their luxS mutant construct, which more closely aligns with our ΔCjNC110 and ΔCjNC110ΔluxS mutant data and not ΔluxS. These findings strongly suggest that as hypothesized by Adler et al. (29), in some previous studies of ΔluxS, the main driver for the transcriptional and phenotypic changes seen was likely an inadvertent inactivation of both CjNC110 and luxS, and not luxS alone. Thus, further work is needed to reevaluate the expression of CjNC110 in various luxS mutant constructs to determine if alteration of expression of CjNC110 is present and may account for some of the differences seen in phenotypes associated with luxS mutation in C. jejuni.

Of the genes identified in both our study and He et al. (67), a large number of the flagellar assembly hook-basal body-associated proteins (FlgD, FlgE, FlgG, FlgG2, FlgH, FlgI, and FlgK) that are under the control of σ⁵⁴ (RpoN) promoters were identified as differentially expressed (69, 70). Expression from σ⁵⁴ promoters has been shown to require activation of the FlgRS two-component regulatory system (69, 71). Many of these genes showed opposite expression based on growth phase (decreased exponential, increased stationary), and it has been previously demonstrated that regulation by these sigma factors is growth cycle dependent, with σ⁵⁴-regulated genes typically expressed between σ⁷⁰- and σ²⁸-associated genes (69). It has been suggested that an additional unknown factor may control the temporal regulation of σ⁵⁴-dependent flagellar genes (72), and the intergenic region between luxS and CJSA_1137 where CjNC110 is located has previously been identified to demonstrate differential expression in an rpoN (σ⁵⁴) mutant in NCTC 11168 (19). Therefore, it is reasonable to consider based on previous studies and our results that the genomic region encompassing luxS and CjNC110 may play a role in the growth-phase-dependent regulation of flagellar assembly.

While not explored further in our phenotypic studies, Rockhopper did identify two additional ncRNAs, CjNC140 and CjNC130/6S, as being differentially expressed in our mutants; these ncRNAs were previously identified and confirmed to be transcribed in multiple other C. jejuni strains (21). Predicted to be transcribed in the intergenic region upstream of porA, the CjNC140 ncRNA was upregulated during exponential growth in all three of the mutant strains compared to the wild-type (Fig. S7). As increased expression was not observed during stationary growth, this suggests that CjNC140 may serve as a regulator involved in mediating changes during different stages of bacterial growth. The fact that its expression is similarly altered under all 3 mutant conditions also suggests that the three genes, luxS, CjNC110, and CjNC140, may normally interact within the cell in some way to facilitate these changes. Point mutations in the nearby porA have also been recently determined to be sufficient to cause the abortion phenotype (10), making this ncRNA particularly interesting for further study.

Finally, while the specific mechanisms by which ncRNA CjNC110 interacts with each individual target were not elucidated in this study, the research presented here clearly identified several specific roles of CjNC110 in the pathophysiology of C. jejuni IA3902. Indeed, even for well-studied ncRNAs such as MicA in Enterobacteriaceae, new publications are still being produced which describe new roles and mechanisms of action even after a decade of research (recently reviewed in reference 73). Future work is necessary to investigate the hypotheses generated here for genes and proteins that may be regulated by CjNC110 and to determine the mechanism by which CjNC110 interacts with each of the potential mRNA targets identified in this study.

In summary, the results of our research clearly demonstrate a significant role for CjNC110 in the pathobiology of C. jejuni IA3902. The collective results of the phenotypic and transcriptomic changes observed in our data complement each other and suggest that CjNC110 may be involved in regulation of energy taxis, flagellar glycosylation, cellular communication via quorum sensing, oxidative stress, and chicken colonization in C. jejuni IA3902. This work provides for the first time valuable insights into the potential regulatory targets of the CjNC110 small ncRNA in the zoonotic pathogen C. jejuni and suggests a wide range of research avenues into the role of ncRNAs in the pathobiology of this important zoonotic pathogen.

MATERIALS AND METHODS

Bacterial strains, plasmids, primers, and culture conditions.

