Tracheal colonization factor A (TcfA) is a biomarker for rapid and specific detection of Bordetella pertussis

Pertussis is a highly contagious disease for which prompt, point-of-care diagnosis remains an unmet clinical need. Results from conventional test modalities (nucleic acid detection, serology, and culture) take hours to days. To overcome this challenge, we identified a new biomarker (tracheal colonization factor A, TcfA) for detection of Bordetella pertussis infection by lateral flow immunoassay (LFIA). We developed a library of 28 epitope-mapped monoclonal antibodies against TcfA and incorporated three antibodies into a LFIA. The LFIA did not cross-react with common bacterial or fungal organisms, but did react with nine distinct B. pertussis strains. The minimal linear epitope sequences targeted by the LFIA were conserved in 98% of 954 B. pertussis isolates collected across 12 countries from 1949–2017. The LFIA’s limit of detection was 3.0 × 105 CFU/mL with B. pertussis cells in buffer, 6.2 × 105 CFU/mL with nasopharyngeal washes from a non-human primate model, and 2.3 ng/mL with recombinant TcfA. The LFIA reacted with patient nasopharyngeal swab specimens containing as few as 1.8 × 106 B. pertussis genomes/mL and showed no false-positives. Rapid (< 20 min) LFIA detection of TcfA as a biomarker for B. pertussis infection is feasible and may facilitate early detection of pertussis.


Results
Biomarker selection and epitope validation with polyclonal antibodies. Candidate B. pertussis biomarkers were selected based on their high abundance 16,17 , known immunogenicity during human infection 9,18 or animal immunization 17,[19][20][21] , and secreted or surface localization 17,20,[22][23][24] . Candidates that were known virulence determinants or cytotoxic factors 9 were prioritized to increase the probability of a diagnostic's long-term utility due to selective pressure against loss of expression. Candidates were also required to have multiple, spatially separated regions with high predicted antigenicity, high conservation across B. pertussis strains, and low homology to other microorganisms reported in nasopharyngeal specimens [25][26][27][28][29][30][31] . Ultimately, we selected five proteins that met the above criteria: tracheal colonization factor A (TcfA), adenylate cyclase toxin (ACT), filamentous hemagglutinin (FHA), pertussis toxin subunit 1 (PTXS1), and autotransporter (Vag8). For each protein, we designed two to five peptide immunogens and used them to generate rabbit pAbs. The resulting 20 affinity-purified rabbit pAbs were evaluated for their performance in ELISA (summarized in Table 1). The TcfA pAbs produced the most sensitive antigen-capture ELISA for formaldehyde-inactivated B. pertussis cells. Thus, we focused our MAb development efforts on TcfA as a biomarker and on the three TcfA peptides that yielded the most highly reactive TcfA pAbs as target epitopes.
To finely map the minimal linear epitopes of the MAbs, a tiled library of biotinylated TcfA peptides was probed by ELISA. Minimal linear epitopes were defined as the minimal overlapping peptide sequence for wells with an OD 450 greater than 0.75 (and in a series of two or more such adjacent wells). The 28 MAbs belonged to 14 minimal linear epitope groups ( Fig. 1 and diagramed in Supplementary Fig. S2).
MAb performance in LfiA. To identify functional MAb pairings for LFIA, we screened each MAb for performance as a test line MAb and as a gold conjugate MAb (Fig. 2). Of 784 MAb pair permutations evaluated by LFIA, we identified 101 functional permutations that showed a twofold or greater increase in signal with formaldehyde-inactivated B. pertussis cells in PBS vs. PBS alone and that showed little to no absolute background (less than 100 units of test line signal with PBS alone) (Fig. 2).
