Rapid Diagnostics of Joint Infections Using IS-Pro

ABSTRACT

Diagnosis of bone and joint infections (BJI) relies on microbiological culture which has a long turnaround time and is challenging for certain bacterial species. Rapid molecular methods may alleviate these obstacles. Here, we investigate the diagnostic performance of IS-pro, a broad-scope molecular technique that can detect and identify most bacteria to the species level. IS-pro additionally informs on the amount of human DNA present in a sample, as a measure of leukocyte levels. This test can be performed in 4 h with standard laboratory equipment. Residual material of 591 synovial fluid samples derived from native and prosthetic joints from patients suspected of joint infections that were sent for routine diagnostics was collected and subjected to the IS-pro test. Bacterial species identification as well as bacterial load and human DNA load outcomes of IS-pro were compared to those of culture. At sample level, percent positive agreement (PPA) between IS-pro and culture was 90.6% (95% CI 85.7- to 94%) and negative percent agreement (NPA) was 87.7% (95% CI 84.1 to 90.6%). At species level PPA was 80% (95% CI 74.3 to 84.7%). IS-pro yielded 83 extra bacterial detections over culture for which we found supporting evidence for true positivity in 40% of the extra detections. Missed detections by IS-pro were mostly related to common skin species in low abundance. Bacterial and human DNA signals measured by IS-pro were comparable to bacterial loads and leukocyte counts reported by routine diagnostics. We conclude that IS-pro showed an excellent performance for fast diagnostics of bacterial BJI.

INTRODUCTION

Bone and joint infections (BJI) are potentially devastating infections of native and prosthetic bone and joints. Native joint infection, or septic arthritis, is caused by bacterial infection of synovial (joint) fluid and joint tissues. This serious, rapidly progressive condition can cause permanent damage to the infected joint. Prosthetic joint associated infection is a complication of orthopedic arthroplasty, in which the prosthetic joint becomes infected by bacteria. Both infections require rapid diagnosis and appropriate antibiotic treatment. Additionally, BJI often require prompt surgical interventions in which the infected joint is either debrided or replaced. Detection of bacteria in the affected joint is essential for diagnosing BJI (1).

Currently, the reference standard for clinical diagnosis of BJI is culture, which is slow to inform physicians on the optimal course of treatment. In acute infections, culture typically takes 2 to 7 days, whereas in cases of chronic infection, where bacteria may be present in biofilms, culture can take up to 14 days (2). Gram-staining of synovial fluid can be performed rapidly but has very low sensitivity (3) and can therefore not be used to exclude joint infection. Because of these issues, physicians are forced to treat patients empirically, based on clinical suspicion of BJI rather than on a confirmed diagnosis. This results in overprescription of broad-spectrum antibiotics and unnecessary operations. Molecular procedures, which use DNA as input, offer a much-needed improvement over culture in timeliness of diagnosis. Several PCR assays, detecting panels of species were developed for use in BJI diagnostics (4–6). However, these tests only detect a predefined set of microorganisms. Consequently, these assays cannot be used to rule out BJI, and can only be used as adjuvants to bacterial culture.

A more comprehensive approach is provided by broad-range PCR or by metagenomics next-generation sequencing (mNGS). Broad-range PCR involves amplification of (part of) the 16S gene, followed by sequencing. This approach has been tested for BJI diagnostics in multiple studies, with a wide range of performances reported (7, 8). Metagenomic NGS involves sequencing of all DNA present in a sample (9–11). The sensitivity for pathogen detection by this assay is strongly affected by the amount of human DNA in the sample, which can be overwhelming compared to the amount of pathogen DNA. To obtain sufficient sensitivity, removal of human DNA is required, which in some cases may also affect detection of certain microorganisms. Both broad-range PCR and mNGS currently involve laborious complex laboratory and bioinformatic workflows and have not been standardized or optimized for implementation in clinical routine.

Here, we evaluate the performance of Molecular Culture (inbiome) for diagnosis of BJI. Molecular Culture is based on the IS-pro assay; a PCR based molecular technique that combines length polymorphisms of the 16S to 23S interspace rDNA with phylum-specific fluorescently labeled primers to detect and identify bacteria to species level (12, 13). This test combines the comprehensiveness and speed of broad-range PCR with simplicity of identification. In addition, the assay assesses the amount of human DNA as an infection marker, and it contains a multi-purpose internal control that is used to control amplification efficiency as well as the downstream automated length calling procedure. The test was successfully applied to diagnose infections in normally sterile body sites (14–16). The assay can be completed in approximately 4 h, thus offering a much faster turnaround time than culture, broad-range PCR, and mNGS.

