INTRODUCTION
Chlamydia trachomatis infections in women differ greatly in the nature and intensity
of disease. A major conundrum with respect to chlamydial infections is why some women
develop clinical disease while the majority remain asymptomatic. Clearly, there can
be multiple reasons for this, including host genetic differences (1), the use of oral
contraceptives (2, 3), the stage of the menstrual cycle in which infection occurs
(4), and the presence of other risk factors such as bacterial vaginosis (5), but it
is more likely that a combination of factors is responsible. Another possible explanation
is that the infecting population consists of multiple genetic variants, with the pathological
phenotype being dependent upon the numerical representation of variants in the infecting
population. Miyairi and colleagues recognized differences among serovars of C. trachomatis,
reporting that ocular serovars had longer developmental cycles and lower growth rates
than genital serovars, and related this to the pathotype (6). Along the same line,
Kari et al. tested two different isolates of C. trachomatis serovar A and observed
that one isolate which had smaller plaques, slower growth, and increased sensitivity
to gamma interferon (IFN-γ) had a lower peak number of organisms with shorter infection
and lesser pathology upon conjunctival infection of nonhuman primates (7). Thus, it
is clear that even within a given serovar, there can be significant variations in
virulence.
The concept that variants within populations exist naturally was presented by Ramsey
and colleagues in their examination of the Nigg and Weiss strains of C. muridarum
which differ in virulence (8). We reported a similar observation using plaque-purified
C. caviae isolated from a conjunctival swab from an infected guinea pig (9). In an
elegant study, Sturdevant and colleagues infected mice with C. trachomatis serovar
D and found that mice resolved infections at vastly different times, from 10 to 77 days
(10). They found that when mice were infected with chlamydiae isolated from mice at
10 days and 49 days postinfection, the infection courses were consistently shorter
for the day 10 isolate and longer for the day 49 isolate. Moreover, the day 49 isolate
produced more pathology in the upper genital tract than the day 10 isolate. The authors
concluded that the “serovar D parental stock was a mixture of organisms varying in
virulence for the mouse” (10). The study results clearly indicate that infecting populations
consist of mixtures of variants with different pathological capabilities and that
the outcome of the infection is dependent upon the virulence of those variants within
the population.
Currently, there is no way to determine whether a woman infected with Chlamydia is
at risk for upper genital tract disease. It would be advantageous if one could identify
specific biomarkers that would help to predict the pathological outcome of the infection
early in the disease process to enable proactive treatment for the prevention of pelvic
inflammatory disease (PID). MicroRNAs (miRNAs) are abundant and evolutionarily conserved
noncoding small (~22-nucleotide) oligonucleotide molecules. They are immune modulators
that serve as an important link between innate and adaptive immune responses (11).
Dysregulation of miRNA expression has been linked to cardiovascular diseases, cancer,
infectious metabolic diseases, and other diseases (12
–
14). Most importantly, miRNAs have been shown to regulate chemokine cytokine responses
during different bacterial infections (15, 16). In addition, several studies have
suggested that miRNAs could be a new class of biomarkers for diagnostic and therapeutic
purposes under different pathological conditions (17, 18). Thus, we hypothesized that
different pathological variants of C. muridarum would elicit different miRNA profiles
and that such profiles would be predictive of the development or lack of development
of upper genital tract disease. In this study, we characterized two chlamydial variants
(C. muridarum Var001 [CmVar001] and CmVar004) that differ in their abilities to induce
upper genital tract pathology in mice by examining microRNA (miRNA) expression profiles
induced by these variants and their relationship to chemokine/cytokine responses within
24 h of infection.
RESULTS
In vitro growth characteristics of chlamydial variants.
We characterized plaque-purified variants of C. muridarum derived from genital tract
swabs of infected mice. These variants differed in their plaque size and in vitro
growth characteristics (Table 1). Growth curve analysis of variants showed different
phenotypes, with CmVar001 being the fastest growing variant followed by CmVar002,
while CmVar001.1, CmVar003, and CmVar004 had lower growth rates. The rate of elementary
body (EB) generation per developmental cycle was calculated by dividing the number
of infectious progeny from supernatants at 36 h by the number of EBs used to infect
the cultures (multiplicity of infection = 1). Similarly, the EB generation rate was
higher for CmVar001 and CmVar002 than for CmVar001.1, CmVar003, and CmVar004. Interestingly,
there was a 6-fold difference in EB production rates between CmVar001 and CmVar004,
and the doubling time results showed that the 2-fold EB increase time for CmVar004
was 136 min versus 78 min for CmVar001, indicating that, in vitro, CmVar004 grows
slower than CmVar001. While, with the exception of the plaque size, the data were
not evaluated statistically, as only two growth curve experiments were performed,
nevertheless, the data do suggest phenotypic differences among the variants.
TABLE 1
In vitro growth phenotype of C. muridarum M variants
a
C. muridarum variant
Plaque size (mm)
EB generation rate (fold increase)
Rate of 2-fold EB increase (min)
CmVar001
1.05 ± 0.29
25.4 ± 3.8
78
CmVar001.1
0.96 ± 0.08
14.4 ± 3.1
121
CmVar002
0.97 ± 0.25
22.3 ± 1.4
87
CmVar003
0.38 ± 0.25*
8.8 ± 0.7
128
CmVar004
0.55 ± 0.16*
3.8 ± 2.2
136
a
*, P < 0.00001 according to a two-tailed t test compared to CmVar001.
In vivo disease phenotype of infection with chlamydial variants.
