OUP user menu

Long-term persistence of Coxiella burnetii after acute primary Q fever

B.P. Marmion, P.A. Storm, J.G. Ayres, L. Semendric, L. Mathews, W. Winslow, M. Turra, R.J. Harris
DOI: http://dx.doi.org/10.1093/qjmed/hci009 7-20 First published online: 29 December 2004


Background: Long-term persistence of C. burnetii in infected animals was established in the 1950s and 60s, but the implications for human Q fever are not fully explored.

Aim: To compare the prevalence of markers of infection in a cohort of Q fever patients in Australia (up to 5 years after infection) with those in the 1989 Birmingham cohort (12 years after infection).

Design: Case follow-up study.

Methods: C. burnetii was tested for by: (i) antibodies to Phase 1 and 2 antigens in the three immunoglobulin classes; (ii) detection of DNA in bone marrow and peripheral blood mononuclear cells by PCR assays directed against several different targets in the genome; and (iii) attempts to isolate coxiellas in cell culture or mice from PCR-positive samples. Amplicon specificity was verified by fluorometric probing and by sequencing. Cross-contamination was excluded by extensive use of non-template controls, and in particular by the use of certain IS1111a target sequences.

Results: Irrespective of clinical state, both groups remained seropositive, principally exhibiting medium levels of IgG antibody against C. burnetii Phase 2 antigen. C. burnetii genomic DNA was detected by PCR in 65% of bone marrow aspirates from Australian patients and ∼88% of Birmingham patients. No coxiella were isolated from PCR positive samples.

Discussion: We propose a provisional model for persistence. In Q fever without sequelae, the process is largely confined to the bone marrow. In Q fever fatigue syndrome (QFS), it is modulated by the patient's immunogenetic background to give higher levels of coxiella genomes in bone marrow and increased shedding into the peripheral blood. In Q fever endocarditis, late pregnancy, or during iatrogenic or other immunosuppression, the multiplication cycle is prolonged, and a potential source of live organisms.


The long-term persistence of Coxiella burnetii in infected domestic ruminants or experimentally infected laboratory animals and its recrudescence during pregnancy or immunosuppression was established in the 1950s and 60s (summary: Harris et al.1). The implications of these findings for human Q fever are not fully explored. Acute primary Q fever is still regarded as essentially benign without serious sequelae. The first discordant finding was subacute Q fever endocarditis. Marmion and Stoker in the UK and Huebner and colleagues in California identified possible but incompletely investigated cases in the 1940s.2 Subsequently, a detailed study of a UK patient with aortic valve endocarditis ∼10 months after acute Q fever revealed live C. burnetii in the blood, along with high titres of CF antibody to both C. burnetii Phase 1 and 2 antigens. Extracellular coxiella microcolonies were found in the aortic valve vegetations, which also had an infectivity titre of 10−6 for guinea pigs.3,,4 Numerous cases were reported in the UK, Europe and later in the USA.5,,6 In the 1950s, a further link with the natural history of infection in animals was made by Syrucek et al.,7 who isolated C. burnetii from placentas of parturient women who had been infected early in pregnancy, or even as long as three or more years before parturition.

More recently the concept of chronic or recrudescent Q fever in humans has been enlarged by culture of C. burnetii, or detection of its genome by PCR, in infections of bone,8,,9 testis and the male genital tract,10–12 liver,13,,14 lung and pleura15,,16 or soft tissue.17 Recently, Ayres and colleagues in the UK18 and Marmion et al.19 in Australia described another sequel to acute Q fever: namely, a long lasting, often highly disabling post-infection fatigue syndrome (QFS) in ∼10% of patients. We view QFS as a failure in some genetically predisposed individuals to ‘switch off’ residual elements of the cytokine-mediated acute-phase reaction in the face of the persistence of C. burnetii cells after the initial attack. This deficiency is illustrated by an inappropriate, increased liberation of some cytokines on stimulation of short-term cultures of QFS peripheral blood mononuclear cells (PBMC) with Q fever antigens.20 In preliminary studies,1 bone marrow aspirates and PBMC were collected from Australian QFS patients at varying intervals up to 5 years after initial infection. PCR amplification of target sequences in the superoxide dismutase (SOD) or IS1111a genes of the coxiella genome gave positive results in 65% of bone marrow aspirates and 17% of PBMCs. Bone marrow from unrelated disease was negative.

Despite these pointers to C. burnetii persistence in humans and the common description of Q fever as a ‘non-sterilizing’ infection, questions remain. Are the genomic sequences detected by PCR located in living, dormant (replication-defective) or dead coxiella cells? Is the persistent coxiella the small or large cell variant, ‘spore’ or another infra-form?21,,22 Which host organ(s) and tissue cells support the persistent focus? Is persistence limited to patients with QFS or other chronic illness? Or does it occur at low level after all initial Q fever infections, but is only symptomatic in those who fail to regulate the level of persistence and the intensity and nature of cell-mediated immune response?

The 1989 outbreak of 149 Q fever cases in Birmingham UK offered an unusual opportunity to study some of these questions. The Q fever infection was probably wind-borne from flocks of infected parturient sheep on pastures just south of the suburb of Solihull. Q fever cases occurred over a distance of 10 miles into the urban area.23 The cohort of Q fever patients was exposed once only, unlike abattoir or agricultural workers, who may be exposed to C. burnetii at intervals, with possible immune re-stimulation and modification of clinical state or immune markers. A single strain of C. burnetii was probably involved in the outbreak, thereby avoiding strain variations in virulence. Finally, clinical presentations were not confounded by claims for compensation as the precise ovine source(s) of the infection was unknown.