The most sensitive permutations were advanced for additional testing with (i) formaldehyde-inactivated B. pertussis cells lysed in SDS, (ii) unfixed 0.2 µm-filtered supernatant of B. pertussis liquid cultures, (iii) rTcfA-His, and (iv) viable B. pertussis cells in PBS (data not shown). LFIAs containing cocktails of three MAbs (two as gold conjugates and one at the test line or vice versa) were also evaluated for a subset of the highest performing MAbs (data not shown). Ultimately, the LFIA containing MAb 10B1 as gold conjugate and MAbs 13E11 and 14D12 at Table 1. Reactivity of purified pAbs raised against predicted epitopes of TcfA, ACT, FHA, PTXS1, and Vag8. All pAbs were affinity purified using each cognate peptide in the solid phase. a pAb reactivity determined by ELISA using (i) B. pertussis cells, (ii) cell lysate, or (iii) purified protein in the solid phase as the detected antigen. b pAb function in antigen-capture ELISA determined by assessing all possible pairwise permutations of epitope-specific pAbs in the capture or detector modes. c LOD was defined as the concentration of formaldehyde-inactivated B. pertussis cells that generated signal equal to 3 × background. The average LOD was calculated by assessing the best pAb pair on three days, each day in duplicate. Directly measured OD 600 values were converted to CFU/mL based on OD 600 measurements and quantitative cultures performed with live B. pertussis. www.nature.com/scientificreports/ the test line was chosen for further study based on its consistently strong reactivity with a diversity of B. pertussis antigen preparations and lack of detectable background (Fig. 3).
LFIA analytical specificity and inclusivity. The chosen LFIA was evaluated for analytical specificity with titered vials of viable bacteria and fungi. Samples were tested in triplicate at 3.3 × 10 7 CFU/mL; B. pertussis (Tohama I) was included as a positive control. The LFIA showed no cross-reactivity with any of the 40 species tested, including two strains each of B. parapertussis and B. bronchiseptica, and one strain of B. holmesii (Supplementary Table S1). Inclusivity of the LFIA was evaluated with formaldehyde-inactivated B. pertussis cells at an OD 600 of 0.1 (equivalent to 4.8 × 10 7 CFU/mL) across 10 LFIA replicates per strain. All strains isolated from 1963 onwards reacted with the LFIA (n = 8, Tohama I, 165, H735, H792, E431, CNCTC Hp 12/63, H973, and A639) (Supplementary Fig. S3). The two strains that did not react were 18323 and 5374(3747), which were isolated prior to 1947 and 1957, respectively 32,33 . Strain 18323 carries the tcfA1 allele, which contains an extra 25 amino acids relative to the tcfA2 allele carried by the Tohama I strain 32,34 . Alignment of the two alleles along with the minimal linear epitopes for the MAbs in the LFIA are diagramed in Supplementary Fig. S4.
To further evaluate how inclusive the LFIA would be with recent clinical isolates, we analyzed the published genomic sequences of 170 B. pertussis isolates selected by the Centers for Disease Control as representative of B. pertussis diversity in the U.S. from 2000 to 2013 35 . All 170 isolates contained sequences encoding the same protein as the tcfA2 allele, which is reactive with the pertussis LFIA. Broadening our geographic focus, we reviewed 784 other B. pertussis isolates for which TcfA sequence information was available 34,[36][37][38][39][40] . Only 19 of these isolates were from the U.S. 34,39 . The remaining 765 were from Poland, the Netherlands, China, the United Kingdom, Finland, Sweden, France, Germany, Japan, Italy, and Belgium (listed in descending order based on the number of isolates sourced from each country) 