C. jejuni IA3902 was initially isolated from an outbreak of sheep abortion in Iowa in 2006 and has been utilized by our laboratory as the prototypical isolate of clone SA (3). W7 is a highly motile variant of the commonly utilized laboratory strain C. jejuni NCTC 11168 (8). C. jejuni strains and their isogenic mutants were routinely grown in Mueller-Hinton (MH) broth or agar plates (Becton, Dickinson, Franklin Lakes, NJ) at 42°C under microaerophilic conditions (5% O₂, 10% CO₂, 85% N₂). For the mutant strains containing a chloramphenicol resistance cassette, 5 μg/ml chloramphenicol was added to either the broth or agar plates when appropriate. For strains containing a kanamycin resistance cassette, 30 μg/ml kanamycin was added to either the broth or agar plates when appropriate.

For genetic manipulations, Escherichia coli competent cells were grown at 37°C on Luria-Bertani (LB) agar plates or broth (Becton, Dickinson, Franklin Lakes, NJ) with shaking at 125 rpm. When appropriate, 50 μg/ml kanamycin, 20 μg/ml chloramphenicol, or 100 μg/ml ampicillin was added to the broth or agar plates for selection of colonies. Vibrio harveyi strains were grown in autoinducer broth (AB) at 30°C with shaking at 175 rpm as described previously (74).

All strains used in this study are described in Table 1, with all relevant plasmids listed in Table S8 and primer sequences and locked nucleic acid-digoxigenin (LNA-DIG)-labeled probe sequences listed in Table S9. Figure S8 illustrates the genetic modification strategy described below to create the mutants utilized in this study. All strains were maintained in 20% glycerol stocks at –80°C and passaged from those stocks as needed for experimental procedures.

Creation of C. jejuni ΔCjNC110 and ΔCjNC110ΔluxS mutants in IA3902.

An isogenic CjNC110 mutant of C. jejuni IA3902 was constructed via deletional mutagenesis utilizing a combination of synthetic double-stranded DNA (dsDNA) fragments and traditional cloning methods. Based on previously published data depicting the proposed transcriptional start site for CjNC110 (21), the coding region of CjNC110 in IA3902 and the prototypical C. jejuni strain NCTC 11168 were first confirmed to be identical. Then, a 200 bp section of the IA3902 genome starting 20 bp upstream and including the entire 137 nucleotide (nt) transcript of the CjNC110 sequence as predicted in Dugar et al. (21) was replaced with 820 bp of the promoter and coding sequence of the chloramphenicol acetyltransferase (cat) gene of Campylobacter coli plasmid C-589 (75). Synthetic dsDNA including approximately 500 bp upstream and 500 bp downstream of the region replaced with the cat cassette was then synthesized in 4 fragments of 500 bp each with overlapping homologous ends (Integrated DNA Technologies, Coralville, IA). The Gibson assembly method was then utilized to assemble the synthetic dsDNA fragments using Gibson assembly master mix (New England Biolabs, Ipswich, MA) (76). Following assembly, primers (CjNC110F2 and CjNC110R2) were designed to amplify a 1,785-bp product of the assembled dsDNA; PCR amplification was achieved using TaKaRa Ex Taq DNA polymerase (ClonTech, Mountain View, CA). This amplified PCR product was then cloned into the pGEM-T easy vector using T4 ligase (Promega, Madison, WI), resulting in the construction of pCjNC110::cat, which was then transformed into chemically competent E. coli DH5α (New England Biolabs, Ipswich, MA). Transformants were then selected on LB agar plates containing chloramphenicol (20 μg/ml), ampicillin (100 μg/ml), and ChromoMax IPTG/X-Gal solution (Fisher Scientific, Pittsburgh, PA). pCjNC110::cat was purified from the transformed E. coli using the QIAprep miniprep kit (Qiagen, Germantown, MD) and confirmed by PCR to contain the construct again using the CjNC110F2 and CjNC110R2 primers.

