Use of Multiplex Real-Time PCR To Diagnose Scrub Typhus

ABSTRACT

Scrub typhus, caused by Orientia tsutsugamushi, is a common cause of acute undifferentiated febrile illness in the Asia-Pacific region. However, its nonspecific clinical manifestation often prevents early diagnosis. We propose the use of PCR and serologic tests as diagnostic tools. Here, we developed a multiplex real-time PCR assay using hydrolysis (TaqMan) probes targeting O. tsutsugamushi 47-kDa, groEL, and human interferon beta (IFN-β gene) genes to improve early diagnosis of scrub typhus. The amplification efficiency was higher than 94%, and the lower detection limit was 10 copies per reaction. We used a human gene as an internal DNA quality and quantity control. To determine the sensitivity of this PCR assay, we selected patients with confirmed scrub typhus who exhibited a clear 4-fold increase in the level of IgG and/or IgM. The PCR assay result was positive in 45 of 52 patients, indicating a sensitivity of 86.5% (95% confidence interval [CI]: 74.2 to 94.4). The PCR assessment was negative for all 136 non-scrub typhus patients, indicating a specificity of 100% (95% CI: 97.3 to 100). In addition, this test helped diagnose patients with inconclusive immunofluorescence assay (IFA) results and using single blood samples. In conclusion, the real-time PCR assay proposed here is sensitive and specific in diagnosing scrub typhus. Combining PCR and serologic tests will improve the diagnosis of scrub typhus among patients presenting with acute febrile illness.

INTRODUCTION

Scrub typhus is a mite-borne infectious disease caused by the obligate intracellular bacterium Orientia tsutsugamushi. A characteristic feature of the disease is patients presenting nonspecific symptoms, including fever, headache, myalgia, cough, and abdominal pain, which cannot be differentiated from symptoms of other systemic infections. The presence of eschar can help diagnose this illness; however, it is found only in some patients (1). Although the clinical course of scrub typhus is usually mild and self-limiting, delaying the treatment in severe cases can lead to complications such as renal failure, myocarditis, meningoencephalitis, and death (2). Since scrub typhus is one of the most common causes of acute undifferentiated febrile illness (AUFI) in areas of endemicity (3, 4), an early, definite diagnosis is essential for providing appropriate treatment and gathering accurate epidemiological data.

Scrub typhus diagnosis mainly relies on serologic tests, particularly the indirect immunofluorescence assay (IFA), whereby the illness is identified by a 4-fold increase in antibody titers in paired sera (5, 6) and/or a positive IgM titer in a single serum sample (7, 8). However, these serologic tests require paired serum samples and good technician expertise, and even then they often return false negatives during the early phase of disease. In addition, reinfection by different O. tsutsugamushi strains is not uncommon in areas of endemicity and reinfected patients may sometimes exhibit high and persistent titers of both IgG and IgM, rendering serologic tests useless to distinguish acute from past infections in these patients (5). Set against this background, molecular techniques such as real-time PCR could be useful to confirm serological results.

The use of real-time PCR assays targeting the 47-kDa gene or 16S rRNA gene (using hydrolysis probes) or the 60-kDa heat shock groEL gene (using SYBR green) has been reported previously (9–11). However, multiplex real-time PCR using several probes labeled with different fluorochromes has never been used to diagnose scrub typhus. Since the genome of O. tsutsugamushi shows a high degree of genetic polymorphism, single-target molecular assays may yield false-negative results related to primer-template or probe-template mismatches. Therefore, we developed a multiplex real-time PCR assay using hydrolysis (TaqMan) probes targeting the genes encoding the 47-kDa antigen and the GroEL protein. We evaluated the levels of sensitivity and specificity of this assay using a large number of clinical samples from patients with AUFI. We used the human interferon beta (IFN-β) gene as an internal control. We compared the real-time PCR data obtained to the IFA results.

RESULTS

Evaluation of PCR efficiency and specificity.