Since the above data demonstrated that the variants have different growth phenotypes
in vitro, it was important to determine whether they would also differ in vivo in
a genital infection of mice. C57BL/6 and BALB/c mice were infected genitally with
3 × 105 inclusion-forming units (IFU), and the course of the infection was monitored
by measuring the number of IFU in genital tract swabs. No statistical significant
differences in the bacterial loads were noted during the primary course of infection
in C57BL/6 mice infected with C. muridarum variants CmVar001 and CmVar004 (2-factor
[day, group] analysis of variance with repeated measures), but a longer infection
was observed in mice infected with CmVar001 (see Fig. S1A in the supplemental material).
In contrast, CmVar001 elicited a significantly higher bacterial burden and longer
infection in BALB/c mice than in mice infected with CmVar004 (P < 0.0001) (see Fig. S1B).
Furthermore, there were no statistical significant differences between the results
determined for mice infected with CmVar001.1, CmVar002, or CmVar003 and those determined
for mice infected with CmVar004 or CmVar001.
An important measure of disease outcome is the development of hydrosalpinx, which
is associated with infertility. Hence; we assessed the presence of hydrosalpinx 35 days
after infection in each of two groups of five C57BL/6 mice infected with one of the
variants (Table 2). Because of major differences in hydrosalpinx development in mice
infected with CmVar001 versus those infected with CmVar004, the experiment with those
groups was repeated twice. Importantly, there was a significant difference with respect
to the development of hydrosalpinx between mice infected with the CmVar001 and CmVar004
variants (chi-square test with Yates correction, P < 0.0001), with only 4 of 15 mice
infected with CmVar004 developing unilateral or bilateral hydrosalpinx compared to
14 of 15 mice infected with CmVar001. Similarly, 6 of 30 (20%) oviducts in CmVar004-infected
mice were positive for hydrosalpinx in contrast to 25 of 30 (83.3%) oviducts in mice
infected with CmVar001. A similar difference was observed in BALB/c mice. Mice infected
with the CmVar001.1, CmVar003, and CmVar002 variants showed an intermediate phenotype
with regard to the production of hydrosalpinx, although all of the results were statistically
significantly different from those determined for the mice infected with CmVar001
(chi-square test with Yates correction, P < 0.01). In addition, CmVar001 also elicited
significantly more hydrosalpinx formation than did the original C. muridarum parent
stock from which the variants were derived (chi-square test with Yates correction,
P < 0.007) (Table 2).
TABLE 2
In vivo phenotype of chlamydial variants
a
C. muridarum variant
No. of mice positive for hydrosalpinx/total no. of mice (%)
No. of oviducts positive for hydrosalpinx/total no. of oviducts (%)
C57BL/6
BALB/c
C57BL/6
BALB/c
CmVar001
14/15 (93.3)
5/5 (100)
25/30 (83.3)
10/10 (100)
CmVar001.1
3/5 (60)
3/10 (30)†
CmVar002
4/5 (80)
4/10 (40)**
CmVar003
4/5 (80)
4/10 (40)**
CmVar004
4/15 (26.7)†
6/30 (20)#
5/10 (50)*
Parent stock
5/8 (62.5)
7/16 (44)†
a
P values compared to CmVar001 using the chi-square test with the Yates correction
(one tailed) in comparison to CmVar001 are indicated as follows: *, P < 0.05; **,
P < 0.02; †, P < 0.008; #, P < 0.0001.
The significant difference in the development of upper genital tract disease raised
the issue of whether CmVar001 and CmVar004 were reaching the upper tract at the same
rate. Therefore, C57BL/6 mice were infected intravaginally, and on day 10, oviducts
and the upper portion of the uterine horns were processed for the isolation and enumeration
of chlamydiae. Similar numbers of chlamydiae were found in the oviducts and upper
uterine horns of CmVar001- and CmVar004-infected mice, suggesting that the bacterial
load is not the cause of the reduced pathology observed in CmVar004-infected mice
(Fig. 1). We observed that 9 animals gave positive CmVar004 results in upper tract
isolations and that all 10 animals gave positive CmVar001 results in the upper genital
tract. We chose to use only CmVar001 and CmVar004 for further experiments because
they represented the opposite extremes in growth rates and pathology induction of
the variants.
FIG 1
Upper genital tract isolations of CmVar001 and CmVar004 variants indicate similar
numbers of IFU in oviducts and upper uterine horns. C57BL/6 mice were infected at
3 × 105 IFU, and cervical swabs were collected once every 3 days to determine the
IFU on a McCoy cell monolayer. On day 10, mice were euthanized to collect oviducts
and samples of an upper portion of uterine horn and of tissue were homogenized, sonicated,
and plated on HeLa cell monolayers for determination of IFU. Log10 IFU/ml values were
calculated from genital swabs and tissue isolations, and the averages of the results
determined for 10 animals were plotted as means ± standard deviations. The numbers
above the bars indicate the ratios of the number of mice positive for unilateral or
bilateral hydrosalpinx to the total number of mice. Cervical swab data indicated that
all mice were infected in the lower genital tract (data not shown). No statistical
significant differences between CmVar001 and CmVar004 upper tract isolations were
observed by a two-way analysis of variance (ANOVA).
Differential expression of chemokines and cytokines during chlamydial infection.