The present report compares the prevalence of markers of infection in the earlier cohort of sporadic Q fever patients in Australia sampled at 9 months to 5 years after acute infection with those in the Birmingham cohort sampled 12 years out from their original infection. Markers included: (i) antibody levels to the two Q fever Phase antigens in the three immunoglobulin classes; (ii) detection of C. burnetii DNA in bone marrow and PBMC by PCR assays directed against several different targets in the genome; and (iii) attempts to isolate coxiellas in cell culture or mice from PCR positive samples.


Selection and classification of patients

Birmingham cohort

Acute primary Q fever patients had a compatible clinical illness and either a four-fold or greater increase in CF antibody titre to C. burnetii Phase 2 antigen between early and late sera, or a CF antibody titre ≥256 in a convalescent phase serum. Patients on record 12 years after the outbreak were asked to volunteer for the study that was approved by the East Birmingham Local Research and Ethics Committee. An explanatory sheet and consent form was issued. Ninety-two patients from the cohort agreed to give blood samples for serum and PBMC and 35 gave bone marrow samples. Clinical grouping followed the hierarchical classification for chronic fatigue syndrome used by Wessely et al.24 and also met the criteria of the CDC group25—see Wildman et al.26 for details and definition of categories. The number of subjects in the various clinical categories together with the number of samples of serum, bone marrow and PBMC received and incorporated into final analysis (see below) is shown in Table 1. Samples were transported to Adelaide on solid CO2.

View this table:
Table 1

Clinical classification and range of samples collected from 92 patients in the Birmingham UK Q fever outbreak, 12 years after infection

GroupClinical categorynSerum samplesBone marrow samplesPBMC samples
1Fatigue, no duration, no comorbidities13133 (3)13 (3)
2Fatigue, no duration, comorbidities present330 (0)3 (0)
3No fatigue, no comorbidities232312 (12)23 (22)
4No fatigue, comorbidities present221 (1)2 (1)
5Chronic fatigue syndrome, no comorbidities886 (5)8 (8)
6Chronic fatigue syndrome, comorbidities present554 (4)5 (5)
7Severe idiopathic fatigue, no comorbidities221 (1)2 (1)
8Severe idiopathic fatigue, comorbidities present886 (6)8 (6)
9Unclassified28282 (2)27 (2)
Total929235 (34)91 (48)
  • Numbers in parentheses indicate number in group finally tested according to cost constraints, and residual specimen volume, etc. (see text).

Australian cohort

The selection and characterization of 18 of the 29 QFS patients eventually recruited to the group are detailed elsewhere.1,,20 The intensity of their illnesses was scored on a locally devised scale, and also met the CDC criteria for CFS.25 Serum and PBMC were collected from all 29 patients, bone marrow aspirates from 20, and thin-needle liver biopsies from 14. Two patients were sampled 9 months after the acute primary Q fever and the rest after between 12 months to 5 years. The mean period from illness to sampling was 31 ± 6 months (95%CI 19–43 months).

Collection of bone marrow and PBMC specimens

Bone marrow was aspirated from the iliac crest under light sedation and placed in EDTA anticoagulant tubes. The use of heparin, inhibitory for PCR reactions, was avoided except for moistening inside of syringes for the aspiration step. Blood was collected direct into EDTA tubes and PBMC retrieved by gradient separation.

Detection of C. burnetii by PCR amplification of genomic sequences

A range of locally designed PCR assays in single round or nested configurations were used, directed against targets in the superoxide dismutase gene,27 in the insertion sequence IS1111 a,28 in the Com1 gene coding for an outer membrane protein of C. burnetii29 and in the gene for 16S ribosomal RNA.30 Formulae for the most effective primer and probe sequences are given below.

Similarly, only critical steps of the DNA extraction method and the PCR reaction conditions are given.

Extraction of DNA from PBMC and bone marrow

The extraction procedure is a crucial step in the PCR testing of the PBMC and bone marrow samples. The C. burnetii small-cell variant has a difficult-to-open, resistant, cross-linked peptidoglycan/protein sacculus encasing its DNA. In addition, inhibition by human DNA reduces sensitivity of the PCR assay.1 Essential steps are: (i) extended digestion of the sample for two or more days at 50°C with Proteinase K (1.3 mg/ml final) and SDS (2.0% w/v final); (ii) chloroform extraction to recover the aqueous phase; (iii) DNA purification by column chromatography (QIAamp DNA kit, Qiagen). A detailed protocol is required to repeat the work, and may be obtained from the corresponding author.

PCR reactions

Primers and probe for IS1111a-based PCR assays for C. burnetii (all sequences 5′–3′)

P1f 508 GCG GTG GGA TTA ACA CCG CGG ATG 531 (amplicon size 343 bp)


P3f 717 TCA TCG TTC CCG GCA GTT 734 (amplicon size 73 bp)


Taqman primers and probe

TP1f 671 TAA CGG CGC TCT CGG TTT 688 (amplicon size 61 bp)



Positions based on first base of initiation codon = +1 See also reference 1.