We obtained a linear standard curve with good correlation coefficients (R² = 0.99) based on our real-time PCR results. The amplification efficiencies for the 47-kDa gene and the groEL gene were 95% and 98%, respectively. The lower limit of quantification, defined as the lowest concentration at which the standard curve was still linear, was 10² copies per reaction (Fig. 1). The amplification curve of this multiplex PCR was comparable to those of separate singleplex PCRs, which yielded efficiencies of 97% and 95% for the 47-kDa gene and the groEL gene, respectively (see Fig. S2 in the supplemental material). The standard curve for the IFN-β gene internal control also showed a good amplification efficiency of 97.5% (Fig. S3). The coefficients of variation (CVs) at 10³ and 10² copies per reaction were 7% and 26% (47-kDa gene) and 12% and 32% (groEL gene), respectively (see Table S1 in the supplemental material). We used real-time PCR in the presence of low concentrations of O. tsutsugamushi DNA to determine the limit of detection (LOD) of this assay. At 10 copies of DNA per reaction, 100% of the samples were still PCR positive for both genes. Therefore, the LOD of this assay was 10 copies per reaction. At 5 copies, 90% and 75% of the samples were positive for the 47-kDa gene and the groEL gene, respectively (Table 1, upper panel). However, 100% of the samples were positive for either of the 2 genes. This multiplex PCR yielded positive results with four strains of O. tsutsugamushi (Karp, Gilliam, Kato, and TA716) available in our laboratory. For the singleplex PCR assay, the LODs for 47-kDa gene and groEL gene PCR were 5 and 10 DNA copies per reaction, respectively. Although the LOD of singleplex PCR for the 47-kDa gene was lower than that of multiplex PCRs, we continued using multiplex PCR for its advantages of being more convenient and economical than individual singleplex PCRs.

FIG 1.

Standard curves of the O. tsutsugamushi 47-kDa gene and groEL gene for the multiplex PCR. The amplification efficiency was calculated using the Bio-Rad program. O. tsutsugamushi standard DNA was negative for the human IFN-β gene. FAM and VIC represent 6-carboxyfluorescein (FAM)-labeled 47-kDa gene and Yakima yellow-labeled groEL probes, respectively. VIC has the same fluorescent spectrum as Yakima yellow.

TABLE 1.

Determination of the detection limits of PCR using pure O. tsutsugamushi DNA and O. tsutsugamushi with human DNA

Strain and copy no.	Value(s)a
	47-kDa gene		groEL
	% positive samples	Cq variation	% positive samples	Cq variation
Pure O. tsutsugamushi DNA
10	100	36.40 (0.86)	100	36.79 (1.07)
5	90 (10)	37.52 (0.93)	75 (8.7)	37.72 (1.17)
O. tsutsugamushi with human DNAb
10	100	36.55 (0.60)	100	36.37 (1.16)
5	95 (8.7)	37.01 (0.92)	80 (14.1)	37.12 (1.51)

^aPercent positive samples data represent means (standard deviations [SD]) of results from four experiments performed in five replicates. C_q variation data represent means (SD) of results from all positive samples.

^bHuman blood samples were spiked with known amounts of O. tsutsugamushi and subjected to DNA extraction.

To determine whether any components in the blood adversely affected the efficiency of this PCR assay, we “spiked” a defined amount of O. tsutsugamushi in the blood samples, extracted the DNA, and determined the LOD of this PCR again. Similarly to the PCR results obtained using a pure O. tsutsugamushi DNA template, the LOD of the PCR using O. tsutsugamushi with human DNA was 10 copies per reaction (Table 1). This finding suggests that the presence of human genomic DNA from blood did not have an inhibitory effect on the PCR assay. Therefore, this PCR assay is applicable for use with patient samples.

For the real-time PCR performed using patients' samples, we used 500 ng of human genomic DNA per reaction. The human genome is approximately 3 × 10⁹ bp in length; thus, 500 ng of human genomic DNA includes approximately 1.4 × 10⁵ copies from 7 × 10⁴ white blood cells (WBCs). Therefore, the LOD for this assay was 10 O. tsutsugamushi organisms in 7 × 10⁴ WBCs, or approximately 2 O. tsutsugamushi organisms per 1 μl of blood containing 1.4 × 10⁴ WBCs.

The specificity of the multiplex PCR assay was tested using DNA extracted from other pathogenic bacteria, including Staphylococcus aureus, Streptococcus pyogenes, Escherichia coli, Klebsiella pneumoniae, Salmonella enterica, Pseudomonas aeruginosa, Burkholderia pseudomallei, Leptospira interrogans, Bartonella henselae, Rickettsia typhi, Rickettsia japonica, and Rickettsia conorii. All real-time PCRs for these species yielded negative results.