Production of specific chemokines and cytokines is inherent in the development of
a pathological response to chlamydial infection. Because there were differences in
the courses of infection and the development of pathology during CmVar001 and CmVar004
infection, we expected to observe a difference in the chemokine and cytokine profiles
elicited by each variant. To evaluate the chemokine and cytokine profile in the cervix,
we analyzed cervical tissue at 24 h postinfection for the expression of 15 key chemokines
and cytokines (tumor necrosis factor alpha [TNF-α], interleukin-1β [IL-1β], IL-6,
IL-10, CXCL10, CXCL1, CXCL2, CCL1, CCL2, CCL3, CCL4, CCL5, IFN-γ, IFN-β, and transforming
growth factor β [TGF-β]) which we previously showed to be upregulated 24 h after chlamydial
infection (19). C57BL/6 mice were infected intracervically rather than intravaginally
since intracervical inoculation places a large number of chlamydiae directly at the
target tissue and results in a synchronous infection for the first developmental cycle
(~42 to ~48 h). We performed an initial screen and determined that animals infected
with CmVar001 and CmVar004 had similar levels of chlamydial rRNA (rs16) (data not
shown) 24 h postinfection. We observed significantly lower levels of CXCL1, CXCL2,
TNF-α, CCL1, and CCL2 in cervical tissue from mice infected with CmVar004 than in
cervical tissue from mice infected with CmVar001 (Fig. 2A) (P < 0.05 according to
a one-tailed t test). A trend toward decreased expression of IL-1β, CXCL10, and IL-6
was also observed with CmVar004 infection, though it was not statistically significant
(data not shown). In addition, the other chemokines and cytokines measured were similarly
expressed in CmVar001 and CmVar004 (data not shown). These data indicated that CmVar004
induced lower levels of important molecules for the initiation of the inflammatory
response (TNF-α, CXCL1, CXCL2) than CmVar001, correlating with the relative pathogenicities
of each variant.
FIG 2
Chemokine and cytokine expression analyses from mice and BM1.11 cells. C57BL/6 mice
were infected with 1 × 107 IFU of CmVar001 or CmVar004 variants by intracervical inoculation.
(A) Expression levels of chemokine and cytokine transcripts are plotted as fold change
relative to control levels, and only CmVar001 and CmVar004 gene results that were
statistically significantly different are shown in the graph. Data are from an average
of 9 animals per group where CmVar004 infection had lower expression of CXCL1, CXCL2,
TNF-α, and CCL1 than CmVar001 infection (*, P < 0.05, one-tailed t test). We observed
that CXCL2 results were significant at 10% (**, P < 0.09) and that CCL2 results were
significant at 6% (**, P < 0.06). (B) Mouse oviduct epithelial cells (BM1.11) were
infected with CmVar001 and CmVar004 and variants. The graph is a representation of
three individual experiments, and data represent an average of three samples per each
group (*, P < 0.01, one-tailed t test). exp, expression.
Since it is likely that these molecules are produced by the infection of epithelial
cells, we wanted to determine if we would see a similar phenomenon in vitro in the
infection of mouse genital tract epithelial cells. Mouse oviduct epithelial cells
(Bm1.11) were infected with CmVar001 and CmVar004 variants at a multiplicity of infection
of 1. At 24 h postinfection, RNA was isolated and expression levels of various chemokines
and cytokines were determined. The data strongly supported the in vivo results, in
that CmVar004 induced significantly lower levels of CXCL1, CXCL2, TNF-α, and CCL2
with similar levels of rs16 than did CmVar001 (Fig. 2B) (P < 0.05). Therefore, both
the in vitro and in vivo chemokine and cytokine expression data correlated with the
pathological outcomes of the attenuated CmVar004 and the virulent CmVar001 variants.
These data indicate that the early chemokine and cytokine responses elicited in mice
within the first developmental cycle of chlamydial infection by defined virulent and
attenuated variants are predictive of the outcome of the disease, resulting in more
and less upper tract pathology, respectively.
Differential expression of miRNA(s) during chlamydial infection.
In mammalian cells, miRNAs play a pivotal role in regulation of different processes
such as development, metabolism, and inflammation as well as in modulating chemokine
and cytokine expression (15, 21
–
26). However, the role inflammation-related miRNAs play during chlamydial pathogenesis
and their correlation to chemokine and cytokine expression has not been studied. The
population of miRNAs elicited by CmVar001, CmVar004, and sham inoculation of control
mice (4 C57BL/6 mice for each group) infected by the intracervical route was first
explored by relatively low-coverage Illumina MiSeq sequencing. This small-RNA sequencing
and bioinformatic analysis identified 2,935 uniquely mapped mouse miRNAs (data not
shown). Using stringent cutoff values, differential expression analysis comparing
the CmVar001- and CmVar004-infected mice to the sham-inoculated mice identified five
(CmVar004 versus controls) and four (CmVar001 versus controls) miRNAs that were significantly
differentially expressed (Table 3). Two miRNAs, miR223-3p and miR18a-5p, were common
to CmVar001 and CmVar004. Further, differential expression analysis of CmVar001 versus
CmVar004 showed that miR223-3p was overexpressed in CmVar004-infected mice (Table 3).
The relatively low-coverage small-RNA sequencing analysis results presented above
provided evidence that CmVar001 and CmVar004 elicit different host miRNA responses.
TABLE 3
DEseq of miRNAs discovered by sequencing
a
Comparison and miRNA
MRC
Log fold change
P value
FDR
Control versus CmVar001
CmVar001
Controls
mmu-miR-223-3p
78.00
29.73
−1.39
0.0000
0.0003
mmu-miR-203-3p*
430.87
121.22
−1.83
0.0000
0.0003
mmu-miR-18a-5p*
22.29
5.64
−1.98
0.0000
0.0021
mmu-miR-215-5p
12.51
0.00
0.0000
0.0000
Control vs CmVar004
CmVar004
Controls
mmu-miR-98-5p
137.96
57.49
−1.26
0.0000
0.0001
mmu-miR-21a-3p
146.06
56.68
−1.37
0.0000
0.0000
mmu-miR-155-5p
34.77
8.91
−1.96
0.0000
0.0001
mmu-miR-18a-5p*
23.50
5.41
−2.12
0.0000
0.0001
mmu-miR-223-3p*
175.75
28.51
−2.62
0.0000
0.0000
CmVar001 versus CmVar004
CmVar001
CmVar004
mmu-miR-223-3p
61.00
100.00
1.00
0.0002
0.04
mmu-miR-203-3p
300.00
100.00
−2.00
0.0001
0.02
a
MRC, mean read count; FDR, false discovery rate. Cutoff of FDR < 0.05; minimum 10
reads; upper fold change = 1; *, lower fold change = −1. DEseq, differential expression
levels.