In this study, we evaluated the performance of IS-pro to detect bacteria on a large (n = 591) set of native and prosthetic joint aspirates by comparing the outcome of IS-pro to that of routine bacterial culture.

MATERIALS AND METHODS

Sample collection.

Residual material of consecutive synovial fluid samples that were sent for routine diagnostics was collected between 2013 and 2016 at the Department of Medical Microbiology of Amsterdam UMC, location VUmc at Amsterdam, The Netherlands. Samples were obtained by arthrocentesis from subjects with clinical suspicion of BJI. The Medical Ethical Review Board of the VU Medical Centre Amsterdam ruled that this study was not subject to the Dutch Medical Research involving human subjects act (WMO), since subjects were not subjected to investigational therapeutic or diagnostic interventions.

Routine diagnostics.

Upon arrival at the laboratory, 100 μL of the sample was aliquoted and stored at −80°C for later analysis by the IS-pro assay. The remaining approximately 900 μL was pelleted and resuspended in a few drops of supernatant. This suspension was used for Gram stain, leukocyte count and culture. Culture was done on chocolate- and blood-agar plates in aerobic and anaerobic conditions, and in Brain Heart Infusion broth. In cases where no growth was detected on the primary plates, secondary plates were inoculated from the broth culture. Colonies growing on plates were identified by MALDI-TOF Vitek-MS (bioMérieux). Culture loads were described in 5 categories: negative, <1 (positive only on secondary plates), 1 (limited growth, only in first streaking segment), 2 (intermediate growth into second streaking segment, 10-100 colonies), and 3 (abundant growth, present in all 3 streaking segments, >100 colonies). Leukocyte counts were reported as none, few (1 or 2 per viewing field), medium (2 to 10 per viewing field), and many (>10 per viewing field).

IS-pro assay.

The IS-pro assay was carried out with the Molecular Culture kit according to the manufacturer’s instructions (inbiome). Briefly, 50 μL of sample was combined with 250 μL of Shock Buffer 1 (inbiome) and incubated at 95°C for 10 min while shaking at 800 rpm. Subsequently, 25 μL of Shock Buffer 2 (inbiome) was added to the samples. One mL of EasyMag lysis buffer (bioMérieux) plus 1 mL of AL buffer (Qiagen) was added before extraction of DNA using the Specific A Protocol in the automated EMAG extraction system (bioMérieux). DNA was eluted in a volume of 70 μL.

Two PCRs were performed, each using 10 μL of sample DNA. One PCR targets the phyla Firmicutes, Actinobacteria, Fusobacteria, Verrucomicrobia, and Bacteroidetes; the other PCR targets Proteobacteria, an internal amplification control (IAC), and human DNA. The final PCR products were combined and analyzed for amplicon length and fluorescence intensity (indicated by relative fluorescence units [RFU]) by an ABI3500 fragment analyzer (ThermoFisher). Analysis of resulting ABI3500 data was performed by Antoni, inbiome’s IS-pro Lab Cloud software which matches amplicon length to bacterial species using a dedicated database (inbiome). Cutoffs were determined by analyzing 50 non-template control samples. The highest peak intensities seen in these samples were set as background cutoffs. Samples yielding peaks above the cut-off intensities were classified as positive. Samples were classified as negative when such peaks were lacking but only if the IAC peaks were present.

Sequencing.

IS-pro PCR products with lengths that did not yet match any entry in the IS-pro database were sequenced on a MinION device (Oxford Nanopore Technologies). To this end, the PCR products obtained from the IS-pro reaction were diluted 1000x and subjected to the same PCR with non-fluorescently labeled primers targeting only the bacterial DNA (leaving out internal control and human DNA amplification). The resulting PCR products were barcoded using SQK-LSK109 and EXP-NBD196 kits (Oxford Nanopore Technologies) and sequenced on R9 flow cells. A custom script detected IS-pro primer sequences in the reads and built consensus sequences from sequences containing both forward and reverse primer sequences. These sequences were classified with BLAST searches at NCBI on the nr/nt and whole genome shotgun databases. Only hits with > 95% query coverage and identity were considered for inclusion in the IS-pro database.