Evaluation of the real-time PCR assay in patients.

We evaluated this PCR assay using samples from 458 in-patients (64.1% males, with a median age of 48 years and an age range of 18 to 101 years) presenting with AUFI. The flow of participants is illustrated in Fig. 2. The median duration of fever was 4 days (range, 3 to 14 days). The overall diagnosis of the febrile illnesses included scrub typhus (124 patients, or 27% [100 cases were confirmed by PCR or dynamic serology, and 24 cases were diagnosed by a positive IgM titer]); leptospirosis (59 patients, 12.9%); dengue virus infection (30 patients, 6.6%); other rickettsial infections (13 patients, 2.9%); bacteremia due to other bacteria (15 patients, 3.3%); suspected coinfections (7 patients or 1.5% [six cases of scrub typhus and leptospirosis coinfection and one case of scrub typhus and murine typhus coinfection]); and unknown causes (210 patients, 45.8%). Detailed demographic data of the patients are summarized in Table S2 in the supplemental material.

FIG 2.

Flow of participants who underwent diagnostic testing by IFA and real-time PCR.

All samples were positive for the internal control (IFN-β gene) with a quantification cycle (C_q) value between 22 and 25. Real-time PCR results were positive for scrub typhus in 100 patients (100 and 98 patients were positive for the 47-kDa gene and the groEL gene, respectively). Two 47-kDa gene-positive and groEL gene-negative patients were confirmed to be acute scrub typhus patients by nested PCR targeting the 56-kDa gene. The primer sequences of the 56-kDa gene are shown in Table S3. Among 52 patients with confirmed scrub typhus, the PCR was positive in 45 (86.5%) patients (95% confidence interval [CI]: 74.2 to 94.4). Among 38 patients with inconclusive IFA results, the PCR was positive in 22 (57.9%) patients (95% CI: 40.8 to 73.7). Fully negative PCR results were obtained from 136 non-scrub typhus patients. The sensitivity and specificity of this assay, which were calculated from confirmed and non-scrub typhus patients, were 86.5% (95% CI: 74.2 to 94.4) and 100% (95% CI: 97.3 to 100), respectively. In addition, using real-time PCR and IFA results, we identified five patients with secondary infection characterized by a high IgG titer with a relatively low IgM titer (12). Among these patients, four exhibited a high (≥1:3,200) IgG titer, with or without an apparent 4-fold increase; a relatively low IgM titer (≤1: 400); and positive PCR results. One patient, on the other hand, exhibited a 4-fold increase in IgG titer, a relatively low IgM titer, and a negative PCR result. Real-time PCR was also evaluated in patients with single blood samples. Among 29 patients with a positive single IgM titer (≥1:400), PCR was positive in 21 (72.4%) patients (95% CI: 52.8 to 87.3). Among 203 patients with a negative IgM titer, PCR still yielded positive results in 12 (5.9%) patients (95% CI: 3.1 to 10.1) (Table 2).

TABLE 2.

Number and percentage of positive PCR results for each patient group

Group (no.)	No. (% of cases: 95% CI) of cases with +ve PCR resulta
Patients with paired sera
Confirmed scrub typhus (n = 52)	45 (86.5: 74.2 to 94.4)
Inconclusive IFA results (n = 38)	22 (57.9: 40.8 to 73.7)**
Non-scrub typhus (n = 136)	0 (0: 0 to 2.7)
Patients with single sera
Positive IgM titer (n = 29)	21 (72.4: 52.8 to 87.3)
Negative IgM titer (n = 203)	12 (5.9: 3.1 to 10.1)
Total (n = 458)	100 (21.8: 18.1 to 25.9)

^a+ve, positive; **, P value = 0.002 (compared to the confirmed scrub typhus group by chi-square test).

In addition, a real-time PCR assay was able to quantify the O. tsutsugamushi organism load in blood. The O. tsutsugamushi DNA copy numbers were higher than the lower limit of quantification of 100 copies in 53 patients and were lower in 47 patients. The median number of DNA copies was 125 O. tsutsugamushi DNA copies per 500 ng of human genomic DNA per reaction. Few patients showed more than 10⁴ or 10⁵ copies of O. tsutsugamushi DNA (Fig. 3).