In order to confirm sequencing results and also focus on the miRNAs involved in immunopathogenesis,
we measured the expression profile of 134 miRNAs by real-time PCR, at 24 h after intracervical
infection, using 5 C57BL/6 mice per group. We observed differential expression of
miRNAs during CmVar001 and CmVar004 infection in comparison to sham-inoculated control
mice. There were distinct differences in the types and levels of the miRNAs expressed
in mice infected by CmVar001 and CmVar004 compared to controls. Data analyses of miRNA
for 1.5-fold or higher differences between CmVar001 and CmVar004 indicated that 12
miRNAs were expressed at higher levels with CmVar004 infection than with CmVar001.
Among the 12, we observed that 10 miRNAs (miR-135a-5p, miR298-5p, miR142-5p, miR223-3p,
miR147-3p, miR105, miR132-3p, miR142-3p, miR155-5p, and miR-410-3p) were upregulated
in mice infected with the avirulent CmVar004 variant in comparison to mice infected
with the virulent CmVar001 variant (Fig. 3 and Table 4). Two miRNAs, miR-299a-3p and
miR-325-3p, were downregulated in both infections; however, the downregulation was
significantly greater in animals infected with CmVar001. As shown in Table 4, miR-155-5p,
miR-142-3p, miR-132-3p, miR105, miR223-3p, and mir-147-3p regulate chemokine and cytokine
responses as well as modulating Toll-like receptor 2 (TLR2) expression under different
pathological conditions (15, 21
–
25). These data indicate that miRNAs expressed at higher levels during CmVar004 infection
are associated with the downregulation of chemokine and cytokine expression in mice,
resulting in lesser upper genital tract pathology during a CmVar004 infection. Conversely,
the low expression level of these inflammatory miRNAs in mice infected with the virulent
CmVar001 variant is associated with the increased incidence of hydrosalpinx in those
mice.
FIG 3
miRNA expression analyses from mice. C57BL/6 mice were infected with 1 × 107 IFU of
CmVar001 or CmVar004 variants by intracervical inoculation, and the miRNA expression
profile was measured as described in Materials and Methods. Results represent an average
of 5 to 7 animals per group; statistical significance was determined by the Wilcoxon
rank sum one-tailed test (Bonferroni correction), with P < 0.05 being considered significant.
Data analyses of miRNA for determinations of differences between CmVar004 and CmVar001
results at a level of 1.5-fold or higher indicate that 12 miRNAs were expressed at
higher levels with CmVar004 infection than with CmVar001 infection and that the results
for 6 miRNAs were statistically significant as well.
TABLE 4
In vivo expression profile of miRNA
a
microRNA
Fold increase (CmVar004 > CmVar001)
One-tailed test P value (CmVar004 > CmVar001)
Related pathway(s) and direction of regulation (source)
miR-135a-5p
1.55
0.003
ND
miR-299a-3p
1.63
0.05
ND
miR-155-5p
1.64
N.S.
CXCL1↓ (15)
miR-142-3p
1.75
N.S.
IRAK1 (23)
miR-132-3p
1.76
N.S.
IRAK4↓, TNF-α (50)
miR-325-3p
1.81
N.S.
ND
miR-105
1.83
N.S.
TLR2↓, TNF-α, IL-6 (21)
miR-410-3p
1.86
N.S.
ND
miR-223-3p
2.04
0.05
IL-1B↓, IL-6↓, TNF-α (22)
miR-298-5p
2.10
0.02
ND
miR-142-5p
2.23
0.02
ND
miR-147-3p
4.54
0.05
TLR2↓, TNF-α, IL-6 (24)
a
N.S., not significant; ND, not determined.
Comparative genome sequencing of chlamydial variants.
High-quality draft genomes were obtained for each C. muridarum plaque variant by deep
sequencing (see Table S1 in the supplemental material). A draft assembly of the conserved
C. muridarum plasmid was also recovered for each plaque variant, excluding the possibility
that the attenuated growth and virulence phenotypes were caused by plasmid-free C. muridarum
(27). Using the published C. muridarum Nigg genome (Genbank accession no. AE002160)
as a reference, comparative genome analysis identified 52 high-quality single nucleotide
polymorphisms (SNPs) and insertions and deletions (indels), supported by very deep
sequence coverage (>2,600×), within the draft plaque variant genomes, including synonymous
and nonsynonymous changes, likely resulting in several frameshift mutations (Table 5;
see also Table S2). The pattern for 18 of these SNPs or indels within or between predicted
genes across the plaque variants correlates with the observed phenotypes of differential
growth rates and gross pathology (Table 5; see also Table S2). Nine of these SNPs
and indels occur within the ompA gene (TC0052) that encodes the major outer membrane
protein (MOMP). These mutations cause an amino acid deletion within the MOMP third
variable domain (VD3) in the CmVar001 variant (Table 5 and data not shown). In the
CmVar004 variant, these mutations cause a double amino acid deletion and a glycine-to-cysteine
substitution within MOMP VD4 (Table 5 and data not shown). The glycine-to-cysteine
substitution may impact MOMP structure and intramolecular complexing through disulfide
cross-linking (28, 29). Phenotype-correlating mutations across plaque variants are
observed in several other genes, including TC0412 (homolog of CT135), TC0155 [3′(2′),5′-bisphosphate
nucleotidase], and TC0682 (tetraacyldisaccharide 4′-kinase) (Table 5; see also Table S2).
