Highly sensitive real-time PCR for specific detection and quantification of Coxiella burnetii.

BACKGROUND: Coxiella burnetii, the bacterium causing Q fever, is an obligate intracellular biosafety level 3 agent. Detection and quantification of these bacteria with conventional methods is time consuming and dangerous. During the last years, several PCR based diagnostic assays were developed to detect C. burnetii DNA in cell cultures and clinical samples. We developed and evaluated TaqMan-based real-time PCR assays that targeted the singular icd (isocitrate dehydrogenase) gene and the transposase of the IS1111a element present in multiple copies in the C. burnetii genome.
RESULTS: To evaluate the precision of the icd and IS1111 real-time PCR assays, we performed different PCR runs with independent DNA dilutions of the C. burnetii Nine Mile RSA493 strain. The results showed very low variability, indicating efficient reproducibility of both assays. Using probit analysis, we determined that the minimal number of genome equivalents per reaction that could be detected with a 95% probability was 10 for the icd marker and 6.5 for the IS marker. Plasmid standards with cloned icd and IS1111 fragments were used to establish standard curves which were linear over a range from 10 to 10(7) starting plasmid copy numbers. We were able to quantify cell numbers of a diluted, heat-inactivated Coxiella isolate with a detection limit of 17 C. burnetii particles per reaction. Real-time PCR targeting both markers was performed with DNA of 75 different C. burnetii isolates originating from all over the world. Using this approach, the number of IS1111 elements in the genome of the Nine Mile strain was determined to be 23, close to 20, the number revealed by genome sequencing. In other isolates, the number of IS1111 elements varied widely (between seven and 110) and seemed to be very high in some isolates.
CONCLUSION: We validated TaqMan-based real-time PCR assays targeting the icd and IS1111 markers of C. burnetii. The assays were shown to be specific, highly sensitive and efficiently reproducible. Cell numbers in dilutions of a C. burnetii isolate were reliably quantified. PCR quantification suggested a high variability of the number of IS1111 elements in different C. burnetii isolates, which may be useful for further phylogenetic studies.

Background

Coxiella burnetii is the causative agent of Q fever, a zoonosis that occurs worldwide and infects a variety of different animals, including domestic mammals like cattle and sheep. Whereas animals in general show no clinical signs of infection except occasional abortions, C. burnetii can cause serious illness in humans, where infections usually occur via aerosols. Acute disease often presents as a self-limiting influenza-like illness with fever and headaches, but severe cases with atypical pneumonia or hepatitis may occur. The disease can become chronic with life-threatening endocarditis as the most frequent clinical manifestation that requires long lasting antibiotic treatment [1]. Although an obligate intracellular organsim, the bacterium is very resistant to environmental conditions due to extracellular spore-like forms, and even a single organism can produce disease. Because of its widespread availability, environmental stability and low infective dose, C. burnetii is considered a potential bioterrorist agent and is classified as a group B agent by the Centers for Disease Control and Prevention in Atlanta, USA [2].

C. burnetii is a slow growing bacterium that can be cultivated in embryonated eggs or eukaryotic cell culture, which is time consuming and must be performed in biosafety level 3 laboratories. Antigen detection of bacteria by capture ELISA or direct immunofluorescence is difficult and has relatively high detection limits. Therefore, diagnosis is still mainly based on serological methods like indirect immunofluorescence, complement fixation or ELISA, with the disadvantage of delayed diagnosis because specific antibodies appear only one to two weeks after infection [3].

During the last years, several PCR based diagnostic assays were developed to detect C. burnetii DNA in cell cultures and clinical samples. These assays used conventional PCR [4-8], nested PCR [9-12] or real-time PCR conditions with LightCycler [13-15], SYBR Green [16] or TaqMan chemistry [17]. The target sequences of the assays originated from singular chromosomal genes like com1 or htpB, on plasmids (QpH1, QpRS) or on the transposase gene of insertion element IS1111 [18] that is present in 20 copies in the genome of the C. burnetii Nine Mile RSA493 strain [19]. Due to the multicopy number of the IS1111 element, the corresponding PCR is very sensitive. However, quantification of cells cannot be performed based on PCR of the IS element, because the numbers of IS1111 elements present in different Coxiella isolates are not known.