Primers and probe for Com1-based single round and nested PCR assays for C. burnetii (all sequences 5′–3′)

P1f 153 AGT AGA AGC ATC CCA AGC ATT G 174 (amplicon size 500 bp)


P2f 187 GAA GCG CAA CAA GAA GAA CAC 207 (amplicon size 438 bp)


Taqman primers and MGB probe

TP1f 419 AAT CGC AAT ACG CTG CCA AA 438 (amplicon size 76 bp)



Positions based on first base of initiation codon = +1.

Primers and probe for 16S rRNA gene-based PCR assay for C. burnetii (all sequences 5′–3′)

P1f 143 GGA TAA CCT GGG GAA ACT GC 162 (amplicon size 88 bp)


Taqman MGB probe


Positions based on first base from start of gene = +1.

Reaction conditions for PCR assays

In general, these followed Harris et al.1 With the COM1 primers and probe, more sensitive results were obtained by following the denaturing step with an annealing temperature at 65°C, then by temperatures of 60°C and 50°C at 1 min each, for all thermal cycles. Hot Star Taq polymerase (Qiagen) generated specific amplicons at expected molecular sizes, with few or no non-specific bands in the gel.

Avoidance of contamination of PCR reactions with exogenous amplicon from other specimens or from positive control suspensions

Routinely, the assay processes took place in five separate locations, each with its own equipment, gowns and glove supplies. They comprised: (i) clean room for preparation and dispensing reagents; (ii) laboratory in a PC3 containment facility and animal house where the samples were processed in a Class 2 biocontainment cabinet; (iii) a laboratory in which reaction mixes were dispensed and DNA added in a Class 2 biocontainment cabinet; (iv) a room for PCR amplification in the 9600 thermal cycler (Applied BioSystems) or the Rotor Gene Cycler (Corbett Industries), the ABI Prism 7700 Sequence Detector (‘Taqman‘) in a separate location also being used; and (v) a room in a separate building in which products were run in gels and bands visualized.

Each test run contained 1–3 times as many ‘non-template’ controls (NTCs) (i.e. all reagents except the target sequence) as there were test samples. NTCs were distributed among the latter and left open until just before addition of positive control suspension (∼5 QVax cells or equivalent extracted DNA). Gloves were changed between the dispensing step for each specimen. The use of cloned C. burnetii DNA as a positive control was embargoed to restrict the number of genome copies in the environment. Only the low copy number of Q Vax coxiella cells (Henzerling strain) were handled in laboratory (iii); the concentrated Q Vax suspension was processed elsewhere.

All NTCs had to be negative for a test run to be accepted as valid. In addition, at the end of the study, to exclude amplicon contamination, critical samples were re-extracted by one worker and the DNA assayed by PCR by a second worker (see ‘Selection and observer bias’).

Identification of amplicons

The criteria for accepting an amplicon as generated from C. burnetii sequences were not limited to bands of the correct size on a gel. They included:

(i) Amplification with a fluorescent probe in real time Rotor Gene or Taqman 7700 instruments. Amplification plots were not accepted as positive unless curves rose exponentially and achieved a value at or above 0.5 log over the Ct line before truncating or reaching a plateau value (e.g. Figure 1); semi-linear curves crossing the Ct line at a narrow angle and at values lower than 0.5 log above the line while possibly specific were not taken as positive.

Figure 1.

PCR amplification curves with Taqman primers and probe for C. burnetii 16S rRNA gene sequences in Com1-sequence positive bone marrow samples from Birmingham Q fever patients. Red, positive Q Vax control, standard exponential curves/positive. Green/blue, patient 37, truncated exponential curves/positive. Purple, patient 122, truncated exponential curves/positive. Yellow, patient 117, ambiguous curve/negative. Black, non-template control/negative.

(ii) Cutting the amplicon with two restriction enzymes (often poorly reproducible).

(iii) Sequencing amplicons by the fluorescent dideoxy-dye-terminator method, which proved optimal. Matches >99% were obtained with GenBank data. As required bands at correct size on a gel were cut out and eluted to provide amplicons for sequencing, these eluates were not re-amplified.

Measurement of antibody to Phase 1 and 2 antigens of C. burnetii by complement fixation (CF) and immunofluorescence (IFA) assays

Antibodies to C. burnetii were measured by the methods of Worswick and Marmion.31 Coxiella Phase 1 and 2 CF antigens came from Institut Serion-Virion GmbH (Wurzburg, Germany). Phase 1 and 2 antigens for microdots in the IFA assay were a generous gift from MG Peacock, Rocky Mountain Laboratory (NIAID) (Hamilton, USA). Sera were pre-tested for rheumatoid factor.

Isolation of C. burnetii in cell culture and laboratory animals

Cell culture

Supernatant fluids from low-speed centrifugation of emulsions of bone marrow were centrifuged into sheets of human embryo lung fibroblasts on cover slips in shell vials.32 Specimens from Q fever placentitis and endocarditis were used as control material. Cultures were incubated for 14 days and the cover slips stained by IF with rabbit antiserum to Phase 1 antigen. Two more serial passes were made with cells and ‘conditioned’ medium to fresh cultures. The cells and medium from the final pass were tested by IFA, Com1 PCR, and sub-inoculation into mice.

Mouse inoculation

Groups of 2–3 A/J and IFN-γ knockout mice were inoculated IP with 0.5 ml of emulsion from bone marrow or PBMC specimens and held for 14 days. Spleen and liver were harvested and passed 2–3 more times in one or other of the two strains of mice. Spleen fragments were placed at once into lysis buffer (Proteinase K 2 mg/ml, SDS 0.5% w/v) and tested by PCR for C. burnetii. The protocol passed the Institutional Animal Ethics Committee.