Phenotype-correlating mutations are also found within several genes with no known
function (annotated as conserved hypothetical genes); this suggests that these genes
may have functions that are related to the observed variant phenotypes. Notably, disruption
of CT135 was recently shown to reduce infectivity in a C. trachomatis mouse model
of infection (30). Thus, comparative genomic analysis of plaque-purified variants
identified mutations that may explain the virulence characteristics of variants in
a particular stock.
TABLE 5
SNPs and indels identified in variants with potential phenotype correlation relative
to the Nigg genome (Genbank accession no. AE002160)
a
SNP location
Gene locant: protein encoded
Gene start (NT)
Gene stop (NT)
CmVar001
CmVar001.1
CmVar002
CmVar003
CmVar004
NT change
AA change
NT change
AA change
NT change
AA change
NT change
AA change
NT change
AA change
58847
TC_0052: major outer membrane protein, porin
58653
59816
C→A
G→C
C→A
G→C
58899
TC_0052: major outer membrane protein, porin
58653
59816
A→
DEL
A→
DEL
58900
TC_0052: major outer membrane protein, porin
58653
59816
A→
DEL
A→
DEL
58901
TC_0052: major outer membrane protein, porin
58653
59816
G→
DEL
G→
DEL
58902
TC_0052: major outer membrane protein, porin
58653
59816
C→
DEL
C→
DEL
58903
TC_0052: major outer membrane protein, porin
58653
59816
T→
DEL
T→
DEL
59061
TC_0052: major outer membrane protein, porin
58653
59816
T→
DEL
T→
DEL
T→
DEL
59062
TC_0052: major outer membrane protein, porin
58653
59816
G→
DEL
G→
DEL
G→
DEL
59063
TC_0052: major outer membrane protein, porin
58653
59816
T→
DEL
T→
DEL
T→
DEL
187432
TC_0155: 3′(2′),5′-bisphosphate nucleotidase, putative
187163
188212
G→A
H→Y
G→A
H→Y
G→A
H→Y
187984
TC_0155: 3′(2′),5′-bisphosphate nucleotidase, putative
187163
188212
C→T
G→S
381736
Intergenic: NA
NA
NA
T→A
NA→NA
473191
TC_0412: conserved hypothetical protein
472757
473854
→A
FS
→A
FS
473705
TC_0412: conserved hypothetical protein
472757
473854
G→
FS
G→
FS
G→
FS
798169
TC_0668: conserved hypothetical protein
797014
798240
T→G
F→V
814976
TC_0682: tetraacyldisaccharide 4′-kinase
814899
816008
C→A
V→L
C→A
V→L
C→A
V→L
1004276
TC_0867: conserved hypothetical protein
1002965
1004440
T→C
S→P
T→C
S→P
T→C
S→P
T→C
S→P
1062351
Intergenic
NA
NA
C→A
NA→NA
C→A
NA→NA
C→A
NA→NA
a
Mutations common to all variants are excluded (see Table S1 in the supplemental material
for the entire list). Chlamydial variants are arranged from highest growth rate to
lowest (left to right). Gross pathology results also followed this pattern. FS, frameshift;
DEL, deletion; INS, insertion; NA, not applicable (intergenic); NT change, nucleotide
change; AA change, amino acid change.
DISCUSSION
Chlamydial genital infections, as in many infectious diseases, vary greatly in the
clinical outcome of the disease, ranging from subclinical infection with little or
no sequelae to full-blown cervicitis followed by pelvic inflammatory disease resulting
in tubal obstruction, ectopic pregnancy, or infertility. Using the C. muridarum mouse
model of genital infection, we undertook studies to expand upon the concept that a
given chlamydial population consists of variants with different levels of pathogenic
potential. In support of this concept, 5 plaque-purified isolates from mice infected
with the less virulent Nigg strain were characterized for their pathology-inducing
capabilities. Interestingly, even though the choosing of the plaques for evaluation
was purely random, a range of pathology-associated phenotypes was identified. While
there was little difference in the infection courses, two variants, CmVar001 and CmVar004,
differed dramatically in the ability to elicit hydrosalpinx. This ability was reflected
in their plaque size, growth rate, and chemokine profile, with the more virulent variant
having an increased growth rate and increased production of proinflammatory chemokines.
Deep genome sequencing of these variants identified patterns of mutations that correlated
with these observed growth and virulence phenotypes, but no definitive conclusions
can be drawn linking a specific mutation to the virulence phenotype. Thus, even though
the initial infection population of chlamydiae was relatively attenuated, variants
were present in that population with a wide range of pathogenic potential. It is interesting
that a virulent variant was contained in the parent population but that virulence
was not reflected in the overall pathological phenotype of the parent population.
Whether the pathogenicity of a given population represents the sum total or the average
of the pathogenicities of its constituent variants cannot be concluded at this point.
What is clear is that the virulence phenotype of a population does not necessarily
reflect the virulence phenotype of all of the individual variants within that population.
Using C. caviae in the guinea pig model of chlamydial genital infection, we previously
observed that when a wild-type variant, capable of eliciting a strong inflammatory
response, and a less virulent and slower-growing variant eliciting significantly less
inflammation are inoculated together into the animal, the outcome of the infection
resembles that of the virulent variant (31). Moreover, even though equal numbers of
the variants are inoculated, the wild-type variant becomes dominant by the end of
the infection (31). This suggested that the composition of a chlamydial population
is dynamic and constantly in flux, governed by host genetics and microenvironment
of the infection site, among other factors.