The prerequisite for a diagnostic PCR is a target sequence that is specific for C. burnetii to exclude false positive results with other organisms and that is conserved and present in all C. burnetii isolates to prevent false negative reactions. The PCR assays mentioned before were usually evaluated with relatively small numbers of characterized isolates or with uncharacterized clinical samples, though it should be noted that most importance was attached on sensitivity of the assay, whereas the suitability of the assays for a great panel of different isolates was less relevant.

The icd gene for the isocitrate dehydrogenase was sequenced in 19 strains and shown to be conserved [20]. We used a fragment of this gene as target for real-time TaqMan PCR based on TaqMan chemistry. In addition, we performed a real-time PCR assay based on a fragment of the transposase gene of the IS element IS1111a. Both assays were validated for specificity and sensitivity, and suitability of the icd assay for the quantification of Coxiella cell numbers was shown. As the exact number of IS elements is only known for the sequenced genome of the Nine Mile strain, we examined the number of IS1111 elements per genome, or per icd copy, respectively, in a large panel of Coxiella isolates of worldwide origin.

Results and discussion

Evaluation of the specificity of the real-time PCR assays

To determine whether false positive reactions occurred in real-time PCR assays with the icd and IS1111 markers, PCR was performed with DNA of the bacterial species listed in the Methods section. Based on the sequence of its 16S rRNA, C. burnetii is classified into the order Legionellales, with Legionella spp. and Francisella spp. as nearest phylogenetic neighbours [1]. Both for these related species and for all other species tested, the PCR was negative, confirming the specificity of both targets.

Determination of precision and detection limit of the assays

Based on the measured DNA concentration (29 ng/μl) and the length of the published sequence of the C. burnetii Nine Mile genome (1,995,275 bp), the theoretical number of genome equivalents (GE) was calculated to be 1.3 × 107 GE per μl. This corresponds to 2.6 × 108 IS1111 elements per μl for the Nine Mile strain (20 per genome). To determine the precision of the icd and IS real-time PCR assays, C(threshold cycle) values for eight replicates of tenfold dilutions of purified C. burnetii Nine Mile genomic DNA were measured (Table 2). The results represent independent dilution series and different PCR runs. The mean Cvalues, standard deviation, and percent CV (coefficient of variation) were calculated for each dilution. The results showed low variability, with CVs ranging from 1.3 to 1.9 % for the icd target and 1.1 to 1.6 % for the IS target, indicating efficient reproducibility of both assays. Standard curves drawn from the copy numbers and mean Cvalues shown in Table 2 had slopes of -3.687 for the icd curve and -3.527 for the IS curve (data not shown), indicating PCR efficiencies of approximately 90 % for both targets (E = 10-1/- 1, where E is the run efficiency and s is the slope of the standard curve).

Table 2

Summary of eight different PCR runs performed on eight separate DNA dilution series of the C. burnetii Nine Mile RSA493 strain

icd marker				IS1111 marker
No. of copies/μl	Mean Ct	SDa	CVb (%)	No. of copies/μl	Mean Ct	SDa	CVb (%)
1.3 × 107	15.81	0.23	1.5	2.6 × 108	15.70	0.25	1.6
1.3 × 106	18.64	0.25	1.3	2.6 × 107	17.63	0.29	1.6
1.3 × 105	22.08	0.33	1.5	2.6 × 106	21.37	0.30	1.4
1.3 × 104	25.93	0.46	1.8	2.6 × 105	25.15	0.27	1.1
1.3 × 103	29.63	0.58	1.9	2.6 × 104	28.76	0.36	1.2
1.3 × 102	33.51	0.60	1.8	2.6 × 103	32.52	0.48	1.5
1.3 × 101	37.79	0.49	1.3	2.6 × 102	36.39	0.51	1.4

aSD, standard deviation of eight replicates

bCV, coefficient of variation

Determination of the detection limit by probit analysis was performed with DNA of the C. burnetii Nine Mile strain. For the singular icd marker, detection of 100 to 0.75 GE/reaction was tested by PCR. For the IS1111 marker, where 20 copies are expected per genome, lower concentrations from 25 to 0.2 GE/reaction, or 500 to 4 copies of the IS1111 element, respectively, were tested. Each PCR was repeated three times with eight replicates for each concentration. The minimal number of genome equivalents per reaction that could be detected with a 95 % probability by real-time PCR was 10 when the icd marker was used (Fig. 1). With the IS1111 marker, 6.5 genome equivalents per reaction were detected with 95 % probability (Fig. 1), corresponding to 130 copies of the target gene. Detection of lower IS1111 copy numbers was possible, as mentioned below for plasmid standards, but less reproducible. PCR products of the icd and IS1111 assays were analysed on agarose gels and showed the expected single bands of 76 bp and 295 bp, respectively.