Selection and observer bias

Ninety two patients from the Birmingham cohort gave blood samples yielding 92 sera and 91 PBMC; 35 gave bone marrow aspirates. The 92 sera were tested in the routine serology laboratory by a worker unaware of their clinical classification. Aside from the ‘unclassified’ Group 9, the Wessely classification for CFS yielded eight subcategories, six of which had fatigue; four with a confounding combination with a co-morbidity (Table 1). Also, the proportion donating the more invasive bone marrow aspirate varied substantially between groups. To ensure reasonable coverage of the PBMC and bone marrow samples, and in view of the expensive and protracted period required to develop the PCR techniques, the two most clearly defined and contrasting clinical groups were chosen for the most intensive comparative analysis, although all bone marrow samples were tested. These were Group 3 (‘acute Q fever not followed by fatigue, other sequelae or co-morbidities’) and Group 5 (‘acute Q fever followed by chronic fatigue syndrome without co-morbidities’). Table 1 gives the number of specimens remaining in each group and used for final analysis after extensive testing, principally to exclude the presence of IS1111a targets, and testing by mouse inoculation. Three workers operated independently over different stages of the extraction and PCR assays, with some cross testing of samples. The ‘readouts’ of the assays (probe positive and/or correct sequence) were machine-based, defined in advance and not open to subjective evaluation.


Serological observations

Sera from the Birmingham cohort of Q fever cases were tested both for complement fixing (CF) and IFA antibody to C. burnetii Phase 1 and 2 antigens. CF tests with Phase 1 antigen were uniformly negative in all clinical subgroups. With phase 2 antigen, the geometric mean (GM) CF antibody titres were low and ranged from 11 to 18 (data not included in Table 2a). Overall, reactions in the more sensitive IFA test were more frequent and predominantly limited to the IgG class; the highest GM titres being against Phase 2 antigen (Table 2a). As with the CFT, the GM titres for IFA antibody were not significantly different in each of the different clinical subgroups, and fell within the range 176–190. In particular, the GM titres of IgG antibody to Phase 2 antigen in the critical Groups 3 and 5 were the same: 186. Comparison of the distributions of individual values within the two groups showed no significant difference (p = 0.9, Mann-Whitney U).

View this table:
Table 2a

Antibody titres to Coxiella burnetii Phase 1 and 2 antigens in various clinical groups in Q fever cases in the Birmingham UK outbreak (time from onset 12 years), assayed by immunofluorescence and in IgG, IgM and IgA immunoglobulin classes

Clinical categoryAntibody to C. burnetii Phase 1 & 2 antigens
3 (n = 23)10 (<10–160)186 (10–640)<10<10<10<10
Q fever, no sequelaea
5 (n = 8)10 (10–80)186 (40–1280)<10<10±10c±10c
1,2,4,6,7,8 (n = 33)10 (<10–160)176 (<10–1280)<10<10<10<10
Fatigue, comorbidityd
9 (n = 28)10 (<10–160)190 (<10–1280)<10<10<10<10
  • Data are geometric means (GM) and ranges where appropriate. aPatients who had Q fever but no sequelae. bPatients who had Q fever followed by post infection fatigue syndrome (QFS) but no comorbidities. cOne patient had IgA antibody titres of 640 to Phase1 and 2 antigens, 3 had titres of 80. dPatients had fatigue, but had other morbidities that might have explained the fatigue.

Sera in all clinical Groups were negative for IgA antibody to the Q fever antigens except for one patient (patient 2) in QFS group 5 who had IgA titres of 320 and 640 to Phase 1 and 2 antigens respectively. The patient had medical imaging evidence of an indolent aortic valve vegetation, was positive in bone marrow and PBMC for C. burnetii genomic DNA, and was considered to have a low-level, cryptic Q fever endocarditis. Only one patient in the total of 92 tested was completely seronegative. Apart from patient 2, the serological profiles clearly differed from those conventionally considered characteristic of Q fever endocarditis. The majority of IgG Phase 1 antibody titres by IFA fell in the range <10 to 80 and the IgG Phase 2 antibody titres from 10 to 320. Note that sensitivities will vary from one laboratory to another.

The serological findings in individual patients in the Australian cohort are detailed by Penttila et al.20 and the geometric mean titres in Table 2b. Only 16 patients were analysed, as two in the subgroup of 18 did not have an identified initial clinical attack of Q fever, but had low levels of antibody and a positive PCR assay for coxiella DNA in PBMC or bone marrow.

View this table:
Table 2b

Antibody titres to Coxiella burnetii Phase 1 and 2 antigens in Australian Q fever cases (time from onset 31±6 months, 95%CI 19–43 months) and in a control group of non-exposed Australian industrial workers, assayed by immunofluorescence and in IgG, IgM and IgA immunoglobulin classes

Clinical categoryAntibody to C. burnetii Phase 1 & 2 antigens
CasesQFS (n = 16)72 (10–320)258 (10–5120)10 (<10–80)10 (<10–640)30a (<10–640)30a (<10–640)
Scores >100
ControlsbNo CFS/QFS (n = 37)<10<10<10<10<10<10
Scores <20
  • Data are geometric means (GM) and ranges where appropriate. aOne patient had IgA antibody titres of 640 to C. burnetii Phase 1 and 2 antigens. bgroup of non-exposed Telstra workers surveyed for antibody and symptom scores for CFS.