If the composition of individual variants within a chlamydial population defines the
potential pathological outcome of the disease, the identification of specific variants
associated with more-severe disease might aid in determining the risk of upper tract
disease. However, while it is technically feasible to genotype a chlamydial population
from a swab (32
–
34), it is not yet logistically and economically feasible to use this methodology
in a clinic setting. Moreover, we are still a long way from being able to definitively
state which gene or group of genes determines chlamydial virulence capability. It
is currently more practical to determine if there is a particular host factor(s) that
demonstrates the potential for more-severe disease, particularly a more intense acute
inflammatory response. The number of polymorphonuclear leukocytes (PMNs) present in
a cervical scraping is an inexact measure and has been of limited diagnostic use for
PID. Nevertheless, it is interesting that the virulent CmVar001 variant results in
the expression of significantly greater levels of key inflammatory cytokines and chemokines,
including TNF-α, CXCL1, CXCL2, and CCL1, than the attenuated CmVar004 variant both
in vitro and in vivo. Importantly, these differences were noted at 24 h after infection,
when chlamydiae were still well within their first developmental cycle and when PMNs
were just beginning to reach the infected cells in the epithelium (35). The reason
for this inflammatory difference between the virulent and attenuated variants is not
known. It could be related to the higher growth rate and more productive infection
associated with the virulent CmVar001 variant; however, if that were the case, it
would be difficult to establish parameters to predict the possibility of different
degrees of upper genital tract pathology, especially because of dilution effects of
the population of multiple variants and the very strong likelihood that individuals
are infected with differing numbers of organisms. Moreover, unlike our experimental
models, women are likely exposed multiple times in an ongoing relationship.
An alternative set of biomarkers is that of host miRNA differentially expressed in
response to infection. A recent report on chlamydial conjunctival infection from human
samples showed differential expression of miRNAs (36). However, it did not show any
evidence of specific miRNAs in predicting pathology, as the study compared samples
from individuals with scarring and those without inflammation in contrast to an early
conjunctivitis stage of trachoma. Our preliminary sequencing of miRNAs in mouse cervical
tissues elicited by CmVar004 and CmVar001 infections demonstrated differential expression
levels, and differences were observed in chemokine and cytokine responses. With this
evidence, we then focused on the expression of 134 inflammation-related miRNAs in
order to determine if the differences seen in chemokine and cytokine responses between
the CmVar004 and CmVar001 variants would also be reflected in specific quantitative
real-time PCR (qRT-PCR) miRNA expression profiles within 24 h of infection. Interestingly,
12 different miRNAs (miR-135a-5p, miR-298-5p, miR-142-5p, miR-223-3p, miR-299a-3p,
miR-147-3p, miR-105, miR-325-3p, miR-132-3p, miR-142-3p, miR155-5p, and miR-410-3p)
were overexpressed following infection with the attenuated variant, CmVar004, in contrast
to infection with the virulent variant, CmVar001. The results for six of these (miR-135a-5p,
miR299a-3p, miR223-3p, miR298-5p, miR142-5p, and miR147-3p) are also statistically
significant. The results with respect to differential expression of miR223-3p were
also supported by the sequencing data. To our knowledge, this is the first report
demonstrating that differences in the early miRNA profile may be predictive of the
outcome of the infection with regard to upper genital tract pathology. Because of
the large number of miRNAs with multiple regulatory functions in the immune response,
it may be possible to define a particular profile from a genital tract specimen early
in the course of infection that will be prognostic for the severity of disease, allowing
physicians to treat those infections more aggressively.
The significant upregulation of 12 different miRNAs by the attenuated variant in comparison
to the virulent variant supports our observation of a lower inflammatory chemokine
and cytokine response elicited by the avirulent variant. These results suggest that
the interaction of the attenuated variant with the host cell initiates a pathway(s)
that results in the production of miRNAs that downregulate inflammatory responses.
Conversely, the virulent variant may also be initiating pathways that suppress the
expression of those miRNAs, resulting in increased inflammatory chemokine and or cytokine
expression. Among those known miRNAs that are upregulated in infection with the avirulent
variant, miR-155, miR-142-3p, miR-147-3p, miR-105, and miR-132 have been shown to
downregulate chemokines CXCL1, TNF-α, and IL-6 and also to modulate TLR2 expression
under different pathological conditions (15, 16, 24). We also observed a lower CXCL1
and TNF-α response in infection with the attenuated variant in addition to the CXCL2,
CCL1, and CCL2 results noted. It is interesting that the activation of the TLR2 pathway,
in contrast to the TLR4 pathway, is associated with the pathological response in chlamydial
genital infections (37). It is possible that activation of certain miRNAs by the avirulent
variant blocks or downregulates the TLR2 pathway, resulting in a lower pathological
response. Nevertheless, the exact mechanism(s) by which chlamydiae initiate these
processes, including the role of the observed pattern of mutations within the variant
genomes, remains to be determined.
Several reports have described the role for miRNAs in regulating the chemokine and
cytokine responses in bacterial induced inflammation. For example, molecules such
as lipopolysaccharide (LPS), TNF-α, and IFN-γ have been shown to induce miRNA expression
(26, 38, 39). During Helicobacter pylori infection, overexpression of miR-155 reduces
the expression of IL-8 and growth-related oncogene–α (GRO-α) (15) in human gastric
epithelial mucosa cells. A total of 28 miRNAs are upregulated and 2 are downregulated
in patients with Mycobacterium tuberculosis infection in comparison to healthy individuals.
Among those, miR-144* is possibly inhibitory for TNF-α and IFN-γ production and T-cell
proliferation, suggesting the involvement of miRNA in antituberculosis immunity (16).
miR-142-3p downregulated interleukin-1 receptor-associated kinase 1 (IRAK-1) expression,
resulting in downregulation of NF-κB, TNF-α, and IL-6 during Mycobacterium bovis bacillus
Calmette-Guérin (BCG) infection in mouse macrophages (23). In a brain inflammation
model, miR-132 was found to have an anti-inflammatory effect by targeting mRNA of
acetyl cholinesterase enzyme, which results in increased levels of acetyl choline,
an important inhibitor of peripheral inflammation (25). In miR-223−/− mouse macrophages,
significantly elevated levels of IL-1β, IL-6, and TNF-α were observed in comparison
to those seen with wild-type macrophages during LPS stimulation, indicating that miR-223
plays a role in anti-inflammatory processes (22). miR-147 is induced by stimulation
of multiple TLRs, and its overexpression in mouse macrophages decreased TNF-α and
IL-6 expression induced by TLR2, TLR3, and TLR4 ligands [PAM3CSK4, poly(I ⋅ C), and
LPS], indicating that miR-147 functions in a negative-feedback mechanism to regulate
inflammation (24). In addition to this, miR-105 has also been shown to modulate TLR2
expression in human keratinocytes (21). All these reports indicate that miRNAs play
a regulatory role in chemokine and cytokine expression.