The pCjNC110::cat plasmid DNA was then introduced to C. jejuni W7, a highly motile variant of NCTC 11168, as a suicide vector, and the deletion was transferred into the genome of C. jejuni W7 via homologous recombination. Transformants were selected on MH agar plates containing chloramphenicol (5 μg/ml), and deletional mutagenesis was again confirmed via PCR analysis and Sanger sequencing to create C. jejuni W7ΔCjNC110. Following confirmation, natural transformation was used to move the gene deletion into C. jejuni IA3902 as previously described to create C. jejuni IA3902ΔCjNC110 (77). Natural transformation was again used to move the CjNC110 gene deletion into the previously created luxS insertional mutant C. jejuni IA3902ΔluxS (8) to create the double-knockout mutant C. jejuni IA3902ΔCjNC110ΔluxS. Transformants were selected on MH agar plates containing chloramphenicol (5 μg/ml) and kanamycin (30 μg/ml) and confirmed via PCR analysis and Sanger sequencing of the CjNC110 region along with the entire upstream (luxS) and downstream (CJSA_1137) genes using primers Cj1198F1 and Cj1199R3. All colonies were screened for the presence/absence of motility as described below, and only colonies with verified motility were used for future studies. Expression of CjNC110 in the wild-type and elimination of expression in the ΔCjNC110 mutant were confirmed via Northern blot analysis using an LNA custom-designed probe and 15 μg of total RNA as described below.

Creation of C. jejuni ΔCjNC110 complement in IA3902.

Complementation of ΔCjNC110 was achieved via insertion of the coding region into the intergenic region of the 16S and 23S rRNA operon (rrs-rrl) of IA3902ΔCjNC110 via homologous recombination using plasmid pRRK. A copy of the CjNC110 gene including 190 bp upstream of the transcription initiation location (total of 570 bp) was cloned into an XbaI site of pRRK located upstream of the kanamycin resistance determinant to create pRRK-CjNC110 as previously described in our lab (78) with some modifications of the original protocol (79). Briefly, the CjNC110 gene and flanking regions of IA3902 were amplified using forward and reverse primers designed with an XBaI restriction enzyme site (CjNC110cF1 and CjNC110cR1). Following PCR amplification with TaKaRa Ex Taq DNA polymerase (ClonTech, Mountain View, CA), the PCR product and pRRK plasmid were digested with the XBaI restriction endonuclease (New England Biolabs, Ipswich, MA). The digested plasmid was then incubated twice with Antarctic phosphatase (New England Biolabs, Ipswich, MA) to prevent self-ligation of cut plasmid. Following ligation and transformation, transformants were selected on LB agar plates containing chloramphenicol (20 μg/ml) and kanamycin (50 μg/ml kanamycin), yielding a plasmid containing the construct pRRK::CjNC110 in E. coli DH5α. Following PCR confirmation and Sanger sequencing to confirm preservation of the correct gene sequence, the plasmid was utilized as a suicide vector, and the gene was inserted into the intergenic region of the 23S rRNA operon (rrs-rrl) of IA3902ΔCjNC110 via homologous recombination to generate IA3902ΔCjNC110c transformants. The transformants were selected on MH agar plates containing chloramphenicol (5 μg/ml) and kanamycin (30 μg/ml) and confirmed via PCR analysis and Sanger sequencing to contain both the expected ΔCjNC110 deletional mutation and the ΔCjNC110c insertion, validating successful genetic complementation. All positive colonies were screened for the presence/absence of motility as described below, and a motile complement was selected to create IA3902ΔCjNC110c. Expression of CjNC110 in ΔCjNC110c was confirmed via Northern blot analysis using an LNA custom-designed probe and 15 μg of total RNA as described below.

Growth curves.