Figure 1

Determination of detection limits for the . The graphs show curves determined by probit analysis for real-time PCR assays targeting the icd and IS1111a sequences of C. burnetii. With the respective targets, 10 and 6.5 genome equivalents per reaction can be detected with a probability of 95 %.

Quantification using plasmid standard curves

Tenfold serial dilutions of plasmids with cloned icd and IS1111 fragments were used to establish standard curves for each PCR run. For both markers, the quantification was linear over a range of 10 to 107 starting plasmid copy numbers, and the detection limit was ten copies per reaction (data not shown).

To assess whether the number of icd and IS1111 copies per genome could be sufficiently calculated by using standard curves derived from plasmid standards, PCR assays for both targets were performed with tenfold serial dilutions of C. burnetii Nine Mile DNA and plasmid standards. The results are shown in Table 3. Especially for lower DNA concentrations, the theoretical numbers of icd and IS1111 copies (calculated from genome size and DNA concentration as shown before) corresponded quite well to the respective copy numbers determined experimentally.

Table 3

PCR quantification of DNA dilutions of the C. burnetii Nine Mile RSA493 strain. The measurements were performed in duplicate; mean values are shown.

	Calculated Valuesa		Measured Valuesb
DNA conc [pg/μl]	icd	IS1111	icd	IS1111	IS per genomec
2900	1.3 × 106	2.6 × 107	4.5 × 106	5.9 × 107	13.1
290	1.3 × 105	2.6 × 106	5.2 × 105	8.1 × 106	15.6
29	1.3 × 104	2.6 × 105	3.9 × 104	5.0 × 105	12.8
2.9	1.3 × 103	2.6 × 104	2.8 × 103	3.5 × 104	12.5
0.29	1.3 × 102	2.6 × 103	2.3 × 102	2.6 × 103	11.3
2.9 × 10-2	1.3 × 101	2.6 × 102	1.5 × 101	2.6 × 102	17.3

aNumber of target copies based on DNA concentration and genome length.

bNumber of target copies based on PCR quantification using plasmid standards.

cCalculated as IS1111 measured per icd measured.

Determination of Coxiella cell numbers by real-time PCR

The cell numbers of purified Coxiella isolates can be determined by Gimenez stain. To assess whether the cell densities quanitified by real-time PCR were comparable, we performed PCR reactions of heat inactivated isolates targeting the icd marker without previous DNA extraction. An exponential dilution series was made from heat inactivated particles of the Nine Mile isolate containing 4.2 × 109 particles per ml, and 1 μl of each dilution was applied per PCR reaction. Cell numbers were quantified using standard curves derived from diluted plasmid standards. The results are shown in Table 4. Given that only one copy of the chromosome is present per bacterial cell, which can be expected for a slow growing bacterium like Coxiella, the number of genome equivalents based on icd quantification should be comparable to the number of bacteria. Indeed, the icd quantity correlated well with the numbers of coxiellae determined microscopically. The detection limit for real-time PCR was 17 particles per reaction, which is in good agreement with the detection limit for purified Coxiella DNA and far below the particle number that can be quantified microscopically.

Table 4

Comparison of microscopical and PCR-based determination of Coxiella cell numbers. The PCR measurements were performed in duplicate; mean values are shown.

Coxiella particles per μl
Determined microscopicallya	Quantified by PCRb
4.2 × 106	5.00 × 106
4.2 × 105	3.05 × 105
4.2 × 104	3.15 × 104
4.2 × 103	2.45 × 103
4.2 × 102	1.45 × 102
4.2 × 101	1.70 × 101
4.2 × 100	0

aCell numbers were only counted from undiluted sample.

bReal-time PCR targeting the icd marker.