FIG 3.

O. tsutsugamushi DNA copies in patients with positive PCR results. The horizontal line represents the median of DNA copy numbers. Data below the lower limit of quantification of 10² copies are depicted as 10⁰.

PCR results in patients with different IFA antibody titers on admission.

The provisional diagnosis of scrub typhus is usually based on a single IgM titer on admission. An IgM titer of 1:400 is routinely accepted in some countries, including Thailand, as the cutoff value to diagnose scrub typhus (7, 8). Therefore, we analyzed the PCR results from patients with different antibody titers on admission (Table 3; see also Table S4). Of 79 patients, 63 (79.7%) presenting with a positive IgM titer on admission were PCR positive. The PCR results from patients with a positive IgM titer (≥1:400) were significantly different from those obtained from patients with an IgM titer in the range of 1:50 to 1:200 (P value = 0.001), suggesting that an IgM titer of 1:400 is an appropriate cutoff value to provisionally diagnose scrub typhus. Since a high (≥1:3,200 or ≥1:12,800) IgM titer has been suggested as a diagnostic criterion to increase the specificity of diagnosis (13, 14), we further classified the patients with a positive IgM titer on admission into three groups: those with high (≥1:6,400), intermediate (1:1,600 to 1:3,200), and low (1:400 to 1:800) titers. There was no statistically significant difference in the PCR results among these groups (Table 3). Collectively, we suggest that a single IgM titer of at least 1:400 is acceptable to diagnose scrub typhus based on serum data; however, the definite diagnosis should be confirmed by detection of a dynamic change in serology using paired sera. Raising the cutoff value of the IgM titer does not seem to increase the accuracy of the diagnosis; rather, if reduces its sensitivity.

TABLE 3.

PCR results from patients with different IgM titers in acute plasma

IgM titer on admissiona	No. of patients
	PCR positive	PCR negative
≥1:400*	63	16
≥1:6,400	10	5
1:1,600–1:3,200	27	7
1:400–1:800	26	4
1:50–1:200*	25	24
<1:50	12	318
Total	100	358

^a*, P value (between the 2 groups) = 0.001 (by chi-square test).

Since provisional diagnosis in clinical practice sometimes relies on both IgG and IgM titers, we analyzed the PCR results obtained from patients with different IgG and IgM IFA results on admission. Among 50 patients with positive IgG (≥1:400) and positive PCR results, most of them (40/50 patients, 80%) were positive for both IgG and IgM, whereas only 10 patients (20%) were positive for only IgG. Among 50 patients with negative IgG but positive PCR results, 23 patients (46%) were positive for IgM. There was only one patient with positive IgM, negative IgG, and negative PCR results; however, this patient was diagnosed with confirmed scrub typhus by ≥4-fold increases in both IgG and IgM titers (Table 4). We then classified the patients into two groups: a PCR-positive group and a PCR-negative group. The positivity rates for IgM and IgG in each group of patients were compared using McNemar's test. In patients with positive PCR results, the positivity rates for IgG and IgM differed significantly, thus confirming that IgM analysis at the time of admission was more sensitive than IgG analysis for a provisional diagnosis of scrub typhus. In patients with negative PCR results, a statistically significant difference between the IgG and IgM titers was also observed, suggesting that the IgG titer at the time of admission may be associated with a past infection to a greater extent than the IgM titer (Table 4).

TABLE 4.

PCR results from patients with positive and negative antibody (IgG and IgM) titers, using a cutoff value of 1:400

Group	No. of patients
	IgM +ve	IgM −ve	Total
PCR positivea
IgG +ve	40	10	50
IgG −ve	23	27	50
Total	63	37	100
PCR negativeb
IgG +ve	15	22	37
IgG −ve	1c	320	321
Total	16	342	358

^aP value = 0.035 (by McNemar's test).

^bP value = <0.001 (by McNemar's test).

^cThe PCR assay results for the patient were negative; however, he was subsequently diagnosed with scrub typhus by 4-fold increases in both IgG and IgM titers.