While we have presented data to demonstrate that the miRNA profile is able to distinguish
between virulent and attenuated variants in our model, the more important point is
that regardless of the specific variants, it is possible that the miRNA profile may
be able to serve as a predicative biomarker for disease severity. From the viewpoint
of human medicine, the nature of the actual variant causing the disease is irrelevant;
it is more important to have a means to determine if there is an increased risk of
severe outcome so that appropriate therapy can be initiated before irreversible pathology
occurs. Moreover, the predictive value of biomarkers such as miRNAs early in the infection
course may be in their use as a measure to determine the effectiveness of a chlamydial
vaccine. Clearly, it is not feasible to use PID or infertility as an outcome for the
measure of vaccine effectiveness; rather, the miRNA profile may have value as a correlate
for PID or infertility or the lack thereof.
In summary, we have presented evidence that an infecting chlamydial population consists
of multiple genetic variants with distinct virulence phenotypes. Moreover, infection
with virulent versus attenuated phenotypes can be detected early in the course of
the infection by the level of expression of specific miRNAs that can regulate various
aspects of the immune response and correlate with the chemokine and cytokine response
elicited by the variants. Importantly, the miRNA expression profile assessed at 24 h
after infection is predictive of the eventual development of hydrosalpinx, the major
pathological outcome for chlamydial infection in females. While 24 h is not a practical
time point to use in analyses of human infections, the results do suggest that observable
differences in the miRNA response can be detected early in the infection course. Considering
the impact of miRNAs on the regulation of the immune response and the association
of specific miRNAs with infection by defined chlamydial pathotypes, they may hold
promise as early-stage biomarkers to be used for prognostics in chlamydial genital
infections.
MATERIALS AND METHODS
Cell lines.
Oviduct epithelial cells (Bm1.11) were provided by Raymond Johnson (Indiana University,
Indianapolis, IN) and were cultured in Ham’s F-12 media supplemented with 10% fetal
bovine serum (FBS), 12.5 ng/ml human recombinant keratinocyte growth factor (Sigma-Aldrich,
St. Louis, MO), 2 mM GlutaMAX (Invitrogen), and 50 μg/ml gentamicin. HeLa cells and
McCoy cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% FBS,
1 mM glutamine, 100 μM nonessential amino acids (Invitrogen), and 50 μg/ml gentamicin.
Chlamydial infection and experimental animals.
Mice were infected with C. muridarum strain Nigg, which has been passaged in our laboratory
since 1977, when it was obtained from the American Type Culture Collection as a yolk
sac preparation. The Nigg stock, as reported by Ramsey and coworkers, consists of
multiple variants, some of which have the same set of SNPs as seen in the genome sequencing
of the variants obtained in vivo in this study (K. H. Ramsey, personal communication).
To obtain chlamydial variants, nude mice were infected with 3 × 105 inclusion-forming
units (IFU) of C. muridarum and swabs were collected on day 6 and day 9. These swabs
were plated for plaques, and random plaques were chosen to be subjected to subplaque
production and expansion. CmVar001.1, CmVar003, and CmVar004 were collected from the
day 6 swab, and CmVar002 and CmVar001 variants were collected from the day 9 swab
from the same mouse. In vitro growth curve analysis was performed as described previously
(8) to determine the differences in growth phenotype. Eight-week-old C57BL/6 and BALB/c
mice were purchased from Jackson Laboratories (Bar Harbor, ME) and Harlan-Sprague
Dawley (Indianapolis, IN), respectively. Mice were given a 2.5-mg subcutaneous injection
of medroxyprogesterone (Depo-Provera; Upjohn, Kalamazoo, MI) 7 days prior to infection
to place the mice in a state of anestrus. Mice were infected intravaginally with 3
× 105 IFU (40) or intracervically with 1 × 107 IFU as described previously (19). Cervices
were collected from mice infected intracervically for total RNA isolation 24 h after
infection. To monitor the course of infection, swabs were collected from mice infected
intravaginally every 3 days for determination of IFU. On day 35, all mice were euthanized
for gross pathology observation. All protocols were approved by the Institutional
Animal Care and Use Committee at the University of Arkansas for Medical Sciences.
Chlamydial culture.
Chlamydiae were quantified from the genital tract using a Dacro swab and then cultured
in McCoy cells according to standard procedures. In order to quantify the number of
chlamydiae from tissues, oviducts and a small portion of the upper uterine horn were
dissected and placed in 1 ml of 2-sucrose-phosphate buffer transport medium with 0.1 mg/ml
gentamicin, 0.2 mg/ml vancomycin, and 2.5 µg/ml of amphotericin B (Fungizone). The
tissues were homogenized and then sonicated for 1 min. After centrifugation to pellet
the large-cell debris, the supernatants were diluted with Eagle’s minimal essential
medium supplemented with 10% fetal bovine serum (FBS) and 5% glucose (pH 7.4) and
inoculated onto confluent HeLa cell monolayers for the determination of IFU. Chlamydial
inclusions were stained with C. muridarum-specific immune serum and with a fluorescent
secondary anti-mouse IgG antibody (Alexa Fluor 488).
Genome sequencing.
