Conceived and designed the experiments: RM KEM MB YP PKW JCL. Performed the experiments: RM LD. Analyzed the data: RM KEM. Contributed reagents/materials/analysis tools: KEM MB LD. Wrote the paper: RM KEM PKW JCL.
The authors have declared that no competing interests exist.
Urinary tract infection (UTI) is a common infection that poses a substantial healthcare burden, yet its definitive diagnosis can be challenging. There is a need for a rapid, sensitive and reliable analytical method that could allow early detection of UTI and reduce unnecessary antibiotics. Pathogen identification along with quantitative detection of lactoferrin, a measure of pyuria, may provide useful information towards the overall diagnosis of UTI. Here, we report an integrated biosensor platform capable of simultaneous pathogen identification and detection of urinary biomarker that could aid the effectiveness of the treatment and clinical management.
The integrated pathogen 16S rRNA and host lactoferrin detection using the biosensor array was performed on 113 clinical urine samples collected from patients at risk for complicated UTI. For pathogen detection, the biosensor used sandwich hybridization of capture and detector oligonucleotides to the target analyte, bacterial 16S rRNA. For detection of the protein biomarker, the biosensor used an analogous electrochemical sandwich assay based on capture and detector antibodies. For this assay, a set of oligonucleotide probes optimized for hybridization at 37°C to facilitate integration with the immunoassay was developed. This probe set targeted common uropathogens including
We successfully demonstrate simultaneous detection of nucleic acid and host immune marker on a single biosensor array in clinical samples. This platform can be used for multiplexed detection of nucleic acid and protein as the next generation of urinary tract infection diagnostics.
Urinary tract infection (UTI) is a common bacterial infection that affects all patient demographics. Diagnostic criteria include presence of urinary symptoms (e.g. frequency, urgency, dysuria), urinalysis showing pyuria, and urine culture showing ≥105 cfu/ml uropathogen. For culture and the associated antimicrobial susceptibility testing (AST), urine samples are sent to a clinical microbiological laboratory, which has a typical delay of 2–3 days. Due to this delay, physicians often prescribe antibiotics empirically based on symptoms and historic antimicrobial susceptibility data. While empiric treatment is sufficient in many patients, a more complete diagnosis is beneficial for patients with recurrent, complicated UTI such as those dependent on urinary catheters for bladder emptying. Catheterized patients are prone to bacterial colonization in the bladder that may not necessitate treatment (i.e. asymptomatic bacteriuria) but are also at a greater risk of infection with resistant pathogens due to frequent exposure to antibiotics
As a part of the host innate immune response, white blood cells (WBCs) are recruited to the urinary tract in response to the presence of bacterial pathogens. Typically, urinary WBC counts are determined by urinalysis in a centralized laboratory or approximated by a dipstick test at the point of care. Although identification of pathogens gives useful information for diagnosis of UTI, it does not distinguish colonization from infection, determine severity of infection, or the degree of host response. Quantitative detection of urinary proteins indicative of host immune response, in addition to pathogen identification, provides a more comprehensive diagnosis of UTI and a significant advancement towards a personalized medicine for UTI treatment
There is significant interest to develop biosensor technology for applications in healthcare, environmental, and food safety monitoring
Using the electrochemical biosensor, we demonstrated a 1-hour biosensor assay for detection of pathogen 16S rRNA from patient urine samples using the biosensor
(A) Schematic of pathogen identification based on sandwich hybridization of bacterial 16S rRNA with capture and detector oligonucleotide probes; (B) Schematic of immunoassay based on sandwich detection host urinary protein with capture and detector antibodies. The two assays share similar assay parameters, including surface functionalization with biotinylated capture probes/antibodies, probe-target binding at 37°C, and amperometric detection using horseradish peroxidase (HRP) as the signaling enzyme
In our previous work, the biosensor assays for nucleic acids and protein detection were based on target binding of nucleic acids at 65°C
Probe selectivity and sensitivity was tested against different uropathogens. A. The biosensor signal for detection of
A systematic approach of combining
After the
Log signal intensities of current (nA) are plotted on the Y-axis. Common uropathogens tested are indicated on the X-axis. The line over the bars indicates the threshold for positive biosensor signal 3SD (log10 unit) over the negative control for UNI, EB, PM, EC, PA, EF, SM, PS, SA and 5SD (log10 unit) over negative control for MM and KE probes. Consistent with their
Probe pairs (Length in bp) | Sequence (5′-3′) | Species detected |
UNI798C (22) |
|
Ec, Mm, Kp, Pm, El, Ea, Cf, Cb, Ko, Ps, Sm, Sf, Pr |
UNI776D (22) |
|
Pa, Ef, Ee Sa, Se, Ss, Pv, Kz, Xm, Dp, Ab, Af, Fi |
EB1275C (23) |
|
Ec, Mm, Kp, Kz, Pm, El, Ea, Cf, Ko, Ps, Sm, Pr, Sf, |
EB1252D (23) |
|
Cb |
PM1019C (22) |
|
Pm, Pv |
PM997D (22) |
|
|
EC471C (24) |
|
Ec |
EC447D (24) |
|
|
PA594C (23) |
|
Pa |
PA570D (23) |
|
|
EF220C (20) |
|
Ef, Ee |
EF200D (20) |
|
|
SM472C (22) |
|
Sm |
SM450D (22) |
|
|
PS151C (24) |
|
Ps, Pr |
PS127D (24) |
|
|
SA91C (22) |
|
Sa, Se, Ss |
SA69D (22) |
|
|
MM181C (22) |
|
Mm |
MM147D (22) |
|
|
KE468C (22) |
|
Ea, El, Kp, Cf, Ko, Kz, Cb |
KE446D (22) |
|
The capture (denoted by “C”) and detector (denoted by “D”) probes were modified with 5′ biotin and 3′ fluorescein, respectively. The degenerate bases R represents bases A or G and Y represents C or T.