34,[36][37][38][39][40] . These isolates were collected between 1949 and 2017 34,[36][37][38][39][40] . Of these 784 isolates, 90% (n = 706 isolates) harbored the tcfA2 allele. In addition, 59 of the 784 isolates carried the tcfA3 or tcfA4 alleles, which contain a single amino acid substitution ≥ 23 amino acids away from the nearest minimal linear epitope sequence of any of the three MAbs in the LFIA (i.e. position 176 for tcfA3 and 174 for tcfA4) 34,36 ( Fig. 1 and Supplementary Fig. S2). As a consequence, 765 of the 784 isolates (98%) are predicted to be reactive with the three MAbs in the LFIA. The remaining 19 isolates comprise tcfA deletions (n = 6), frameshifts (tcfA5, n = 7; tcfA9, n = 1) or alternative alleles (tcfA1, n = 3; tcfA6, n = 2) 34,36-39 .   www.nature.com/scientificreports/ LFIA performance was assessed with NP wash specimens from baboons directly challenged with B. pertussis (D420) via a well-established non-human primate model 1,41 . Building on the inclusivity testing of Supplementary  Fig. S3, D420 was the ninth strain of eleven tested strains that reacted with the LFIA. A photograph of representative LFIAs run with NP washes having a range of bacterial burdens as well as the visual interpretation of those LFIAs by three blinded readers are shown (Fig. 5A). Each NP wash (n = 41) was tested on duplicate LFIAs, with the exception of 4 specimens for which low sample volume permitted only a single replicate. The LFIA produced no false-positives with any of the 11 NP washes containing 0 CFU/mL (100% specificity) (Fig. 5B). All NP washes with ≥ 4.9 × 10 5 CFU/mL were positive, and all NP washes with ≤ 3.4 × 10 4 CFU/mL were negative (Fig. 5B). The lowest concentration scored as positive was 5.4 × 10 4 CFU/mL (3.2 × 10 3 CFU/test) (Fig. 5B). The LOD with baboon NP washes (n = 78 tests) was 6.2 × 10 5 CFU/mL (95% CI, 4.9 × 10 5 to 1.1 × 10 6 CFU/mL) ( Fig. 5 and Supplementary Fig. S6).
LfiA assessment with patient np swab specimens. NP swab specimens from nine human patients with suspected pertussis were evaluated by LFIA with three blinded readers ( Table 2, Supplementary Fig. S7). Four specimens that were negative by qPCR were also negative by LFIA (Table 2). Similarly, two qPCR-positive specimens with Ct values corresponding to ≥ 1.8 × 10 6 genomes/mL were also positive by LFIA (Table 2). Two specimens from qPCR-positive patients had Ct values equivalent to 3.6 × 10 5 and 8.1 × 10 4 genomes/mL, resulting in split visual interpretations of the LFIA by the blinded readers: two of three readers interpreted the specimen with 3.6 × 10 5 genomes/mL as positive, and one of three readers interpreted the specimen with 8.1 × 10 4 genomes/mL as positive ( Table 2). The concentration range covered by these values is within the concentra-  Table 2. Nasopharyngeal swab specimens from nine patients with suspected pertussis were evaluated by qPCR and LFIA. a Observed qPCR Ct values for sample loaded onto LFIA were converted to genomes/mL based on a standard curve with purified B. pertussis genomic DNA. b Ct values < 35 were diagnosed as positive; qPCR was performed as described 52 . c Consensus visual interpretation by three blinded readers. Where all three readers were not in agreement, independent reader scores are listed (d and c). d Two readers scored the LFIA as positive; one reader scored the LFIA as negative. e One reader scored the LFIA positive; two readers scored the LFIA as negative. f No Ct value because target was not detected. g Not applicable because there was no Ct value. www.nature.com/scientificreports/ tion range that also produced both positive and negative LFIAs in the baboon NP wash testing (i.e. 3.4 × 10 4 to 4.9 × 10 5 CFU/mL, Fig. 5).