DISCUSSION

Scrub typhus is a common cause of acute undifferentiated febrile illness in areas of endemicity. Although the diagnosis is usually confirmed by IFA, real-time PCR can improve early diagnosis before seroconversion and confirm IFA results. In this study, we developed a multiplex PCR assay targeting the 47-kDa gene and the groEL gene of O. tsutsugamushi, the scrub typhus causal agent. We included the human IFN-β gene as an internal control. The amplification efficiency of the reaction was higher than 90%, and the lower limit of detection was 10 DNA copies per reaction. The LOD of this assay to detect O. tsutsugamushi is comparable to that for detection of other blood-borne pathogens, ranging from 3 to 20 copies per reaction (15–17). We assessed the effectiveness of this real-time PCR assay in a large number of clinical samples. The sensitivity of this real-time PCR was 86%, and the specificity was 100%. In addition, real-time PCR helped confirm the diagnosis of patients with inconclusive IFA results. Taking the data together, this multiplex real-time PCR with an internal control is sensitive, specific, and practical for scrub typhus diagnosis.

Although several real-time PCR assays for scrub typhus diagnosis have been reported in previous studies (9–11), the use of multiplex real-time PCR with an internal control has never been published. The level of genetic polymorphism found in O. tsutsugamushi is quite high; consequently, assays based on a single gene sequence can yield false negatives due to mismatches between the template and the primers or probes. Here, we intentionally designed long probes (33 and 29 bp), which are known to be more tolerant of probe-template mismatches than short probes (18). In addition, a real-time PCR result with a quantification cycle (C_q) value of over 37 to 38 cycles may be considered to represent a negative or a false-positive result (19); therefore, detection of two genes increases the accuracy of the results. For example, samples positive for the two genes can be interpreted as true positives, regardless of the C_q values, whereas nested or repeated real-time PCRs would be required to confirm samples positive for only one gene. The presence of an internal control in our real-time PCR helps to ensure the quality and quantity of genomic DNA template. Thus, the detection of several genes by multiplex PCR yields more-accurate results. Using multiplex real-time PCRs is cheaper in time and cost than using individual singleplex PCRs. Real-time PCR is also more convenient than nested PCR and can reduce the turnaround time and the risk of carryover contamination. A previous study proposed that loop-mediated isothermal amplification (LAMP) is another nucleic acid detection test with the potential of becoming a point-of-care diagnostic test (13). However, multiplex LAMPs are difficult to design due to the complicated structure of the amplicons (20). Therefore, real-time PCR is more practical for multiplex reactions. The presence of an internal control in this real-time PCR renders this assay more practical and promising for use in service laboratories.

Due to the genetic diversity of O. tsutsugamushi organisms, this assay should be further evaluated in other areas of endemicity. In addition, the DNA sequence alignments in Fig. S1 in the supplemental material are mainly based on the O. tsutsugamushi strains isolated from Thailand and Southeast Asia. The O. tsutsugamushi strains Karp, Gilliam, and TA763 are predominantly found in Thailand (21). Other areas of endemicity, including East and South Asia, may harbor different O. tsutsugamushi strains, such as Boryong, Kato, Kawasaki, Kuroki, and Ikeda (22–24). Although these two O. tsutsugamushi genes are more conserved than the 56-kDa gene, further studies are required to ensure the diagnostic accuracy of this PCR assay with other O. tsutsugamushi strains not commonly found in Thailand. In addition, since the groEL gene is also present in other bacteria, this real-time PCR assay should be further tested with other closely related bacteria (i.e., Ehrlichia chaffeensis, Anaplasma phagocytophilum, and Neorickettsia spp.) to exclude the possibility of cross-reactivity associated with this assay.