Genomic DNAs extracted from plaque-purified C. muridarum variants were subjected to
library preparation and next-generation whole-genome sequencing (paired end, 100 bp)
using a single channel of an Illumina HiSeq 2500 flow cell (1/8 channel per variant).
Sequence reads were assembled using CLC Genomics Workbench software (version 6.0.2)
by mapping reads to the C. muridarum Nigg (Genbank accession no. AE002160) reference
genome (similarity fraction = 0.85) (41) (see Table S1). For each sequenced isolate,
consensus contigs were extracted and ordered into a pseudomolecule, using the reference
genome for contig order. Gene identification and annotation were performed as previously
described (42). SNPs and indels between C. muridarum variants and the reference C. muridarum
genome were identified as previously described using a NUCmer-based bioinformatics
pipeline (43
–
45).
Total RNA isolation and expression analysis of miRNA and mRNA transcripts.
Total RNA was isolated from cervices and cell samples by using an miRNeasy kit (Qiagen)
after 24 h of infection with CmVar004 or CmVar001. Animals in a control group were
sham inoculated. The concentration of RNA was determined by Nanodrop, and absorbance
at 260 and 280 and absorbance at 260 and 230 were taken into consideration to check
for RNA quality. RNA was reverse transcribed with SuperScript III enzyme (Invitrogen)
according to the manufacturer’s instructions using random hexamer and dT for priming.
Quantitative PCR was performed on samples using IQ-SYBR mix (Bio-Rad) and a CFX96
PCR detection system (Bio-Rad). Levels of chlamydial rRNA (rs16) were measured to
determine chlamydial growth, and samples that showed rs16 levels similar to those
determined for CmVar004 and CmVar001 were chosen for chemokine and cytokine (9 individual
animals per [control, CmVar004, or CmVar001] group) analyses and miRNA analyses (4
to 5/group). All primers were designed using Beacon Design software (Bio-Rad).
To analyze miRNAs by sequencing, total RNAs of 4 samples from each group were subjected
to small-RNA sequencing. To analyze miRNA by sequencing, total RNA was size selected
and sequenced. Total RNA quality was assessed using an RNA Nano chip on an Agilent
BioAnalyzer 2100 instrument (Agilent Technologies) prior to library preparation. Sequencing
libraries were constructed using a DGE Small RNA Library Preparation kit (Illumina).
Library preparations were validated by the use of a BioAnalyzer 2100 system using
High Sensitivity DNA chips (Agilent Technologies) prior to sequencing. Small-RNA sequencing
was performed using an Illumina MiSeq instrument. miRNAs were mapped to the mouse
reference genome (GRCm38.70) using Bow tie v 0.12.9 (maximum number of mismatches
= 2) (46). The number of reads mapped to each miRNA was counted by HTSeq against the
annotated mouse miRNA gff3 file downloaded from miRBase (release 20) (47). Data normalization
and differential gene expression analyses were performed using DESeq (48) and EdgeR
(49), independently comparing C. muridarum CmVar004-infected and CmVar001-infected
replicates to uninfected controls. Up- and downregulated miRNAs were filtered for
significantly differentially expressed miRNAs using the following parameters: a cutoff
false-discovery rate (FDR) = <0.05; minimum read count = 10; an upper fold change
value of 1; and a lower fold change value of −1. To confirm the miRNA sequencing results
and also to focus on immune-related miRNAs, we measured miRNA expression levels for
5 samples per group to run miRNA arrays from Qiagen that contained 134 different miRNAs
related to inflammation (combination of inflammatory response and autoimmunity and
immunopathogenesis miScript miRNA PCR arrays). Data analyses were performed using
Web-based software provided by SABiosciences. miRNA expression analyses were performed
during CmVar004 and CmVar001 infection with respect to control animals, and small
nucleolar RNAs (SNORDs) were also used as internal controls (SNORD 61, 68, 72, 95,
and 96A), which had the least variability of all the samples with respect to CmVar004
and CmVar001 infection. Data are presented as a clustergram and indicate fold change
differences in regulation (up or down) of miRNAs. Data obtained by comparisons to
the control were used to determine the fold change differences between CmVar004 and
CmVar001 with respect to expression of individual miRNAs by using the one-tailed t test.
Statistical analysis.
Data are presented as the mean ± 1 standard deviation. The course of the infection
among variants was analyzed with a 2-factor (days, group) analysis of variance with
repeated measures. Statistical analyses for chemokine and cytokine data were carried
out only with the measurements that showed a higher fold increase with the CmVar001
variant than with the CmVar004 variant, and data were analyzed by one-tailed t test.
MicroRNAs that showed fold increases in mice that were infected with CmVar004 in comparison
to those infected with CmVar001 were analyzed for statistical significance by a nonparametric
one-tailed (with Bonferroni correction) Wilcoxon rank sum test.
Nucleotide sequence accession numbers.
The whole genome shotgun project has been deposited at DDBJ/EMBL/GenBank under accession
numbers JNON00000000, JNOO00000000, JNOP00000000, JNOQ00000000, and JNOR00000000.
SUPPLEMENTAL MATERIAL
Figure S1
In vivo course of infection of chlamydial variants (CmVar004 and CmVar001) in two
different strains of mice (C57BL/6 and BALB/c mice). Download
Figure S1, PPTX file, 0.1 MB
Figure S2
Alignment of C. muridarum MOMP (ompA) from Nigg, CmVar001, and CmVar004. Download
Figure S2, PPTX file, 0.1 MB
Table S1
Sequencing metrics for plaque variants and the original source culture. Reads are
mapped onto the published C. muridarum Nigg genome (Genbank accession no. ).
Table S1, PPTX file, 0.1 MB.
Table S2
Comparative genome sequences of C. muridarum variants.
Table S2, PPTX file, 0.1 MB.