Quantitative real-time PCR (qPCR).

Detection of Staphylococcus aureus by qPCR was performed as described previously (17).

Data analysis.

IS-pro results are reported as found by the software platform Antoni or by sequencing (Tables 1, 2, and 3). Comparison to culture is reported as positive percent agreement (PPA) and negative percent agreement (NPA) since culture is regarded as an imperfect reference standard for diagnosis of joint infections (18). Statistical analyses were performed using the Mann-Whitney U test. P < 0.05 was considered statistically significant.

TABLE 1.

Sample characteristics^a

Sample type	Analyzed samples	Culture positive	Is-pro positive	Concordant positive	Total positive	Culture polymicrobial	Is-pro polymicrobial
Native	293	48	59	43	64	1	3
Prosthetic	298	144	164	131	177	22	23
Total	591	192	223	174	241	23	26

^aData are the number of samples in the indicated categories at the top for the sample types indicated in the left outer column.

TABLE 2.

Detected species in native joint aspirates by culture and IS-pro^a

Species	Culture	IS-pro	Concordant
Staphylococcus aureus	28	33	28
Cutibacterium acnes	4	1	0
Streptococcus dysgalactiae	4	6	4
Staphylococcus epidermidis	3	4	3
Streptococcus pyogenes	3	4	3
Corynebacterium striatum	1	2	1
Escherichia coli	1	1	1
Mixed gram-positivesb	1	1	1
Staphylococcus hominis	1	0	0
Staphylococcus lugdunensis	1	1	1
Staphylococcus warneri	1	0	0
Bacillus licheniformis	0	1c1	0
Enterobacter cloacae /asburiae	0	1	0
Finegoldia magna	0	1	0
Granulicatella adiacens	0	1	0
Neisseria gonorrhoeae	0	1	0
Streptococcus bovis /salivarius	0	2	0
Turicibacter sp	0	1c1	0
Veillonella parvula	0	1	0
Total detections	48	62	42

^aData are the number of detections per assay and per species.

^bThe IS-pro result for this detection was Streptococcus mitis/pneumoniae group, which was considered concordant with the culture outcome.

^cn detections (out of the stated number) were obtained by sequencing.

TABLE 3.

Detected species in prosthetic joint aspirates by culture and IS-pro^a

Species	Culture	IS-pro	Concordant
Staphylococcus epidermidis	32	30	26
Staphylococcus aureus	31	40	28
Cutibacterium acnes	14	10	6
Enterococcus faecalis	10	8	8
Escherichia coli	10	16	9
Streptococcus dysgalactiae	10	14	10
Pseudomonas aeruginosa	6	5	5
Streptococcus agalactiae	6	6	6
Streptococcus mitis group	6	7	6
Enterobacter cloacae /asburiae	4	4	4
Morganella morganii	4	3	3
Bacteroides fragilis	3	3	3
Proteus mirabilis	3	3	3
Staphylococcus capitis	3	0	0
Aerococcus urinae	2	2	2
Aerococcus viridans	2	2	2
Citrobacter sedlakii	2	2	2
Corynebacterium striatum	2	0	0
Fusobacterium necrophorum	2	2	2
Klebsiella aerogenes /michiganensis/oxytoca	2	2	2
Peptoniphilus asaccharolyticus	2	0	0
Staphylococcus caprae	2	2	1
Streptococcus sanguinis	2	2	2
Anaerobic gram-negative rodsb	1	1	1
Bacillus species	1	0	0
Candida albicans	1	0	0
Citrobacter koseri	1	1	1
Clostridium perfringens	1	1	1
Coagulase-negative Staphylococcus (CNS)	1	0	0
Corynebacterium sp	1	0	0
Finegoldia magna	1	1c1	1
Gemella morbillorum	1	0	0
Klebsiella pneumoniae	1	2	1
Micrococcus luteus /lylea	1	0	0
Salmonella enteritidis	1	1	1
Staphylococcus hominis	1	0	0
Staphylococcus lugdunensis	1	0	0
Staphylococcus warneri	1	2	1
Streptococcus anginosus	1	1c1	1
Veillonella species	1	0	0
Anaerococcus sp	0	1c1	0
Anoxybacillus flavithermus	0	1	0
Bacillus cereus group	0	6c2	0
Bacteroides ovatus	0	1	0
Clostridium innocuum	0	1	0
Facklamia hominis	0	1c1	0
Fusobacterium sp	0	1	0
Kocuria sp	0	1c1	0
Lactococcus lactis subsp cremoris	0	4c4	0
Lysinibacillus sp	0	3c3	0
Massilia sp	0	2c2	0
Prevotella intermedia	0	1	0
Rothia mucilaginosa	0	1	0
Sneathia vaginalis	0	2c2	0
Staphylococcus auricularis	0	1	0
Streptococcus bovis /salivarius	0	1	0
Total detections	177	201	138