The serological profiles in this cohort of QFS patients although sampled closer to an initial illness were broadly similar to those of the Birmingham patients. As with the latter, the dominant response by IFA was in the IgG class and greatest to the Phase 2 antigen.20 However, there were five patients with medium levels of IgM antibody, and four with low levels of IgA antibody, presumably reflecting a more recent primary infection (Table 2b). Again, the serological profiles differed from those of Q fever endocarditis. Sera from a control group of 37 volunteers without QFS/CFS symptoms and working in the industrial section of a telecom (Telstra Australia) were negative in the IFA test for Q fever antibody (Table 2b).

Other serological observations

Previously, we detected raised serum levels of IL-6 in a proportion of QFS patients.20 Accordingly, present sera were tested for a range of shed cytokine receptors and other markers as possibly less transient indicators of increased cellular immune activity. There were no significant differences between serum levels of IL-6 receptor, sTNF R1 and R11, TGFβ R1 and R11, E selectin and ICAM1 in QFS patients, when compared to the Telstra group of volunteers.

Detection of C. burnetii genomic DNA sequences in the two cohorts of Q fever patients

With the Birmingham UK patients (Table 3a) various one-round and nested versions of the IS1111a PCR were consistently negative with the bone marrow and PBMC samples, even though these primer sets and probe reacted with the positive Q Vax (Henzerling) controls, and also with specimens from the Australian QFS and 5 local Q fever endocarditis patients (Table 3b). In marked contrast, the single-round or nested Com1 PCR assays detected positives in both types of specimens, and in all clinical subgroups of the Birmingham cohort. Response curves with the latter in probe-based fluorometric assays were acceptable (Figure 1), and all non-template controls from the test runs were negative. In all, 22 Com1 amplicons from the samples in the various clinical groups in the Birmingham cohort were sequenced and shown to be C. burnetii (Table 3a).

View this table:
Table 3a

Detection of Coxiella burnetii genomic DNA sequences in bone marrow and PBMC samples from Q fever cases in the Birmingham UK outbreak, 12 years after initial infection: PCR assays directed against targets in the Insertion sequence (IS1111a), COM1 and 16S rRNA genes of C. burnetii

PCR assays and gene targets (positives/total tested)
Clinical categorySampleIS1111aCom116S rRNACorrect sequence Com1Correct sequence 16S rRNA
3 (n = 22) Q fever,Bone marrow0/1111/126/97/73/3
    no sequelaeaPBMC0/223/22NT3/3NT
5 (n = 8) QFSbBone marrow0/54/53/34/42/2
1,2,4,6,7,8c (n = 33)Bone marrow0/1513/155/123/33/3
    Fatigue, comorbiditydPBMC0/160/16NTNTNT
Positive control Q Vax5 cells3/33/33/3
  • aPatients who had Q fever but had no chronic sequelae or comorbidity. bPatients who had Q fever followed by post infection fatigue syndrome (QFS/CFS) but no co morbidity. cUnclassified group not shown in table. dPatients had fatigue but had other morbidity that might have explained the fatigue. NT, not tested.

View this table:
Table 3b

Detection of Coxiella burnetii genomic DNA sequences in bone marrow, PBMC and liver samples from sporadic Q fever patients with post infection fatigue syndrome (QFS) in Australia (31±6 months after onset, 95%CI 19–43 months) and valve extracts from five local Q fever endocarditis cases (additional control): PCR assays directed against targets in the insertion sequence (IS1111a) and COM1 genes of C. burnetii

PCR ranges and gene targets (positives/total tested)
Clinical categorySampleIS1111aCom1Correct sequence IS1111a
QFS (n = 29)Bone marrow13/202/84/4
Q fever endocarditis (n = 5)Valve extract5/55/5NT
Positive control Q vax5 cells3/33/3
  • NT, not tested.

The difference in reaction patterns with IS1111a and Com1 primers indicated that the positive results with the Birmingham samples were not due to contamination with coxiella cells or DNA originating from the positive controls or Australian samples. But it could be argued that amplicons generated with the Com1 primer sets from the Q Vax controls had (in some undefined way) contaminated all the subsequent assays with the Birmingham samples, despite the fact that the non-template controls had been negative.

Accordingly, DNA was re-extracted by worker 1 from the collection of 31 bone marrow samples that remained, starting from the original tubes. A non-template/reagent control tube accompanied each sample through a parallel extraction process up to the final column treatment. Worker 2 then tested the freshly derived DNA extracts and the linked non-template controls with the Com1 primer and probe sets. Only the DNA extracts were positive. Second, 24 of the 31 freshly-extracted bone marrow DNA samples were also tested with primers and probe directed against a target in the 16S rRNA gene; reagents that had not been used previously in the laboratory for sample testing. Of the 24 samples, 14 (58%) were clearly positive, matching the positive fluorometric responses as defined in Figure 1. Eight amplicons from the 14 were sequenced and were correct for C. burnetii 16S rRNA gene.

The findings with targets in two genes widely separated in the coxella genome33,,34 confirmed that longer sequences of DNA than that of a hypothetical contaminant amplicon from the Com1 primer set were present in the bone marrow samples.