Determination of the number of IS elements in 75 different Coxiella isolates

Although the measured icd and IS1111 copy numbers shown in Table 3 exceeded the calculated numbers in some cases, the number of IS1111 elements per genome (i.e., per icd copy) varied between 13 and 17 for different DNA concentrations, which is close to the published number of 20 IS1111 elements for the Nine Mile isolate. Therefore, with this assay, DNA samples of 75 isolates of C. burnetii from all over the world were assessed for presence of the icd and IS markers and the numbers of IS1111 elements per genome were calculated. Quantification of icd and IS markers was based on standard curves obtained from diluted plasmids. Each DNA sample was tested in duplicate in three independent PCR runs targeting both markers except for DNA from the Nine Mile isolate, where six runs were performed.

In a recent study where Q fever patients were examined 12 years after infection, the IS1111 element could not be amplified, whereas PCR for other targets was positive [21]. Our results indicated that all isolates contained both the icd and the IS1111 markers. Different PCR runs resulted in discrepancies of the measured quantities and accordingly, different values and standard deviations for the number of IS1111 elements per genome equivalent were obtained (data not shown). For the Nine Mile RSA493 strain the number of IS elements was determined to be 23 (± 3.43), which is in good agreement with the number revealed by sequencing. The mean number of IS1111 elements per genome varied between seven (isolate J 3) and 110 (isolate Z2534), and between 10 and 30 for the majority of isolates. In French isolates of the related restriction groups 12 to 16 (Table 1), however, the number of IS1111 elements was found to be above 30, being highest in strain "Raphael" (around 95). All isolates of restriction group I had numbers below 30 insertion elements, so that for these isolates a correlation of the number of IS1111 elements with the restriction group seems likely. In other restriction groups, however, the number of IS1111 elements was highly variable. Although the standard deviations were very high for some values, our data suggest that the number of IS1111 elements can vary widely between different C. burnetii isolates and some isolates seem to contain a very high number of IS1111 elements. To further confirm our real-time PCR based quantification, Southern blot analyses should be performed.

Table 1

Characteristics of C. burnetii isolates used in this study

C. burnetii Isolate	Restriction Group [23,24]	Geographical Origin
Nine Mile RSA493	I	USA
Balaceanu	I	Romania
Hardthof	I	Germany
Bernard	I	France
CS 1	I	Slovak Republic
CS 3	I	Slovak Republic
CS 4	I	Slovak Republic
CS 5	I	Slovak Republic
CS 6	I	Slovak Republic
CS 7	I	Slovak Republic
CS 8	I	Slovak Republic
CS 9	I	Slovak Republic
CS 10	I	Slovak Republic
CS 11	I	Slovak Republic
CS Dayer	I	Slovak Republic
CS L 35	I	Slovak Republic
CS Poland	I	Poland
J 1	I	Japan
J 3	I	Japan
J 27	I	Japan
Priscilla Q177	IV	USA
Scurry Q217	V	USA
CS S	V	USA
Dugway 5J108-111	VI	USA
Z 3027	VI	Germany
Z 3205a	VI	Germany
Z 3205b	VI	Germany
Z 3351	VI	Germany
Z 3568	VI	Germany
Z 3749	VI	Germany
Z 257	VI	Germany
Boren	1	USA
CS 48	1	Slovak Republic
CS II/Ia	1	Slovak Republic
CS F	1	Slovak Republic
CS Ixodes	1	Russia
CS S1	1	Russia
Florian	1	Slovak Republic
Frankfurt	1	Germany
München	1	Germany
Henzerling	1	Italy
RT 1	1	North Western Russia
RT 3	1	North Western Russia
Gbud	1	Slovak Republic
Geier	1	Romania
Andelfingen	2	Switzerland
Herzberg	2	Greece
CS Z 57	2	Slovak Republic
S 1	2	Sweden
S 4	2	Sweden
Soyta	2	Switzerland
Utvinis	2	Romania
Stanica	2	Romania
Z 3478	2	Germany
Z 3574	2	Germany
Z 4313	2	Germany
Z 4485	2	Germany
Z 104	2	Germany
Z 3464	4	Germany
Z 3567	5	Germany
Brustel	6	France
Z 2534	6	Austria
Z 3055	6	Germany
Z 2775	7	Germany
Brasov	8	Romania
Namibia	9	Namibia
Schperling	11	Kirgisia
Ouaret	12	France
Jaquemot	13	France
Campoy	13	France
Pallier	14	France
Lombardi	15	France
Raphael	16	France
Butin	16	France
Z 349-36/94	unknown	Germany