Discussion
There is an unmet clinical need for rapid, point-of-care diagnosis of pertussis. This study identified TcfA as a biomarker that could be used to detect B. pertussis in < 20 min by LFIA. Because B. pertussis cells are present in NP specimens, we pursued antibodies that recognized formaldehyde-inactivated B. pertussis cells (Table 1). Moreover, TcfA has a secreted and a cell surface associated isoform 24 . Thus, to increase the potential clinical sensitivity of the LFIA, we targeted regions conserved in both isoforms (i.e. amino acids 40-374) 24 for antibody development. All 28 TcfA MAbs bound formaldehyde-inactivated B. pertussis cells, and all TcfA MAbs bound epitopes conserved in both TcfA isoforms ( Fig. 1 and Supplementary Fig. S1 and S2). LFIA testing demonstrated that some MAbs formed functional LFIA pairs only as test line MAbs (e.g. MAbs 15A9, 14D9, and 19F9) or only as gold conjugate MAbs (e.g. MAbs 7A10 and 22B7) (Fig. 2). Some MAbs worked equally well in both positions (e.g. MAbs 13E11 and 18B2) (Fig. 2). The least amount of background and strongest signal was observed with the LFIA containing MAb 10B1 as gold conjugate and a combination of MAbs 13E11 and 14D12 at the test line (Figs. 2 and 3).
This LFIA configuration was highly specific for B. pertussis when evaluated with 40 other microorganisms potentially found in the nasopharynx, including B. parapertussis, B. holmesii, and B. bronchiseptica (Supplementary Table S1). Infections with these Bordetella species are rarer than infections with B. pertussis, and optimal treatment regimens for them have yet to be established 9,42,43 . Where clinical guidelines do exist, they generally call for different actions to be triggered for a diagnosis of B. pertussis than for other Bordetella species. Thus, the lack of cross-reactivity of the pertussis LFIA with B. bronchiseptica, B. parapertussis, and B. holmesii (Supplementary Table S1) may be advantageous for guiding appropriate actions regarding patient treatment, post-exposure prophylaxis, and public health notifications.
Inclusivity testing indicated that the LFIA was reactive with multiple strains of B. pertussis. Specifically, the LFIA reacted with 9 of 11 B. pertussis strains and two untyped isolates from patient NP specimens (Supplementary Fig. S3, Fig. 5, and Table 2). One of the two strains that did not react with the LFIA was 18323, which was isolated over 70 years ago 32 . The 25 additional amino acids in the tcfA1 allele of 18323 34 (relative to the tcfA2 allele found in Tohama I) may change the local secondary structure of TcfA near the epitope of MAb 10B1 (Supplementary Fig. S4), which may explain the LFIA's lack of reactivity with this strain. The other strain that did not react with the LFIA was 5374(3747), which was isolated over 60 years ago 33 . To our knowledge, tcfA sequence data for 5374(3747) has not been published. Notably, this strain is known to differ from more recent clinical isolates due to its lack of expression of the type III secretion system (TTSS) effector, Bsp22 44 . Like TcfA, the TTSS is a virulence factor that may be important for efficient respiratory tract colonization 24,44 .
Both high inclusivity and high analytical sensitivity are necessary for high clinical sensitivity. Two reports describing the bacterial burden in NP washes from four infants, all aged less than 3 months, have been published 16,25 . These specimens contained 8 × 10 10 genomes/mL, 1.2 × 10 9 genomes/mL, 1 × 10 8 CFU/mL, and 7.5 × 10 7 CFU/mL 16,25 . These bacterial burdens are 100-fold to 125,000-fold greater than the LOD of the pertussis LFIA with baboon NP washes, and 250-fold to 250,000-fold greater than the LFIA's LOD with B. pertussis cells in PBS, which further suggests that using TcfA as a biomarker for LFIA-based detection of infant pertussis may be feasible.
Specimen type will likely be an important variable in assay performance. Although both NP swabs and NP washes are compatible specimen types for the pertussis LFIA (Table 2 and Fig. 5), NP washes would likely be the preferred sample specimen for clinical use due to the higher concentration of bacteria reported in infant NP washes vs. NP swab extracts 16,25,[45][46][47] . In a direct comparison of specimen collection methods, washes were found to be a much more sensitive sampling method than swabs for PCR-mediated detection of B. pertussis: in cases where a patient had both a nasal wash and a pernasal swab collected at the same time and the wash was positive, the corresponding swab was positive only approximately 50% of the time 48 .