Two separate growth curves were completed in triplicate utilizing the same method as described below, with the only variation being the addition of the complement strain ΔCjNC110c to the second growth curve. To begin the growth curve, the A₆₀₀ of overnight cultures were adjusted to 0.5 using sterile MH broth on a Genesys 10S VIS spectrophotometer (Thermo Scientific, Waltham, MA). Cultures were then diluted 1:10 for a final targeted starting A₆₀₀ of 0.05 in 90 ml MH broth and placed in a sterile 250-ml Erlenmeyer glass flask. Cultures were incubated at 42°C under microaerophilic conditions with shaking at 125 rpm for 30 h with removal of samples from the flasks at designated time points (3, 6, 9, 12, 24, 27 [2nd growth curve only] and 30 h). For the first growth curve (which included the following strains: IA3902 wild-type, ΔCjNC110, ΔluxS, and ΔCjNC110ΔluxS), samples were processed as described below for RNA isolation and assessed for A₆₀₀ and actual colony counts using the drop-plate method as previously described (80). For the second growth curve (which included ΔCjNC110c in addition to the four previous strains), samples were collected and processed for A₆₀₀ and as described below for qRT-PCR and assessment of AI-2 levels via the bioluminescence assay. The A₆₀₀ over time were statistically analyzed for the two independent growth curves using a two-way ANOVA with repeated measures and Dunnett’s multiple-comparison test (GraphPad Prism).

Total RNA extraction.

Broth culture samples processed for total RNA extraction were centrifuged at either 8,000 × g for 2 min (for sequencing and qRT-PCR) or 10,000 × g for 4 min (Northern blot) at 4°C immediately following collection. Following pelleting of the cells, the supernatant was decanted, and 1 ml QIAzol lysis reagent (Qiagen) was added to resuspend the cell pellet and protect the RNA. QIAzol-protected cultures were then stored at –80°C for up to 2 months prior to proceeding with total RNA isolation. Total RNA isolation was performed using the miRNeasy minikit (Qiagen) followed by purification using the RNeasy MinElute cleanup kit (Qiagen) as previously described (24). RNA quality was measured using the Agilent 2100 bioanalyzer RNA 6000 nano kit (Agilent Technologies, Santa Clara, CA), and all RNA samples utilized for downstream RNAseq library preparation had an RNA integrity number (RIN) of >9.0, indicating high-quality RNA. Verification of complete removal of any contaminating DNA was performed via PCR amplification of a portion of the CJSA_1356 gene, which is unique to C. jejuni IA3902, using primers SA1356F and SA1356R (81).

RNAseq library preparation and sequencing.

The 3-h (exponential phase) and 12-h (stationary phase) time points were selected for RNAseq analysis based on assessment of log₁₀ CFU/ml (Fig. S2A). Then, 2.5 μg of confirmed DNA-free total RNA was treated with a Ribo-Zero rRNA removal kit for bacteria according to the manufacturer’s instructions (Illumina, San Diego, CA). Following rRNA removal, the ribosomal depleted total RNA was purified using the RNeasy MinElute cleanup kit (Qiagen) using the same modifications as described previously (24). Following cleanup, the RNA quality and quantity and the rRNA removal efficiency were analyzed using the Agilent 2100 bioanalyzer RNA 6000 Pico kit (Agilent Technologies). Library preparation for sequencing on the Illumina HiSeq platform was completed using the TruSeq stranded mRNA high-throughput (HT) library preparation kit (Illumina) with some modifications as described previously (24). The pooled library was then submitted to the Iowa State University DNA facility and sequenced on an Illumina HiSeq 2500 machine in high-output single-read mode with 100 cycles.

Differential gene expression analysis of RNAseq data.

Rockhopper (http://cs.wellesley.edu/~btjaden/Rockhopper/) was used to analyze the differences in gene expression between strains and time points; the standard settings of the program were utilized for analysis (30). Following computational analysis via Rockhopper, a change in gene expression was deemed significant when the q value (false-discovery rate) was <0.05 and a >1.5 fold change in expression was observed. Visual assessment of read count data was performed using the Integrated Genome Viewer (IGV) (https://www.broadinstitute.org/igv/) (82, 83). Assessment of function of the differentially expressed genes was performed using the Clusters of Orthologous Groups (COG) (84) as previously described for IA3902 (5). Venn diagrams were used to depict overlap of genes differentially regulated in multiple mutant strains and were generated using BioVenn (http://www.cmbi.ru.nl/cdd/biovenn/index.php) (85). KEGG Pathways was used to perform metabolic pathway analysis (http://www.genome.jp/kegg/pathway.html) (31).