The efficacy of our real-time PCR assay was evaluated using a large number of patients presenting with AUFI. The patients were classified into different groups based on their IFA patterns. Since IFA results are sometimes equivocal for some patients and cannot discriminate between acute and past infections, we determined the sensitivity of this real-time PCR assay using only patients with a clear 4-fold increase in IFA titers. The level of sensitivity reported here, 86.5% (95% CI: 74.2 to 94.4), is higher than that reported in previous studies, where the level ranged from 60% to 85% (19, 25). The higher sensitivity of our assay probably relates to a more accurate classification of the patients and differences in the strains of causative O. tsutsugamushi organisms. In addition, we also used a larger number of clinical samples than previous studies (11, 19, 25, 26). A substantial proportion of patients (38 cases) showed equivocal IFA results, including persistently high titers of both IgG and IgM without an apparent 4-fold increase. Real-time PCR was helpful to confirm the diagnosis in these patients. As expected, PCR was positive in only 58% of the patients with inconclusive IFA results, compared to 86% of patients with confirmed IFA results. This suggests that high and persistent antibody titers can be observed in patients with past or recent infection and that detecting the causal organism by PCR is necessary to distinguish between acute and past infections.

Although IFA is considered the gold standard test because of its high (96% to 98%) specificity (6, 7), its sensitivity is only approximately 60% to 80% (7, 27). The low sensitivity of the diagnostic criterion using a 4-fold increase in the titer of paired sera is likely due to the serotypic diversity of O. tsutsugamushi organisms and inappropriate timing of specimen collection (7). Indeed, we found that a few of the patients with the inconclusive IFA and positive PCR results exhibited a rapid decline in IgM and/or IgG titers within 2 weeks. On the other hand, some patients exhibited high antibody titers (≥1:6,400) in the paired plasma samples. Those patients would not be diagnosed with scrub typhus using the dynamic serology criterion. On the basis of the results of this study, only 52 patients would be diagnosed with scrub typhus by the IFA titers of the paired plasma. However, 22 additional patients with inconclusive IFA results could be diagnosed by PCR. Thus, the sensitivity of the dynamic serology criterion in this study was 52/74 or approximately 70%, which is comparable to that of previous studies (7, 27). Combining the PCR and IFA titers of the paired sera will increase the diagnostic sensitivity without compromising specificity.

In clinical practice, the provisional diagnosis is typically based on the antibody titer on admission. We have shown that about 80% of patients with an IgM titer of ≥1:400 had positive PCR results, whereas only 50% to 60% of the patients with an IgM titer of between 1:50 and 1:200 exhibited positive PCR results (Table 3; see also Table S4 in the supplemental material). Therefore, the cutoff value of an IgM titer of ≥1:400 is appropriate for the area of this study. However, some experts in different areas of endemicity may use a different cutoff value, depending on the local evidence (5, 27). Since severe scrub typhus may present with clinical manifestations of sepsis (28), broad-spectrum antimicrobial treatment is usually prescribed to these patients. Knowledge of the positive PCR or IFA results combined with negative hemoculture results may benefit the deescalation of antimicrobial treatment.

In conclusion, we demonstrate the usefulness of a novel multiplex real-time PCR assay to diagnose scrub typhus. This multiplex PCR assay includes two target O. tsutsugamushi genes and one human gene used as an internal control. The amplification efficiency of the reaction and the lower limit of detection were acceptable for this purpose. After applying this real-time PCR assay to a large number of clinical samples, it showed sensitivity and specificity levels of 86% and 100%, respectively. Therefore, this real-time PCR assay could represent a promising new tool for early diagnosis of scrub typhus. Used in combination with serologic tests, this real-time PCR assay can provide more-accurate prevalence figures for scrub typhus in areas of endemicity.

MATERIALS AND METHODS

Primer and probe design.

We obtained previously published DNA sequences of the 47-kDa gene and the groEL gene of several O. tsutsugamushi strains (11, 29) from GenBank, aligned the sequences using the BioEdit program, and manually designed the primers and probes from the most conserved regions. The 47-kDa gene is unique in O. tsutsugamushi and is absent in members of the genus Rickettsia (9). Although the groEL gene is present in other bacteria as well, there are some regions that are highly conserved among O. tsutsugamushi strains but differ from those in members of the genus Rickettsia (11). DNA sequence alignments of these two genes are shown in Fig. S1 in the supplemental material. The human IFN-β gene was selected as an internal control because it is a small gene with approximately 50% GC content. We used the Oligo Analyzer program (freeware provided by TeemuKuulasmaa, Finland) to check for potential hairpin loop and primer dimer formation. PCR products of 47-kDa, groEL, and human IFN-β genes were located at positions 700 to 817 (118 bp), 1069 to 1164 (96 bp), and 419 to 503 (85 bp), respectively. Table 5 shows the sequences and final concentrations of the primers and probes used.