Insertion sequences play a major role in determining band pattern differences between isolates produced by methods such as PFGE (pulsed-field gel electrophoresis) in many bacterial species [22]. C. burnetii expresses a low degree of genetic heterogeneity among strains by DNA-DNA hybridization. However, Not I restriction of total DNA followed by PFGE resulted in the characterization of 20 restriction groups among 80 C. burnetii isolates collected worldwide, as indicated in Table 1[1,23,24]. Typing C. burnetii based on restriction fragment length polymorphisms of the locations of the IS1111 element, like published for the insertion sequence IS100 of Yersinia pestis [25], may add to the elucidation of the phylogenetic relationship of Coxiella isolates. Moreover, the insertion sites of IS1111 could be examined by inverse PCR or by a recently described technique, the so called vectorette PCR [26].

So far, our data are too incomplete for judgements on clinical outcome, namely, to find any correlation between the number of IS1111 elements and the virulence of an isolate. Nevertheless, it is tempting to speculate that an increased number of IS elements in the genome of an isolate could have a deteriorating effect on its fitness, because essential genes might be interrupted by the insertion sequences.

Conclusion

We validated TaqMan-based real-time PCR assays targeting the singular icd gene and the transposase of the IS1111a element present in multiple copies in the genome of C. burnetii. The assays were evaluated with a variety of other bacterial species and shown to be specific for C. burnetii. Dilution series of C. burnetii DNA and of plasmids with cloned icd and IS1111 inserts demonstrated the sensitivity of the assays. Less than 10 genome equivalents per reaction were reproducibly detected. Using the icd marker, cell numbers of C. burnetii isolates were quantified also at very low cell concentrations. As a first approximation, the combination of both assays was useful to assess the numbers of IS1111 elements in 75 C. burnetii isolates from all over the world. Our data indicate that the numbers of this insertion element in the different isolates seem to be highly variable. The differences in the content of IS1111 elements might be of importance for further phylogentic analyses of C. burnetii isolates.

Methods

Bacterial strains and growth conditions

The C. burnetii isolates used in this study are shown in Table 1. C. burnetii bacteria were grown in Buffalo green monkey cell cultures and isolated as described [7]. To determine bacterial concentrations, a defined volume of a diluted suspension was fixed on a slide and stained by the Gimenez method. Bacteria were counted and the concentration of the suspension was calculated.

The following DNA samples from other bacterial species were used as negative controls for PCR: Legionella pneumophila (ATCC 33152, JR32 and 130b), Francisella tularensis ssp. novicida (ATCC 15482) and ssp. tularensis (Schu4), Bacillus subtilis (DSM 347), Bacillus anthracis (UD III-7), Bacillus cereus (DSM 31), Bacillus thuringiensis (DSM 350), Bacillus megaterium (DSM 90), Bacillus licheniformis (DSM 13), Staphylococcus aureus (DSM 20231), Streptococcus equi (ATCC 9528), Pseudomonas putida (ATCC 12633), Pseudomonas aeruginosa (ATCC 9027), Pseudomonas fluorescens (ATCC 49838), Burkholderia mallei (RR0053), Burkholderia pseudomallei (ATCC 23343), Burkholderia stabilis (CCUG 34168), Burkholderia multivorans (CCUG 37240), Yersinia enterocolitica (O:8 Ye/80), Yersinia pseudotuberculosis (DSM 8992), Yersinia pestis (Kim), Brucella melitensis biotype 1 (16M Weybridge), Brucella abortus biotype 1 (544 Weybridge), Brucella suis biotype 1 (1330 Weybridge), Brucella ovis biotype 1 (63/290 Weybridge), Klebsiella oxytoca (CCUG 15788), Serratia marcescens, Proteus mirabilis, and Escherichia coli (DSM 30083). The DNA preparations of L. pneumophila were kind gifts from Dr. A. Flieger (NG 5, Robert Koch-Institut).

DNA extraction

C. burnetii isolates were mixed with an equal volume of ATL Tissue Lysis Buffer (Qiagen, Hilden, Germany) and heat inactivated (90°C, 20 min). DNA was extracted from 400 μl of this suspension according to the protocol for Gram-negative bacteria of the DNeasy Tissue Kit (Qiagen) and eluted in 100 μl of AE buffer.