In addition to determining differential gene expression between the mutant and wild-type strains of IA3902, Rockhopper has the capability to predict noncoding RNAs present within the data. Prior to further manual analysis, Rockhopper predicted a total of 59 ncRNAs present in IA3902 (57 chromosomal, 2 pVir). Manual curation was performed to remove candidate noncoding RNAs if they were related to the 16S or 23S genes, as these were thought to be spuriously identified due to Ribo-Zero depletion differences between replicates. In addition, a single antisense RNA was discarded due to incorrect annotation of the rnpB gene to the wrong strand in IA3902.

Northern blot detection of CjNC110.

For Northern blot detection of CjNC110, two separate experiments utilizing identical conditions were performed as previously described with some modifications (86). Experiment 1 included wild-type IA3902, ΔCjNC110, and ΔCjNC110c, while experiment 2 included the wild-type and ΔluxS. For both, bacterial cells were grown to the early stationary phase of growth in MH broth, and total RNA isolation was performed as described above. The total RNA input was normalized to 15 μg using a Qubit BR RNA assay kit (Invitrogen, USA) for each respective strain. For Northern blot detection, total RNA from bacterial cells (15 μg) and a prestained DynaMarker RNA high ladder (Diagnocine LLC, Hackensack, NJ) were loaded onto 1.3% denaturing agarose gel (Bio-Rad, Hercules, CA), consisting of 1× MOPS buffer (Fisher, USA) and 5% solution of 37% formaldehyde (Fisher, USA), electrophoresed at 100 V for 1 h, and transferred to positively charged nylon membrane (Roche, Indianapolis, IN) using a vacuum blot system (Bio-Rad) transferring at 60 mbar for 2 h. To make the RNA cross-linking solution, 245 μl of 12.5 M 1-methylimidazole (Sigma, St. Louis, MO) was suspended in 9 ml RNase-free water, maintaining a pH of 8.0. Then, a total of 0.753 g 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) was added, and the total volume adjusted to 24 ml with RNase-free water. RNA was then cross-linked to the membrane by placing the nylon membrane, side devoid of RNA, on saturated 3 mm Whatman chromatography paper and incubating it in the freshly prepared EDC solution at 60°C for 2 h. After cross-linking, the membrane was washed with RNase-free water. Next, CjNC110-LNA DIG-labeled probe (/5′DigN/GCACATCAGTTTCAT/3′Dig_N/) (Qiagen) was added to 15 ml DIG EasyHyb buffer (Roche) to reach a final concentration of 25 ng/ml. Hybridization was performed using a hybridization oven and bottle, rotating at 54°C for 12 h. After overnight incubation, subsequent washes and incubations were performed, again using the hybridization chamber. The blot was first washed twice with low-stringency buffer (2 × SSC and 0.1% SDS) for 15 min at 60°C and three times with high-stringency buffer (0.1% SSC and 0.1% SDS) for 10 min at 60°C. DIG washing and DIG blocking buffers were prepared using the DIG wash and block set (Roche, USA). The blot was washed twice with washing buffer for 10 min at 37°C. Next, the blot was incubated in 1× blocking buffer for 3 h at room temperature. DIG antibody solution was added by mixing blocking buffer with DIG antibody at a ratio of 1:5,000. The blot was then incubated for 30 min at room temperature. Next, excess DIG antibody was removed by rinsing with washing buffer at 25°C a total of four times for 15 min each. For development and detection, the membrane was placed in development/detection buffer for 10 min, and 2 ml ready-to-use CDP-Star (Roche, USA) was added to cover the blot completely. To image the blot, Chemidoc imager (Bio-Rad) was used, exposing the blot for 5 min on the chemiluminescence setting. Image layout and annotations were performed using GraphPad Prism. ImageLab software (Bio-Rad) was used to generate a standard curve using replicative form (RF) and nucleotide size of the ladder to predict the average band size of the target RNA.

Enumeration of luxS transcriptional levels utilizing qRT-PCR.