^aData are the number of detections per assay and per species.

^bThe IS-pro result for this detection was Prevotella sp, which was considered concordant with culture outcome.

^cn detections (out of the stated number) were obtained by sequencing.

RESULTS

Sample characteristics.

A total of 591 samples were analyzed: 293 native joint samples from 243 subjects and 298 prosthetic joint samples from 168 subjects. Sample characteristics are shown in Table 1. Prosthetic joint samples yielded significantly more positive cultures than native joint specimens (48% versus 16%, respectively). All but 1 of the 48 culture-positive native joint specimens grew monomicrobial cultures, most commonly S. aureus (58%). The prosthetic joint samples yielded a larger variety of microorganisms, and more often showed a mixed microbial culture (15% of positive samples). IS-pro detected more positive samples for each of the 2 sample types. Overall, 192 out of 591 (32%) samples were found positive by culture, whereas 223 (38%) were positive by IS-pro. A total of 174 samples (29%) were found positive with both methods. PPA between IS-pro and culture was 174/192 = 90.6% (95% CI 85.7 to 94%) and NPA was 350/399 = 87.7% (95% CI 84.1 to 90.6%) at the sample level.

Native joints: microorganisms detected by culture and IS-pro.

Of the 293 samples from native joints, 48 (16%) were positive with culture, whereas 59 of 293 (20%) were positive with IS-pro. All species detected by both methods are shown in Table 2. Six culture detections were found to be discrepant with IS-pro. Five of these were negative in IS-pro. Three involved a very low load (<1) of Cutibacterium acnes reported by culture, in the other 2 Staphylococcus hominis (load 1) and Staphylococcus warneri (load < 1) were found by culture. In the sixth discrepant detection, culture detected a <1 load of C. acnes, while IS-pro found a high signal for Staphylococcus aureus. Only 1 sample was scored as polymicrobial, with a culture result defined as mixed gram-positives. IS-pro identified a Streptococcus from the mitis/pneumoniae/sanguinis group which is a common species in such samples.

IS-pro yielded 20 additional bacterial detections in 17 samples compared with culture. The most common additional detected bacteria were S. aureus (5/20, 25%) and Streptococcus species (5/20, 25%). The remaining 10 additional detected bacteria consisted of various Gram-positive and Gram-negative bacteria (Table 2). Notably, 1 of these was a high Neisseria gonorrhoeae signal, a species difficult to culture (19). Of the 5 extra S. aureus detections, 1 came from a patient of whom another S. aureus culture-positive sample was collected in this study. The same was true for the extra Corynebacterium striatum detection. Seven of the extra detections came from samples that also contained high levels of leucocytes, indicative of a true infection.

Prosthetic joints: microorganisms detected by routine culture and IS-pro.

Of the 298 samples from prosthetic joints, 144 (48%) were positive with routine culture, and 164 (55%) were positive with IS-pro. A congruent positive detection for culture and IS-pro was found in 131 samples. Thirteen culture-positive samples were negative in IS-pro. Similar to the findings with native joint samples, these missed detections involved mostly low load detections of skin microbiota: C. acnes (6x), Staphylococcus capitis (2x), Corynebacterium (1x), Bacillus sp (1x), and Staphylococcus epidermidis (3x). IS-pro yielded an additional 33 positive samples, with a total of 14 different detected species, among which S. aureus (8 samples), Escherichia coli (4 samples), and Streptococcus dysgalactiae (3 samples). One-hundred and six samples out of the 131 samples that were positive with both methods yielded completely concordant results. All detections in the prosthetic joint samples are shown in Table 3. The results of all polymicrobial prosthetic joints samples are shown in Table S1 together with a subject identifier. Discrepancies were more prevalent in the polymicrobial samples: PPA at the species level between culture and IS-pro was 102/122 (83.6%, 95% CI 76.0 to 89.1%) for culture monomicrobial prosthetic joint samples, whereas this was 36/55 (65.5%, 95% CI 52.3 to 76.6%) for culture polymicrobial samples.