The limit of detection (LOD) of the biosensor assay was 104 cfu/ml from bacterial culture and clinical urine samples. Similar to our previous results
In order to validate the new probe pairs in an integrated biosensor assay combining pathogenic nucleic acids and host protein biomarker detection, we used clinical urine samples collected from spinal cord injury (SCI) patients, who are at significant risks of developing complicated, polymicrobial urinary tract infections due to neurogenic bladder and frequent use of catheters. From 111 patients recruited (109 male, mean age 59±13), 113 samples were analyzed, 85 from outpatients and 28 from inpatients. In our study population, 27 samples were collected through spontaneous voiding, 85 samples from catheters, and 1 sample from a patient with ileovesicostomy. Of the 85 catheterized samples, 21 were from standard indwelling Foley catheter, 21 from clean intermittent catheterization, 11 from suprapubic (SP) catheter, and 32 from condom (C) catheter.
Number of samples | |
|
|
|
|
|
9 |
|
4 |
|
2 |
|
2 |
|
2 |
|
1 |
|
1 |
|
1 |
|
|
1 | |
1 | |
2 | |
1 | |
2 | |
1 | |
1 | |
1 | |
1 | |
1 | |
1 | |
|
|
The collected urine samples were tested using the integrated biosensor array for quantitative detection of pathogen(s) and LTF. Each urine sample was tested with a single 16-sensor array. The sensors were functionalized with the new panel of the capture probes against 16S rRNA and the capture antibody against LTF. Both hybridization and the antibody-antigen binding steps were performed at 37°C.
Each array consisted of 16 electrochemical biosensors. In this study, eleven sensors were used for nucleic acid assay for pathogen identification and 5 sensors for immunoassay for LTF detection. For measurement of LTF directly from urine, two urine dilutions were tested and the concentration of LTF was determined using the standard curve generated from three known concentrations of LTF. For this urine sample, positive signals from the UNI, EB and EC probes indicated this sample contained
Of the 34 samples that were culture negative, biosensor results agreed with clinical microbiology in all but one sample, yielding a specificity of 97%. In this particular sample, the clinical microbiology laboratory reported it to be culture negative, while independent culturing in our laboratory indicated 4×107 cfu/ml bacteria. The clinical microbiology laboratory reported 79 samples as culture positive, out of which biosensor detected 70 as positive, yielding a sensitivity of 89%. The samples that were not detected by the biosensor had bacterial concentrations ranging from 103 to 104 cfu/ml, which was at or below the LOD of the biosensor assay.
The biosensor assay detected bacteria in all 22 urine samples containing a single uropathogen and further identified the bacterial species in 16 of these samples and the family (e.g.
The majority of the samples with polymicrobial infection were reported as ‘mixed flora’ and not speciated by the clinical microbiology laboratory, thus precluding direct comparison with the biosensor results at the species level. Out of the 44 ‘mixed’ samples identified by the clinical microbiology laboratory, the biosensor detected 37 of these samples as positive by the UNI probe. The bacterial concentration in four of the missed samples was below the biosensor's limit of detection.
Of the 13 polymicrobial samples that were speciated, the biosensor accurately identified all species in four of these samples and at least one of the species in six samples. The samples missed by the biosensor either had pathogens below the limit of the detection of the biosensor or specific probes for the pathogens were not available.