To validate the RNAseq results and compare luxS expression between mutants, quantitative reverse transcriptase PCR (qRT-PCR) was performed using total RNA isolated from the IA3902 wild-type and ΔCjNC110 and ΔCjNC110c mutants purified in three separate biological replicates of broth cultures grown to 12 h from the second growth curve described above. Total RNA was extracted, and qRT-PCR was performed using 1,000 ng of total RNA and the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA), according to the manufacturer’s instructions as previously described (64). Each sample was normalized to the same starting amount of cDNA using a Qubit BR DNA assay kit (Invitrogen). cDNA purity was measured using the NanoDrop ND-1000 spectrophotometer. qPCR assays were run using the SsoAdvanced universal SYBR green supermix (Bio-Rad, Hercules, CA) and the CFX Maestro real-time PCR detection system (Bio-Rad). Dilutions of cDNA template for both standards and all unknowns were run in triplicate with reaction volumes of 10 μl. Amplification of converted cDNA occurred with 35 cycles of denaturation at 95°C for 10 s and then annealing for each primer pair (Table S9) at 58°C for 30 s. Prior to analysis, both standard curves (16S and luxS) were experimentally validated to have high efficiency of >90% of amplification and precision of R² = 0.98 or greater. The relative fold change in luxS mRNA expression between the IA3902 wild-type, ΔCjNC110, and ΔCjNC110c was calculated using the ISU Gallup method equation (87). Statistical analyses were performed using one-way ANOVA (GraphPad Prism) to determine significance in gene expression levels. A P value of <0.05 was considered significant.

Liquid chromatography with tandem mass spectrometry-based proteomic analysis.

For extraction of whole cell protein, bacterial cells from the IA3902 wild-type, ΔCjNC110, and ΔCjNC110c were grown to the stationary phase of growth at 16 h on MH agar. Lawn cultures were then harvested from the plates using 1 ml of MH broth. Collected lawns were normalized to an A₆₀₀ of 0.2 and aliquoted into 1-ml technical replicates. This was repeated for three biological replicates. To wash the cells, each aliquot was resuspended in 1 ml of phosphate-buffered saline (PBS), cells were spun down at 8,000 × g for 3 min, and the supernatant was decanted. Next, 50 μl of lysis buffer (filtered sterilized water + 1% Triton X-100) was added to the protein pellets for resuspension (adapted from reference 88). The cells were then mechanically sheared by boiling at 96°C for 5 min followed by rapid pipetting. After 3 cycles of boiling-shearing, total cell extracts were centrifuged at 10,000 × g for 5 min to remove cell debris. The supernatants were collected, and protein concentration was determined using the Qubit protein assay and fluorometer (Invitrogen). Crude protein was then submitted to the Iowa State Protein Facility. Crude protein extracts were digested overnight in solution with trypsin/Lys-C (25 to 50 μg). After digestion, Pierce peptide retention-time calibration mixture (PRTC) (Thermo Fisher Scientific) was spiked into the samples to serve as an internal control (25 fmol/μl). The PRTC areas were used to normalize collision energies, retention times, and peak intensities to allow quantitative analysis between samples.

Proteomic analysis was performed via liquid chromatography with tandem mass spectrometry (LC-MS/MS) using a Q Exactive hybrid quadrupole-orbitrap mass spectrometer (Thermo Scientific) system. The system was coupled with an EASY-nLC 1200 nanopump with integrated auto-sampler (Thermo Scientific). Liquid chromatography was used to separate the peptides, followed by fragmentation and MS/MS analysis. The resulting intact fragmentation pattern was compared to a theoretical fragmentation pattern (MASCOT) to find peptides that can be used to identify high-confidence proteins. The Minora feature detector was used for label-free quantification to detect and quantify discovered isotopic clusters.

Abundances of proteins were compared by grouping by strain and averaging the abundances of each protein identified. For data filtering and data imputation, PANDA-view, a freely available proteomic software (https://sourceforge.net/projects/panda-view/), was used (89). Briefly, only proteins identified in two out of three biological replicates were included in the analysis. Nearest neighbor (kNN) imputation was used to fill any missing protein values (90). All data were log transformed using log₂. For statistical analysis, two-way ANOVA with false-discovery rate (q value) correction set at 0.05 was performed (GraphPad Prism). Graphical hierarchical clustering heat maps were generated using freely available MetaboAnalyst software (91). The top 20 proteins were grouped and analyzed using the following heat map settings: distance measurement of Euclidean and the average clustering algorithm (91, 92).