Patient age may also impact assay performance. Younger patients have generally been found to have higher bacterial burdens at the time of specimen collection than older patients 45,46 . This may limit the pertussis LFIA's clinical utility to younger patients (e.g. infants). However, an infant-use-only indication might not materially reduce the public health impact of the assay, since (i) infants are at greater risk of hospitalization, morbidity, and mortality from pertussis than older patients, and (ii) infants require more healthcare visits on average to receive a pertussis diagnosis than do older patients 14,15 .
Our examination of human samples provided strong proof-of-concept that TcfA is a biomarker for infection, but likely underestimated the potential sensitivity of the LFIA for detection of infection. First, the remnant patient specimens we analyzed were from swabs, not washes. Second, the swabs had been placed into 1 mL of Amies medium instead of a minimal volume (e.g. 0.1 mL). Third, a further two-fold dilution of the specimens into SDS/ PBS extraction buffer was required to mitigate interference of the Amies medium with the LFIA. In the future, a prospective clinical trial with NP washes from symptomatic patients will be needed to rigorously determine the assay's clinical sensitivity and specificity as a function of patient age and disease stage.
In conclusion, we have identified a novel biomarker (TcfA) for detecting B. pertussis infection by LFIA. The LFIA's LOD with NP washes from a non-human primate model and reactivity with NP swab specimens from www.nature.com/scientificreports/ human patients raises the possibility that a proven diagnostic approach for many other infectious diseases (i.e. LFIA) can be brought to bear on pertussis. A pertussis LFIA would differ from existing pertussis tests in that it would (i) require no electricity or specialized equipment, (ii) provide results in less than 20 min, and (iii) eliminate the need to send specimens to the laboratory. A rapid test that could be used at the point-of-care even in resource limited settings would enable immediate patient triage and prompt initiation of appropriate antibiotic therapy, which in turn should help reduce patient morbidity, limit further transmission, and save infant lives. Antibody production. Immunization of rabbits for polyclonal antibody (pAb) production and of mice for MAb production was approved by the University of Nevada, Reno Institutional Animal Care and Use Committee (IACUC); all experiments were performed in accordance with relevant guidelines and regulations. ProSci, Inc. (Poway, CA) prepared pAbs by immunizing rabbits with peptide immunogens coupled to keyhole limpet hemocyanin (KLH). The resulting pAbs were affinity-purified over peptide-resin columns.

Methods
MAbs were prepared in-house by immunizing mice with (i) TcfA peptides conjugated to KLH, (ii) formaldehyde-inactivated B. pertussis cells (Tohama I) in PBS, or (iii) rTcfA, peptide:KLH conjugates, and formaldehydeinactivated B. pertussis cells. Splenocytes were isolated and cryopreserved as described 51 . Hybridomas were generated from cryopreserved splenocytes via standard protocols and subjected to multiple rounds of cloning by limiting dilution. MAbs were purified from cell culture supernatant using rProtein A (GE). eLiSA. Bovine serum albumin (BSA) and TcfA peptides conjugated to BSA were diluted in 50 mM carbonate, pH 9.6. All other antigens were diluted in PBS, pH 7.4. For ELISAs with MAbs, microtiter plates were incubated with antigens overnight, washed with PBS-Tween (PBS containing 0.05% Tween-20), and blocked for 90 min with blocking buffer (PBS containing 0.5% Tween-20 and 5% w/v non-fat powdered milk) at room temperature (RT). Plates were washed, incubated for 90 min at RT with antibodies in blocking buffer, washed again, then incubated 90 min at RT with horseradish peroxidase-labeled (HRP-labeled) goat anti-mouse secondary antibodies (Southern Biotech, Birmingham, AL and Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) in blocking buffer. After a final wash, plates were incubated for 30 min at RT with tetramethylbenzidine substrate (SeraCare, Milford, MA). The reaction was stopped with 1 M H 3 PO 4 , and plates were read at OD 450 .