Evaluation of discrepant detections.

We looked for supporting clinical evidence for the IS-pro detections that were not found by culture, such as whether the species was present in culture in other samples from the same patient, or whether high leucocytes levels were reported. We found such evidence for 33 out of all 83 additional detections (Fig. 1A). Most of the IS-pro signals measured in the additional detections were higher than 10.000 RFU, indicating they were unlikely to be contaminants. We corroborated 13 out of the 17 additional S. aureus detections by qPCR (Fig. 1A). Only 4 samples with low S. aureus IS-pro signals were not positive by qPCR, possibly because of differences in sensitivity between the assays. We performed a similar analysis for detections that were found by culture, but not by IS-pro (Fig. 1B). Supporting clinical evidence was found for 18 out of the 44 detections. The M. morganii and E. faecalis (load 2) detections came from 1 subject with 4 samples in the study (subject 3260 in Table S1). Remarkably, IS-pro detected the same 2 species in 3 of the 4 samples, but in the fourth sample only a low load of S. epidermidis was detected. Possibly, this sample was mislabeled.

FIG 1.

Additional detections by IS-pro and culture with corroborating evidence for true positivity. (A) Signals of all additional IS-pro detections per species. (B) All additional detections by culture grouped per species and per culture load. (A) and (B) Colors indicate additional clinical evidence: in red: the species detected is present in culture in another sample from the same subject; in yellow: leukocyte levels fall into the category “many”; in blue: both of the categories red and yellow apply; in gray: none of the above apply. Crosses indicate samples that were positive by S. aureus qPCR. CNS: all non-S. aureus Staphylococci detections combined.

Bacterial loads.

Culture not only reports on pathogen identification but also on the abundance of microorganisms found. The fluorescence measured in IS-pro is a correlate for the amount of PCR products, which is directly related to the amount of DNA input in the PCR. We assessed whether the total amount of fluorescence associated with specific species peaks correlated to culture loads. To that end, we compared the IS-pro signals of all concordant positive detections to loads reported in culture which were given in categories (Fig. 2). This analysis showed that IS-pro signals are generally consistent with culture loads.

FIG 2.

Bacterial load comparison between culture and IS-pro. IS-pro signals are indicated as fluorescence (in relative fluorescence units [RFU]) of the PCR products of the detected species. Culture loads are given in categories as explained in the Methods section. P values calculated by the Mann-Whitney U test are indicated between the compared categories. Shown are all concordant detections.

Human DNA as a proxy for leukocyte count.

Additional to the detection of Proteobacteria, the Proteo-IC PCR in the IS-pro assay yields a few PCR products from human DNA. These products can be quantified in the same fashion as bacterial products and reflect the abundance of human cells in a sample. We assessed whether the human signals correlated with the reported leukocyte count. This relationship, shown in Fig. 3, highlights that the human DNA derived signals correlate well with leukocyte counts and may serve as a proxy for leukocyte count.

FIG 3.

Box plot showing human DNA signal intensities measured by IS-pro in RFU in relation to leukocyte levels reported by routine diagnostics. In the bottom table, the top row indicates the leukocyte levels as reported by routine diagnostics. The bottom rows indicate the total number of samples (count), the calculated median RFU and the number of outliers. Leukocyte levels were not reported for 83 samples. P values calculated by the Mann-Whitney U test are indicated between the compared categories.