Overall, carriage of the coxiella (as judged by detection of its genomic DNA by Com1 PCR) was present at a similar proportion (p = 0.4) in bone marrow of all classified clinical groups (Table 3a). In all, 28 (88%) of the 32 bone marrows finally remaining for test were positive, a proportion higher (p = 0.03) than that observed previously (65%) in the Australian patients with the multiple copy IS1111a gene1 (Table 3b). The unclassified Group 9 contributed two bone marrows, one of which was positive. On the other hand, the proportion of positives in the PBMC samples overall was much lower (5/46, 11%) and similar to that (17%) in the Australian group.

However, in contrast to the almost uniform distribution of positive reactors in the bone marrow samples, there were significant differences in the proportion of reactors in PBMC samples from the different clinical Groups in the Birmingham cohort. Thus the proportion of positives in PBMC from Group 5 (‘QFS/CFS without comorbidity‘) (5/8) was significantly higher than that (3/22) in Group 3 (‘acute Q fever, no sequelae‘) (p = 0.015, Fisher's exact test). On the other hand, the proportion of positives in the bone marrow samples of the two groups did not differ significantly (p = 1.0, Fisher's exact test) (Table 3a).

In a further analysis to detect different levels of coxiella genomes in the two groups, the distribution of Ct values in the Taqman assays of bone marrow samples from group 5 and group 3 patients was compared. The mean value in group 5 was 32.5, whereas that in Group 3 was 37.2. While the difference indicated between 32 to 64 as many ‘targets’ in group 5 (lower Ct values indicate higher numbers present), the trend did not reach statistical significance (p = 0.22, Mann Whitney U statistic). Comparisons were also made of the number of Com1 positives obtained in the sets of triplicates in nested PCR tests on each bone marrow sample. (Variations in frequency of positives in triplicates arise from the small number of coxiellas in the samples and sampling variability around an endpoint). In nested round 1 of Com1 PCR, group 3 had 22 positive/44 replicates vs. 6 positive/6 in group 5 (p = 0.028). And in nested round 2, group 3 had 31 positive/60 vs. 20 positives/22 in group 5 (p = 0.0002). This reinforced the conclusion that there were more targets in Group 5 specimens.

With the Australian QFS bone marrow and PBMC samples, and those from endocarditis patients, C. burnetii DNA was detected1 in the probe-based versions of the IS1111a and by the single-round and nested Com1 PCR assays (Table 3b). Four of the earlier1 IS1111a positives from bone marrows had been confirmed by sequencing, and gave >98% agreement with C. burnetii sequences on GenBank. The positive control of ∼5 cells of C. burnetii Henzerling strain (Q Vax) reacted consistently with all three assay systems used with the Australian and Birmingham cohort of cases. NTC in all systems were negative. During the study, 21 bone marrow samples from leukaemia and other haematological conditions unrelated to Q fever also tested negative with the IS1111a PCR.

Tests for viable C. burnetii in PCR-positive samples

Cell culture

Bone marrow suspensions from 14 Australian patients were given three serial passes in human embryo lung fibroblasts using shell vial culture.32 IFA staining of cell sheets did not reveal convincing intracellular microcolonies. ‘Conditioned’ cell culture media from the third pass was tested by Com1 PCR and were negative. Control suspensions from 5/9 known positive Q fever endocarditis valve vegetations and uterine curettings from Q fever placentitis gave intracellular microcolonies and were positive by IFA. All nine were positive by COM1 PCR, presumably due to the additional presence of dormant or dead but undegraded coxiella cells.

Mouse inoculation

Bone marrow and/or PBMC suspensions from 12 patients in the Birmingham cohort positive by Com1 PCR were passed serially 2 to 3 times in A/J or INF-γ ‘knockout’ mice. Mice in each passage were held for 14 days, and spleen fragments harvested for passage to fresh animals and PCR tested. Results were negative.


At the start of the investigation, we anticipated that persistence of C. burnetii would be limited to patients with QFS, Q fever endocarditis or a granulomatous recrudescence. In the event, our results indicated a more complex interaction between a host-regulated, persistent carriage of C. burnetii and disease. Twelve years after infection, all clinical subgroups of volunteers from the Birmingham cohort of Q fever cases had essentially the same geometric mean titres of Q fever antibody. Only one patient was serologically negative. A high proportion (∼88%) of individuals tested had C. burnetii DNA in the bone marrow, irrespective of clinical state. It seems that, in essence, all subjects had persistent coxiellas; conceivably either live, dormant, or dead but with un-degraded DNA. On the other hand, only a small minority (group 5, ∼10%) had QFS symptoms with no co-morbidity—a clinical state shown previously to be associated with a hyper-reactive cellular immune state to Q fever antigens, as indicated by cytokine dysregulation.20 Another subject had probable Q fever endocarditis, but with general symptoms similar to QFS, as is often the case. Other patients had fatigue but with a co-morbidity; at present, final conclusions can not be drawn about the relation, if any, between carriage of the coxiella and their illnesses. Given the serological and PCR evidence that most group 3 patients remained asymptomatic, but continued to carry the organism, it follows that there must be an additional but variable factor of host regulation of cellular immune responses which determines the level of persistence and symptomatic outcomes—as for example, in group 5 with QFS (see below).

PCR detection of an organism in terms of its genomic DNA sequences unsupported by positive culture at once invites scepticism either on the grounds of ‘specificity’ (e.g. the PCR products are not authentic but are, for example, low level hybridization artifacts coded by eukaryotic or extraneous prokaryotic DNA in the specimen) or, if the sequences are shown to be correct, on the grounds of ‘contamination’, e.g. from suspensions of the control prototype strain of C. burnetii or its derived amplicons.