TABLE 5.

Primer and probe sequences

Primer or probe	Sequence (5′→3′)a	Concn (nM)
Primers
OT 47 kDa F1	CCA TCT AAT ACT GTA CTT GAA GCA GTT GA	400
OC 47 kDa F2b	CCA ACA CTG TAT TAG AAG CAG TTG A	200
OT 47 kDa R	GTC CTA AAT TCT CAT TTA ATT CTG GAG T	400
OT groEL F	GCW GTT GCT CAT ACT GGC AA dICC	400
OT groEL R	GGA ACC TTT TAA ATT GTT TAA TAT CAA TGC	400
hu IFN-β F	GTC TGC ACC TGA AAA GAT ATT ATG G	67
hu IFN-β R	CTG ACT ATG GTC CAG GCA CA	67
Probes
OT 47-kDa Rc	FAM-TCA TTA AGC/ZEN/ATA ACA TTT AAC ATA CCA CGA CGA-IBFQ	200
OT groEL Rd	Yakima yellow-AAG AGC TTC TCC GTC TAC ATC ATC AGC AA-BHQ1	200
hu IFN-β Rd	Cy5-CTC CTT GGC CTT CAG GTA ATG CAG-BHQ3	67

^aBHQ1, black hole quencher 1; BHQ3, black hole quencher 3; dI, 2′-deoxyinosine; FAM, 6-carboxyfluorescein; IBFQ, Iowa Black FQ quencher.

^bSelected from O. tsutsugamushi strain Chuto (GenBank accession no. HM156063.1).

^cPurchased from Integrated DNA Technologies (Coralville, IA).

^dPurchased from Jena Bioscience (Jena, Germany).

Patients and specimen collection.

In this study, we used blood specimens collected at the Maharat Nakhon Ratchasima Hospital, Northeast Thailand, for a previously published prospective hospital-based study in adult patients with acute undifferentiated fever (30). The participants recruited were adult in-patients (≥18 years) who presented with acute fever (oral temperature of 38.0°C or higher for less than 15 days) in the absence of an obvious focus of infection. The median duration of fever was 4 days (range, 3 to 14 days). Patients with malaria and those with clinically obvious dengue virus infection (who met the criteria of the World Health Organization for diagnosing dengue infection) were excluded. The study protocols were approved by the Ethical Review Subcommittee of the Public Health Ministry of Thailand and the Ethical Review Subcommittee of the Faculty of Medicine, Siriraj Hospital, Mahidol University. All participants provided written inform consent to be included in the study.

Acute and convalescent blood samples were obtained on admission and at an outpatient visit at 2 to 4 weeks after discharge. EDTA blood samples were centrifuged at 2,000 × g for 5 min. Plasma and buffy coats were subjected to serologic tests and PCR, respectively. Serologic testing for the diagnosis of leptospirosis, scrub typhus, murine typhus, and spotted fever rickettsioses was performed by IFA using Leptospira interrogans serogroup Autumnalis, O. tsutsugamushi (pooled Karp, Gilliam, and Kato strains), Rickettsia typhi, and pooled spotted fever group rickettsiae (R. japonica and R. conorii), respectively. Dengue infection was diagnosed by complete blood counts (CBC) and an SD Bioline rapid immunochromatography test for NS1Ag, IgM, and IgG (Standard Diagnostics, South Korea). For IFA, multiwell slides were coated with paraformaldehyde-killed organisms, fixed, and permeabilized in ice-cold acetone for 10 min. Patients' plasma samples were 2-fold serially diluted from 1:50 to 1:6,400 in phosphate-buffered saline (PBS), transferred to IFA slides, and incubated for 30 min at 37°C in a humidified chamber. The IFA slides were washed three times in PBS for 5 min each time and incubated with fluorescein isothiocyanate (FITC)-conjugated rabbit anti-human IgM and anti-human IgG (Dako, Glostrup, Denmark) diluted at 1:40 for 30 min at 37°C. The IFA slides were washed three times in PBS, and the host cells were counterstained with 0.00125% Evans blue during the last washing step. For the purpose of this study, the patients were classified into the following groups on the basis of their O. tsutsugamushi IFA results as follows: (i) patients with confirmed scrub typhus, defined as patients with at least a 4-fold increase in IgG and/or IgM titers; (ii) patients with equivocal IFA results, including those with very high (≥1:6,400) IgG and IgM titers upon admission and those with positive (≥1:400) IgG and IgM titers without a 4-fold increase in convalescent plasma; (iii) patients with non-scrub typhus, defined as patients with IgG and IgM titers lower than 1:400 without a dynamic increase in paired plasma; (iv) patients with a positive (≥1:400) IgM titer in single plasma; and (v) patients with a negative (<1:400) IgM titer in single plasma. Leptospirosis was diagnosed by a 4-fold increase in IFA titer, a positive IgM titer (≥1:400) in single plasma, or positive PCR results (30). Other rickettsioses were diagnosed by a 4-fold increase in IFA titer or a positive (≥1:400) IgM titer in single plasma. Dengue virus infection was diagnosed by positive NS1Ag or IgM or positive IgG results with a compatible clinical presentation.