Vibrio harveyi bioluminescence assay.

Culture samples collected from time points 3, 6, 9, and 12 h (2 ml each time point) from the second growth curve described above were centrifuged at 20,000 × g for 5 min at 4°C. The supernatant was then removed from the cell pellet and filter-sterilized using a 0.2-μm syringe filter to create cell-free supernatant representative of the extracellular environment (E-CFS). The cell pellet was then washed twice with PBS; following each wash, the cells were spun again at 8,000 × g, and the resulting supernatant was completely removed from the cell pellet. Both the E-CFS and the cell pellet were then frozen at –80°C. For processing of the cell pellet to create intracellular cell-free supernatant (I-CFS), samples were thawed on ice and then resuspended in 600 μl of ice-cold MH broth. The resuspended cell pellets were then lysed using a Bullet Blender (Next Advance, Troy, NY). To lyse the cells, 1.5-ml microcentrifuge tubes (Corning, Corning, NY) were loaded with sterilized 0.9- to 2.0-mm-diameter stainless steel beads (Next Advance). The Bullet Blender was used for 5 min at maximum speed and stored at 4°C. Lysed cellular debris was spun at 8,000 × g for 2 min using a microcentrifuge, and the resulting supernatant was filtered using a 0.20-μm sterile syringe filter, resulting in intracellular cell-free supernatant (I-CFS). Both E-CFS and I-CFS were frozen at –80°C until the bioluminescence assay was continued.

Autoinducer-2 levels within the collected supernatant for both intracellular (I-CFS) and extracellular (E-CFS) samples were then enumerated using the Vibrio harveyi bioluminescence assay as previously described (32). Briefly, 10 μl of each CFS sample was added to 90 μl AB media containing a 1:5,000 dilution of the reporter strain, V. harveyi strain BB170 (93), in triplicate. Relative light units (RLU) were measured every 15 min over 8 h using FLUOstar Omega (BMG Labtech, Ortenburg, Germany). MH broth and AB medium were used as negative controls, while E-CFS collected from V. harveyi strain BB152 (74) was used as a positive control. The time points utilized for analysis were those occurring during the nadir of values for the negative control wells and at a standardized duration of time (approximately 3 h) following initiation of increasing values for the positive wells. The results reported are the average of three independent growth curves with three technical replicates used at each time point. Statistical analysis was conducted using two-way ANOVA with repeated measures and Sidak’s multiple-comparison test (GraphPad Prism). A P value of <0.05 was considered significant.

Motility assay.

Motility was determined via inoculation of plates consisting of MH broth with 0.4% agar as previously described in our laboratory (8). Briefly, the A₆₀₀ of overnight cultures was adjusted to 0.3 using sterile MH broth on a Genesys 10S VIS spectrophotometer (Thermo Scientific). A 1-μl-volume inoculation stick was then dipped into a set volume of the standardized culture contained in the bottom of a 15-ml conical tube, which was then used to make a stab inoculation into the center of the freshly made motility agar (MH broth with 0.4% Bacto agar), with a new inoculation stick for each plate. Plates were incubated at 42°C under microaerophilic conditions as described above, with the exception that the plates were incubated right-side up and in a single layer. Measurement of the outermost reach of the halo was performed at 30 h following inoculation for the first set of experiments and at 24 h for the second set of experiments. All strains were assessed for average motility using six biological replicates consisting of six technical replicates during each independent study. The six experiments were statistically analyzed using one-way ANOVA, and differences between each strain were assessed via Tukey’s multiple-comparison test (GraphPad Prism). A P value of <0.05 was considered significant.

Autoagglutination assay.