As regards ‘specificity’, 22 (61%) of the total of 36 Com1 amplicons from bone marrow and PBMC samples from clinical groups (other than group 9, unclassified) in the Birmingham cohort were submitted for sequencing, and shown to be over 99% identical with authentic C. burnetii sequences in GenBank (Table 3a). All positive amplicons had also been verified by probe identification in the Taqman or Rotor Gene instruments. Additionally, 14 (58%) of a residual group of 24 bone marrow samples from Birmingham patients were shown to contain a C. burnetii 16S rRNA gene sequence (Table 3a).

As regards ‘contamination’, routine NTC and the experiment with intercalated NT/reagent controls during a reprocessing of bone marrow aspirates did not uncover amplicon or coxiella cell contamination (see Results). A definitive finding was that, unlike the strains of C. burnetii involved in the Australian QFS and endocarditis patients and the Henzerling strain in the Q vax control, the Birmingham outbreak strain had no complementary sequences to the primer pairs for our chosen targets in the 1S1111a gene. This difference excludes possible contamination from the Australian samples or Q Vax cells. The recent sequencing of the C. burnetii genome33 confirms that the Nine Mile strain (at least) has multiple (∼20) copies of the IS1111a gene. The latter have the primer sites we have chosen as targets. Our PCR assays consistently detect ∼ 5 coxiella cells with the single copy Com1 and 16S rRNA gene targets. As the Birmingham outbreak strain gave no specific product with any of three primer pairs directed against different sequences in the IS1111a gene, we conclude that the insertion elements are most likely not present. The Birmingham strain genome needs to be sequenced or otherwise characterized to establish this with certainty. An isolate is not available at present, but efforts to isolate the outbreak strain continue as it may be either a new species of coxiella or, alternatively, merely a locally prevalent variant without insertion sequences. In the latter context, attention is drawn to analogous variations in the copy number of the insertion sequence IS6110 in Mycobacterium tuberculosis isolates, and its absence in a minority of geographically-restricted strains.35

Overall, it appears that authentic C. burnetii genomic DNA is present in the specimens. The nature of the cell containing it has not been visualized directly because of the small numbers present. Indirect indications are: (i) that the DNA is enclosed in a resistant sacculus requiring prolonged Proteinase K and SDS to liberate the DNA; (ii) that taking the Birmingham and Australian data together, the SOD, IS1111a, Com1 and the 16S rRNA genes can be present (as these genes are situated at different locations in the genome,33,,34 this indicates that a major part or all of the genome is present); and (iii) that a majority of subjects in the Birmingham and Australian cohorts have persisting antibody to Phase 1 (LPS) and Phase 2 (protein) epitopes measured by IFA techniques directed against surface antigens of coxiella cells.

McCaul21 described three cell forms of the coxiella in EM preparations: (i) a small-cell variant (SCV) with a condensed nucleus, a rigid peptidoglycan–protein sacculus and an external membrane with Phase 1 LPS and Phase 2 antigenic epitopes; (ii) a large-cell variant (LCV) (perhaps a replicating form of the coxiella) with a more open peptidoglycan and diffuse nucleus, evidently without Phase 1 LPS epitopes, but with a unique 29.5 kDa membrane protein not present in resting SCV; and (iii) within some LCV, there is a morphological’ spore form’ containing DNA. A sporulation gene is claimed,36 but it is unknown whether the ‘spore’ has the complete genome, or if it is infective. Also, the membrane antigens of the SCV or LCV do not appear to be present. Comparison of our findings with the attributes of the three C. burnetii cell forms suggests the persistent cells are likely to be SCV rather than ‘spores’.

The C. burnetii DNA detected by PCR might conceivably be from those coxiellas seeded throughout the body during the original primary infection, but dead or ‘dormant/replication defective’ at the time of bone marrow sampling. Alternatively, they may have originated recently from a tightly host-regulated focus of low-level coxiella multiplication in bone-marrow cells, but have been killed and liberated during the maturation and immune activation of their host cells.

The capacity of dead or dormant coxiella cells and their antigenic components for long term survival in the host should not be dismissed lightly. It is illustrated, for example, by the persistence of C. burnetii in endocarditis valve vegetations even after long-term chemotherapy. The coxiella cells—visualized by Giemsa or Macchiavello staining, immunofluorescence or electron microscopy—may be found inside macrophages5,,37 and more frequently extracellularly as small microcolonies or aggregates.4,38–43 Presumably the latter were once in host macrophages subsequently destroyed by immune reactions and apoptosis.44 The extracellular microcolonies contain antigen demonstrable by specific IFA staining of surface epitopes with antibody to Phase 1 antigen, and may also be coated with host antibody.39 Material containing them is PCR-positive. A limited number of studies in which vegetations have been stained for organisms, titrated for infectivity in animals, and for coxiella genomes by PCR reveals discrepancies between visual amounts of cell aggregates and infectivity. This suggests that the majority of the coxiella cells may be dead or dormant, but the matter requires more investigation.4,,10