Indirect and direct ELISAs with rabbit pAbs were performed similarly except for the following changes: blocking buffer was PBS containing 0.5% Tween-20 and 1% (v/v) normal rabbit serum (MP Biomedicals), incubations were done for 1 h at 37ºC, washes after the blocking step and after the primary antibody incubation were done with blocking buffer, and rat anti-rabbit IgG-HRP (Southern Biotech) was used. Secondary antibody steps were omitted when using HRP-conjugated rabbit pAbs.
For antigen-capture ELISAs with rabbit pAbs, plates were coated overnight with affinity purified pAbs, washed with PBS-Tween, blocked with 0.5% Tween-20 and 1% (v/v) normal rabbit serum for 90 min at RT, then incubated with antigen in blocking buffer for 90 min at RT. After washing in PBS-Tween, plates were incubated with substrate, stopped, and read as above. Data was analyzed with SoftMax Pro (version 6.4), using a log-log plot.
MAb epitope mapping. A tiled peptide library covering amino acids 40-374 of TcfA was constructed (Mimotopes, Melbourne, Australia). The library contained 108 peptides (15mers, offset by three amino acids). Peptides were biotinylated with an "SGSG" spacer at the N-terminus and an amide at the C-terminus. Standard binding streptavidin-coated microtiter plates pre-blocked with bovine serum albumin (BSA) (ThermoScientific Pierce) were incubated for 1 h with shaking with peptides at 0.1 µg/mL in PBS containing 0.1% (w/v) BSA. Plates were washed with PBS containing 0.05% Tween-20 (PBS-Tween), incubated for 1 h with shaking with purified MAbs at 1 µg/mL in PBS containing 2% (w/v) BSA, washed again, and incubated for 1 h with shaking with goat was sprayed with goat anti-mouse Ig, Human adsorbed (Southern Biotech) at 1 mg/mL in PBS using a BioDot XYZ3050 Biojet system at 1 µL/cm. LFIAs were assembled onto 60 mm tall adhesive backing cards (DCN Diagnostics, Carlsbad, CA) with a CFSP203000 (Millipore) wicking pad and cut into 4-mm-wide strips. All other testing was done with LFIAs optimized to the following specifications. MAb 10B1 was passively adsorbed onto 40 nm colloidal gold particles (DCN Diagnostics) and brought to an OD 530 of 20 in PBS, pH 8 containing 0.5% (v/v) Pluronic F127, 1% (w/v) BSA, 20% (w/v) sucrose, and 5% (w/v) trehalose. Gold conjugated MAb was dispensed using a BioDot XYZ3050 Airjet system at 10µL/cm onto a Fusion 5 conjugate pad (blocked as above) and then dried at 37ºC. Antibody solutions for the test line (1 mg/mL MAb 13E11 and 1 mg/mL MAb 14D12) and control line were contact-dispensed onto FF80HP nitrocellulose (GE) using a BioDot XYZ3050 Biojet system at 1µL/cm. LFIAs were assembled and cut as above with CFSP203000 wicking pads and blocked Standard 14 sample pads. LfiA evaluation with nasopharyngeal (np) specimens. Remnant NP washes from baboons directly challenged with B. pertussis (D420 strain) and CFU/mL data were provided by Dr. Tod Merkel at the Food and Drug Administration's Center for Biological Evaluation and Research. Baboon challenge experiments were approved by the FDA's IACUC and were performed in accordance with relevant guidelines and regulations. Remnant, deidentified NP swab specimens in Eswab Amies buffer (BD, Franklin Lakes, NJ) from patients with suspected pertussis infection and qPCR results 52 were provided by Dr. James Dunn at Texas Children's Hospital. NP washes were combined with 10% SDS, while NP swab specimens were combined with 0.5% SDS for a final concentration of 0.25% SDS for LFIA analysis.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).