Primers and probes for real-time PCR

The icd assay targets a 76 bp fragment of the C. burnetii icd gene.

Primers:

forward, icd-439F = CGTTATTTTACGGGTGTGCCA (439–459)

reverse, icd-514R = CAGAATTTTCGCGGAAAATCA (494–514)

TaqMan probe:

icd-464TM = FAM-CATATTCACCTTTTCAGGCGTTTTGACCGT-TAMRA-T (464–492).

The numbers in brackets show the positions based on the GenBank accession no. AF146284.

The IS1111 assay targets a 295 bp fragment of the transposase gene of the C. burnetii IS1111a element.

Primers:

forward, Cox-F = GTCTTAAGGTGGGCTGCGTG (219–238)

reverse, Cox-R = CCCCGAATCTCATTGATCAGC (493–513)

TaqMan probe:

Cox-TM = FAM-AGCGAACCATTGGTATCGGACGTT-TAMRA-TATGG (259–287).

The numbers in brackets show the positions based on the GenBank accession no. M80806.

All sequences are given in 5'-3' orientation. Primers and probes were designed using the Primer Express software (Applied Biosystems, Darmstadt, Germany) and purchased from TIB Molbiol (Berlin, Germany).

Preparation of plasmid standards

The target sequences were amplified by conventional PCR using DNA from C. burnetii Nine Mile RSA493 strain as template and with the same primers as for real-time PCR in the case of the IS1111 marker and with primers icd-418F (5'-TATGTTTGCCTTAGGCCCGT) and icd-818R (5'-AAGGGCTTTGCTCCAAATTC) in the case of the icd marker, for which a 401 bp long amplicon was obtained. Plasmid standards with cloned (TOPO TA Cloning System, Invitrogen, Karlsruhe, Germany) and sequenced inserts were generated by GenExpress (Berlin, Germany). Plasmid preparations were quantified spectrophotometrically, and plasmid copy numbers were calculated. Dilutions of the plasmids were used in real-time PCR reactions to prepare standard curves for quantification of the initial copy numbers.

PCR assay conditions

Real-time PCR reaction mix consisted of 6.25 μl of Universal Master Mix (Applied Biosystems, Darmstadt, Germany) containing dNUTPs, MgCl2, reaction buffer and AmpliTaq Gold DNA polymerase, 300 nM of each primer and 100 nM of fluorescence-labeled TaqMan probe. For most assays, water was added to a final volume of 24 μl, and 1 μl of purified template DNA or heat inactivated C. burnetii isolate was used as template. For determination of the IS1111 copy numbers in the 75 C. burnetii isolates, water was added to a final volume of 15 μl, and 10 μl of 10-fold dilutions of the DNA were used as templates to minimise pipetting errors. All real-time PCR reactions were performed in duplicate in a 7700 Sequence Detection System (Applied Biosystems) as follows: 2 min at 50°C, 10 min at 95°C, 40 cycles at 15 s 95°C and 30 s at 60°C. Data were analyzed with the corresponding software.

Probit analysis

The number of C. burnetii Nine Mile genome equivalents (GE) with a genome size of 1,995,275 bp in a DNA preparation with a concentration of 29 ng/μl was calculated to be 1.3 × 107 GE/μl. To determine the number of GE that can be detected with a probability of 95 %, eight replicates of serial DNA dilutions from 100 GE/reaction to 0.75 GE/reaction for icd or 25 GE/reaction to 0.2 GE/reaction for IS1111 were tested in independent PCR reactions performed by different persons. The reaction volume was 1 μl. Each PCR gave a positive or negative result at the concentration tested. The detection probability was obtained by plotting the proportion of positive PCRs observed against the genome equivalents. Statistical analysis was performed using the SAS version 9.1 software.

Authors' contributions

SRK designed and coordinated the study, drafted the manuscript and participated in performing real-time PCR assays. JT and GB were responsible for the cultivation of the C. burnetii isolates, participated in the design of the study and helped to draft the manuscript. HE participated in the design of the study, the evaluation of the PCR assays and helped to draft the manuscript. TF isolated C. burnetii DNA and participated in performing PCR assays. SL participated in DNA isolation and evaluation of the PCR assays. BA participated in the design of the study, provided technical and financial support, and helped to draft the manuscript. All authors read and approved the final manuscript.