The long lasting immunity after inoculation with a killed Q Fever vaccine45 and the finding of C. burnetii DNA in occasional granulomatous reactions at vaccination sites months after inoculation (C. Taylor, J. Robson and S. Graves, personal communication and Q Fever Research Group10) are other pointers to the survival of dead but antigenic coxiella cells in the host. Nevertheless the survival of dead organisms alone cannot explain recrudescence of live coxiellas some years after initial infection, e.g. at the end of pregnancy, or recrudescence in the endocardium, on prostheses, in the testis or other organs, or that following iatrogenic immunosuppression. Note that living C. burnetii has been isolated from bone marrow in chronic Q fever granulomatous hepatitis.13

A provisional, hypothetical model to reconcile these observations is that in the great majority of acute Q fever patients (i.e. those without sequelae) there is a continuing low level but strictly regulated multiplication of live coxiella (perhaps, for example, in early lineage bone marrow cells such as monoblasts or premonocytes), which may have a lower capacity for activation and antigen presentation (C. burnetii multiplies in cells in vitro without direct cytopathic effects). The coxiella might also subvert IL-2 and IFN-γ production45 and so prolong the period of replication. Eventually, however, with maturation of the infected host cell to monocyte/macrophage, there is activation, and antigen presentation. The parasitized cells are then destroyed by CMI and apoptosis,44 with liberation of dead but unopened coxiella cells still detectable by PCR.

It is further proposed that the levels of the immune regulation of multiplication, generation of dead cells and chronic outcomes are determined by allelic polymorphisms in the immune response and control gene repertoire of the subject. Thus compared with controls, QFS patients show an increased proportion of HLA DR B1*11 and allelic differences in a control sequence for an IFN-γ gene,46 whereas Q fever endocarditis patients show allelic differences in IL-10 and TNF receptor genes.47 Irrespective of genetic background, immune regulation might also be subverted by changes in cellular immune function, e.g. in late pregnancy, or by intense iatrogenic immunosuppression, as when patients develop Q fever in the late stage of bone marrow transplantation.48 The robust nature of the immune regulation is indicated by a surprising absence of reports of recrudescent or chronic Q fever when HIV infection is superimposed on persistent carriage of C. burnetii.49,,50

Loss of immune control would permit the survival and emergence of monocytes/macrophages containing live coxiellas beyond the boundaries of the bone marrow to seed, e.g. endocardium on damaged or abnormal valves.

There are similarities between host-pathogen interactions in Q fever and those of mycobacterial infections. Recent reports51,,52 on Mycobacterium avium subspecies paratuberculosis (MAP) in patients with Crohn's disease describe the contrast between the ease of PCR detection of MAP DNA and the exacting, prolonged process of culturing the cell-wall defective organism in cell-free, artificial media.

Clearly the validity of this provisional model turns on the sensitivity of the methods used to detect living coxiellas in bone marrow or PBMC samples that were positive for C. burnetii DNA by PCR. Reports are ambiguous. Ormsbee et al.53 found close agreement between the direct fluorescence microscopic count of cells (DFC) in an optimally prepared suspension of the virulent Nine Mile strain (Phase 1) of C. burnetii and the numbers making up one ID50 (0.5–0.6; essentially one organism) on titration in out-bred Swiss mice or chick embryo yolk sac (CE). In contrast, the sensitivity of various forms of cell culture was lower than that of mice or CE.

Scott et al.54 inoculated (i.p.) 47 different inbred lines of mice with a dose of 106.5 of Nine Mile (Phase 1) strain. A/J mice—as used in our assays—showed the highest mortality and concentration of coxiellas in tissues, and were estimated to need only 1 to 5 coxiellas for infection. All genetic lines of mice seroconverted. On the other hand, Japanese workers55 found that A/J mice were infected (seroconversion; spleen PCR positive) by 10 TCID50 of the Nine Mile strain, but survived with minor pathological changes, whereas 66% of SCID mice were killed by 10−4 TCID50 of the coxiella suspension, with gross pathological changes and high tissue concentrations of the organism.

Similarly, while the sensitivity of the shell vial technique32 with HEL cells is satisfactory for isolation of C. burnetii from valve vegetations or bone marrow from Q fever endocarditis (perhaps reflecting large numbers of viable coxiellas), its absolute sensitivity does not appear to have been assessed against a total cell count (e.g. DFC as used by Ormsbee et al.53) or against a sensitive PCR system such as IS1111a calibrated with a coxiella suspension of known cell numbers. Thompson and colleagues (personal communication, 2004) using a shell vial technique but with Vero or Buffalo green monkey cells, found that inocula of 103 ID50 from the Nine Mile Phase 1 strain were required to initiate infection, even though these cell substrates subsequently supported satisfactory growth. In a previous report,1 we estimated that there might be 40–700 coxiella genomes/ml of QFS bone marrow aspirates. So the standard, conventional culture systems we used may not have been sensitive enough. Moreover not all strains of C. burnetii may be as virulent as Nine Mile. Validation of the provisional model for persistence should include not only the use of SCID mice, but also more complex, prolonged primary and co-cultivation cell culture techniques, coupled with measures to prevent activation of the (postulated) infected monocyte precursors in the bone marrow sample.


We are much indebted to Dr Martin Wildman, Mrs Jayne Groves, and to Dr Chris Fegan and his staff in the Department of Haematology, for the collection, processing, and shipment of samples from the Birmingham Q fever cases. Dr Adrian Easterman gave much appreciated statistical advice. Generous grant support was given by Meat and Livestock Australia Pty Ltd and the TRUST Co. Australia Ltd (via Woodend Foundation; Mrs Isabel Milner, Sydney, NSW) to the Q Fever Research Group over many years.


View Abstract