DISCUSSION

In the present study, we evaluate the performance of IS-pro for diagnosis of BJI in a large set of 591 samples. IS-pro showed a PPA of 90.6% (95% CI 85.7 to 94%) and NPA of 87.7% (95% CI 84.1 to 90.6%) to culture at the sample level. The PPA at species level was better for monomicrobial infections (84%) than for polymicrobial samples (66%), with overall PPA of 80% (95% CI 74.3 to 84.7%). In cases where IS-pro missed the cultured species, culture often showed a low load of cutaneous species. Interestingly, IS-pro detected a number of pathogens with high signals (e.g., S. aureus, N. gonorrhoeae, and S. dysgalactiae) that were missed by culture (Fig. 1).

IS-pro yielded 83 detections that were not reported by culture. We found a number of uncommon species such as Sneathia and Lysinibacillus species. These were found in multiple samples from the same subject, arguing against them being contaminants. Both species have been recognized as human pathogens (20, 21). Overall, IS-pro outperformed culture in terms of breadth of species detection. We also detected species, often in high loads, that should readily grow in culture. These might be bacteria affected by antibiotic treatment or could represent DNA originating from biofilms. Regardless of the origin, the identification of bacterial DNA in an infected joint may guide antibiotic treatment.

There are some aspects that might further improve the IS-pro assay. In this study, 20 of the 264 (7.6%) IS-pro detections were based on sequencing which inevitably increases analysis time. Eleven of these detections involved minor species in polymicrobial samples, where our automated software did detect the most prevalent species. The majority of the sequenced detections involved uncommon species. As the IS-pro database continues to expand with uncommon bacteria, sequencing will become an increasingly dispensable part of the process. Furthermore, in the current study, the sample input volume was much higher in the culture procedure than in IS-pro (see Materials and Methods section), potentially explaining why low load positive samples were sometimes not detected by IS-pro. If concentrated joint fluid material (i.e., centrifuged as in the routine culture procedure) were used as input, IS-pro would likely do better on low load samples. Currently, IS-pro does not detect fungal pathogens such as Candida species which occurred once in the current sample collection in line with its being a rare BJI causative agent (22). Also, direct (genotypic) antibiotic resistance testing is not part of the IS-pro kit. Culturing of positive samples is therefore needed to assess phenotypic resistance. However, most samples are negative, and the species identified by IS-pro can guide antibiotic treatment according to the usual sensitivity profiles of the pathogens. A limitation of IS-pro is that some bacteria cannot be identified up to species level. For instance, IS-pro does not differentiate all members of the S. mitis or S. bovis groups, similar to other broad bacterial identification procedures such as MALDI-TOF or 16S rDNA sequencing (23).

The most discordant outcomes between culture and IS-pro were the C. acnes identifications. C. acnes occurs both as an infectious agent of BJI and as a common contaminant in BJI cultures, complicating the clinical interpretation of C. acnes detections. Furthermore, C. acnes is an often-encountered contaminant in molecular studies (24). Accordingly, we found that C. acnes signals required a higher background cut-off setting, based on negative control samples, than other species. The difficulty of interpreting a C. acnes finding in a suspected BJI is not new. It has been a longstanding issue in the diagnosis of BJI (25) for which guidelines for interpretation are in place (26). These guidelines require (among others) 2 or more positive samples, an elevated leukocyte count plus a load threshold to establish a confirmed infection. Of note, in our study, 4 of the concordant C. acnes samples showed high loads, both in routine and IS-pro, and came from the same patient. Thus, load may be a discriminative factor in assessing the clinical significance of C. acnes detections and may further assist physicians in interpreting results.

Recently, studies on the use of mNGS for BJI diagnosis have been published (8, 9, 27). The detection potential of mNGS is extensive as it enables identification of all infectious agents in a sample. However, it is still far from being implementable in routine clinical diagnostics. For instance, its interlaboratory variation was found to be unacceptably high (28). At present, the average turnaround time for most mNGS platforms from sample receipt to result is still 48 h (29) although Nanopore technology should yield faster turn-around times (30). In an independent study, IS-pro was shown to be faster and cheaper than other broad approaches (31).

We showed a significant correlation between IS-pro generated human signals and leukocyte counts reported by histology. The IS-pro signals are not necessarily an improvement over histology but may be relevant when only IS-pro is performed.

In conclusion, IS-pro showed excellent diagnostic performance for BJI. Although other broad molecular sequencing approaches may yield similar performance for positive samples, IS-pro is unique in that it is easily manageable in a routine diagnostic laboratory.