Autoagglutination was assessed according to the method described previously (40) with some modifications. Briefly, the A₆₀₀ of overnight cultures was adjusted to 1.0 in sterile Dulbecco’s phosphate-buffered saline (PBS) (Cellgro; Corning, Manassas, VA) using a Genesys 10S VIS spectrophotometer (Thermo Scientific). The suspension was then aliquoted (2 ml each) into standard glass culture tubes. One subset of cultures was kept at controlled room temperature (25°C) under microaerophilic conditions; the others were incubated at 37°C under microaerophilic conditions. At 24 h, 1 ml of the upper aqueous phase was carefully removed, and the A₆₀₀ was measured to determine autoagglutination activity. All strains were assessed in quadruplicate at each temperature in three independent experiments. The three experiments were statistically analyzed using one-way ANOVA, and differences between each strain were assessed via Tukey’s multiple-comparison test (GraphPad Prism). A P value of <0.05 was considered significant.

Hydrogen peroxide disk inhibition assay.

The hydrogen peroxide (H₂O₂) disk inhibition assay was performed as previously described with some modifications (63). Briefly, C. jejuni IA3902 wild-type and mutant strains were grown to log phase in MH broth. The bacteria were harvested by centrifugation at 12,000 × g for 2 min and subsequently normalized by resuspension into MH broth to an A₆₀₀ of 1.0. Next, 24 ml of melted MH agar was combined with 1 ml of the normalized bacterial suspension on petri plates, mixed well, and allowed to solidify. 20 μl of aqueous 3% H₂O₂ solution (Sigma-Aldrich, St. Louis, MO) was then pipetted onto a 6-mm-diameter disk, which was placed at the center of the plate containing the bacterial agar suspension. Plates were incubated for 24 h at 42°C under microaerophilic conditions. The zones of H₂O₂ sensitivity were measured in millimeters (mm) by measuring the diameter of the clear zones. The H₂O₂ disk inhibition assay was repeated three times using freshly prepared 3% H₂O₂ with 4 technical replicates per strain. The data were statistically evaluated using one-way ANOVA and Dunnett’s multiple-comparison test. A P value of <0.05 was deemed significant.

Chicken colonization.

All studies involving animals were approved by the Iowa State University Institutional Animal Care and Use Committee (IACUC) prior to initiation (1-18-8675-G) and followed all appropriate animal care guidelines. Chicken colonization studies were performed as previously described in our laboratory utilizing 1-day-old broiler chicks obtained from a commercial hatchery (8). A subset of chicks from each group were screened via cloacal swabs plated on MH agar containing both selective supplement and growth supplement (MH+sel+sup) (Oxoid, Thermo Fisher Scientific) and found to be negative for Campylobacter carriage prior to inoculation. At 3 days of age, chicks were inoculated by oral gavage with 200 μl of bacterial suspension containing approximately 1 × 10⁷ of one of the tested strains. Groups of chicks inoculated with specific strains were housed in separate brooders with no contact between the groups. At predetermined time points post-inoculation (5 days, 12 days, and 19 days), 6 chicks from each group were randomly selected for humane euthanasia, which was immediately followed by necropsy. Cecal contents were harvested aseptically and stored on ice until further processing could be completed. Cecal samples were then weighed and subjected to a 10-fold serial dilution series, plated on MH+sel+sup agar, and grown under routine culture conditions as described above. To confirm that the isolates recovered were, in fact, the mutant strain and no cross-contamination occurred, representative colonies were selected from each strain and time point, grown in MH+sel+sup broth, and plated on the respective antimicrobial selective agar of the mutant strain as described above. In addition, a minimum of two CFU from each group were collected for PCR analysis and confirmation of no cross-contamination between biological groups. The detection limit of the assay was determined to be 100 CFU/g; for samples that failed to reach the optimal target range of 30 to 300 CFU on the initial dilution plate, a count of 3 × 10³ was used to enable statistical analysis. For statistical analysis, one-way ANOVA and Tukey’s multiple-comparison test was used to determine significant differences in colonization between biological groups (GraphPad Prism), using the null hypothesis that colonization rates between groups are the same. A P value of <0.05 was considered significant.

Data availability.

The RNAseq data set generated in this publication has been deposited in the NCBI database under BioProject number PRJNA590513.