LTF concentration was determined in each urine sample along with the detection of pathogens by 16S rRNA in the integrated biosensor assay. Data from the integrated biosensor assay was correlated with results obtained from urinalysis. Spearman's correlation coefficient was used to evaluate relationship between different clinical parameters, bacterial and LTF concentration. Consistent with our previous report
System integration is one of the most challenging aspects of developing point-of-care diagnostics
For this study, we developed new panel of oligonucleotide probes for pathogen identification that facilitate integration with synchronous protein detection under common assay conditions. While nucleic acid detection works well at higher temperature
We have expanded our probe panels to target additional uropathogens including
We successfully demonstrated simultaneous detection of nucleic acid and host immune marker on a single biosensor array in clinical samples. The two electrochemical assays share similar protocol parameters, including surface functionalization with biotinylated capture probes/antibodies, probe-target binding at 37°C, and amperometric detection using horseradish peroxidase (HRP) as the signaling enzyme. Furthermore, integration improves the efficiency and reduces the potential human errors associated with running two separate assays. One sample reported to be culture negative, was found to have significant pyuria (75–100 WBC/HPF) and bacteria by microscopy in separate clinical laboratory urinalysis. In our integrated assay, we found this sample to be biosensor positive for UNI, EB and KE probes and contained 9950 ng/ml of LTF. Independent plating in our laboratory further corroborated pathogen identification, indicating the presence of 4×107 cfu/ml bacteria.
LTF is a molecular marker of pyuria
Rapid molecular diagnosis of UTI represents a significant advance in the management of UTI and potentially reduce the practice of prescribing unnecessary antibiotics
Patient urine samples were collected with approval from Stanford University's Institutional Review Board (IRB) and Veterans Affairs Palo Alto Health Care System's (VAPAHCS) Research and Development committee. Oral informed consents were obtained from all study subjects due to spinal cord related physical impairment and low risk nature of the study. Oral consent was approved by both Stanford IRB and VAPAHCS R&D committee and was documented with a written record of participants name, date of birth, medical record number and date of sample collection, which served as both the record of oral consent and enrollment record.
Oligonucleotide probes targeting the 16S rRNA of uropathogens were designed using Primrose v2.17 and RDP Release 8. The selectivity of the potential probes identified by Primrose was assessed using the probe match function of RDP Release 10. Probe selectivity was further verified by alignment against 16S rDNA sequences from bacterial isolates from urine (at least 3 sequences for each species) using Geneious. 16S rRNA gene sequences from bacterial isolates were kindly provided by David Haake (unpublished data).
Oligonucleotide probes are designated as capital letters to detect all Eubacteria (UNI),
To validate their selectivity, oligonucleotide probes were tested against uropathogen isolates from patient urine samples obtained at VAPAHCS as well as strains obtained from American Type Culture Collection (ATCC):
The electrochemical biosensor array (GeneFluidics, USA) surface was functionalized as described previously
Urine samples were collected from the VAPAHCS spinal cord injury unit between July 2009 and April 2010. The decision to collect urine was based on clinical judgment of the treating physician (patients suspected of a UTI or as part of their routine care). Depending on the patient's bladder emptying status, urine samples were collected from an indwelling catheter, straight catheterization, or voiding. For each sample, one aliquot was used for the biosensor experiment and the second was sent to clinical laboratory for urinalysis (white blood cells per high power field (WBC/HPF), pH, specific gravity, leukocyte esterase, nitrite) and culture and susceptibility. Qualitative and quantitative plating was done in our laboratory on BBL CHROMagar™ Orientation and TSA with 5% sheep blood (BD Diagnostics, Sparks, MD) and LB media, respectively. For biosensor pathogen identification assay, 1.5 ml of the sample was centrifuged and the pelleted fraction was stored. Whole urine was stored for the LTF biosensor immunoassay. The samples were stored at −80°C until tested.
For the integrated assay, the surface of 11 of the 16 sensors on the array was functionalized with the capture probes for pathogen identification and 5 sensors were functionalized with capture antibodies for LTF detection (rabbit biotinylated polyclonal anti-LTF, ab25811, Abcam). The biosensor with the capture probes and antibody was incubated at 37°C for 30 min. The urine pellet was lysed as previously described
Clinical microbiology results were used as the standard to determine the diagnostic sensitivity and specificity of the biosensor assay for pathogen identification. Per VAPAHCS clinical microbiology laboratory protocol, samples containing bacterial concentrations of 10,000 cfu/ml or less were not speciated. In addition, polymicrobial samples containing 3 or more species were typically not speciated and reported as mixed flora. Similar to our previous report, signal from UNI probe was used to calculate the diagnostic sensitivity of the biosensor
The urinary LTF concentrations were compared with results obtained from urinalysis (WBC/HPF and leukocyte esterase) to assess validity of the biosensor. To facilitate analysis, WBC/HPF obtained from urinalysis were divided into four groups (0–2, 3–10, 11–50, >50). Samples containing <2 WBC/HPF were considered negative for pyuria. Descriptive analyses and ranked medians among categorical groups were conducted using Kruskal-Wallis test with GraphPad Prism version 5.0b. Spearman correlations were performed using SAS version 9.1.3 to assess ranked correlation between clinical and biological variables.
We thank the nursing staff of spinal cord injury unit at VAPAHCS, Katherine Wu, and Shelly Teng Teng Hsiao for technical assistance. We thank Vincent Gau for helpful discussions and Xuefeng B. Ling and Gigi Liu for assistance with biostatistical analysis.