Real-time PCR.

Genomic DNA was extracted from patients' buffy coats using a QIAamp DNA minikit (Qiagen, Hilden, Germany). Genomic DNA concentrations were measured in duplicate by the use of a NanoDrop spectrophotometer (Thermo Scientific, Waltham, MA). To prepare the working DNA solutions, we diluted the DNA samples to a final concentration of 65 ng/μl. The PCR mixture contained 20 μl of 2× Kapa Probe Fast quantitative PCR (qPCR) master mix (Kapa Biosystems, Woburn, MA) and included 5 mM MgCl₂, primers and probes at the concentrations provided in Table 5, and 8 μl of DNA template in a final volume of 40 μl. We used 500 ng of genomic DNA from blood samples per PCR. The PCRs were performed in a Bio-Rad CFX96 real-time thermocycler (Bio-Rad, Hercules, CA) with the following programmed conditions: initial denaturation and hot-start enzyme activation at 95°C for 3 min followed by 45 cycles of denaturation at 95°C for 1 s, combined annealing/extension at 58°C for 30 s with data acquisition for 12 s, and extension at 70°C for 3 s. To prepare O. tsutsugamushi DNA standards for PCR, O. tsutsugamushi-infected EA-hy926 human endothelial cells were disrupted by repeated passage through a 25-gauge needle using a syringe. The cell suspension was centrifuged at 400 × g for 5 min, and the supernatant containing extracellular O. tsutsugamushi organisms was filtered through a 5-μm-pore-size filter using a syringe to remove host cell nuclei. The absence of human nuclear DNA was confirmed by targeting the human IFN-β gene during PCR. To quantify the O. tsutsugamushi genomic DNA, eight replicates of 2-μl aliquots from an O. tsutsugamushi DNA standard were measured using a NanoDrop spectrophotometer (Thermo Scientific). The number of copies of O. tsutsugamushi DNA was calculated using the DNA Copy Number and Dilution Calculator (Thermo Scientific). As the whole-genome length of O. tsutsugamushi is approximately 2.1 Mbp, DNA standards were serially diluted from 10⁷ copies per reaction mixture to 10² copies. To create a standard curve, we performed the real-time PCR using triplicate samples for each standard concentration. Then, to estimate amplification efficiency, we used the Bio-Rad program and the following formula: efficiency = (10^−1/slope − 1) × 100%. To demonstrate the repeatability (intra-assay variation), we calculated the coefficient of variation (CV) (for copy number variance) using the following formula: CV = (standard deviation [SD]/mean of copy number) × 100%. To determine the lower limit of detection (LOD), real-time PCRs in the presence of low DNA concentrations (20, 10, and 5 copies) were performed in five replicates and repeated four times. The LOD was determined from the lowest concentration of DNA that yielded a positive PCR result in more than 95% of the samples. We used both pure O. tsutsugamushi DNA (without human DNA) and O. tsutsugamushi with human genomic DNA for LOD determinations. To prepare O. tsutsugamushi with human genomic DNA, we spiked known amounts of O. tsutsugamushi organisms into human blood samples and extracted DNA.

Statistical analysis.

Data were analyzed by nonparametric tests (chi-square or McNemar's test), as indicated. P values of <0.05 were considered significant.