Systemic and pulmonary C1q as biomarker of progressive disease in experimental non-human primate tuberculosis

Tuberculosis (TB) causes 1.6 million deaths annually. Early differential diagnosis of active TB infection is essential in optimizing treatment and reducing TB mortality, but is hampered by a lack of accurate and accessible diagnostics. Previously, we reported on complement component C1q, measured in serum by ELISA, as a candidate biomarker for active tuberculosis. In this work we further examine the dynamics of C1q as a marker of progressive TB disease in non-human primates (NHP). We assessed systemic and pulmonary C1q levels after experimental infection using high or low single dose as well as repeated limiting dose Mycobacterium tuberculosis (Mtb) challenge of macaques. We show that increasing C1q levels, either peripherally or locally, correlate with progressive TB disease, assessed by PET-CT imaging or post-mortem evaluation. Upregulation of C1q did not precede detection of Mtb infection by a conventional interferon-gamma release assay, confirming its association with disease progression. Finally, pulmonary vaccination with Bacillus Calmette Guérin also increased local production of C1q, which might contribute to the generation of pulmonary protective immunity. Our data demonstrate that NHP modelling of TB can be utilized to study the role of C1q as a liquid biomarker in TB protection and disease, complementing findings in TB patients.

Tuberculosis (TB) remains a highly significant burden to global health. In 2018, 6.4 million new cases of TB were officially notified to the World Health Organization (WHO) 1 . However, the WHO estimates the actual number of TB cases to be 10 million 1 , implying an underestimation of the true number of TB cases. Closing this gap in TB detection has the potential to prevent millions of deaths and to curb further dissemination of TB. Accordingly, the WHO has made early diagnosis of tuberculosis an integral part of their EndTB strategy.
After infection with Mycobacterium tuberculosis (Mtb), the causative agent of TB, individuals may develop latent TB infection (LTBI) or active progressive disease (although some may also clear the infection). Definitive diagnosis of active TB can only be made by detection of Mycobacterium tuberculosis (Mtb) in sputum, either by smear microscopy, culture or GeneXpert technology. These assays, however, are time-consuming, require specific (expensive) infrastructure and suffer from sampling error and low sensitivity. Immunological tests, such as Tuberculin Skin Testing (TST) or Interferon Gamma Release Assays (IGRAs) can detect exposure to mycobacteria, but are unable to differentiate between active and latent TB infection 2,3 . These host response measures can remain positive even after treatment and clearance of infection. Accurate tests for risk of relapse after drug-treatment or risk of progression to active disease are currently unavailable. A wide range of new TB diagnostics are currently under development, investigating the diagnostic potential of pathogen-derived components as well as host-derived biomarkers in various body fluids 2,3 .
Over the last few years many host-derived candidate biomarkers of progressive TB disease have been identified, and amongst them several components of the complement system [4][5][6][7][8][9] . The complement system consists of a number of serum proteins with the capacity to recognize and neutralize invading pathogens by opsonization

Results
C1q is a predictive marker of disease progression in NHP TB. In addition to having described serum C1q as a biomarker of active TB in human patients, we also reported on elevated C1q levels in a preliminary analysis of serum and BAL samples from Mtb-infected non-human primates 18 . Here we set out to further investigate C1q as a marker of TB disease, under various model conditions in NHPs.
We started with retrospective measurement of C1q in serum from rhesus macaques (Macaca mulatta) that were either vaccinated or not with a standard human dose of intradermal BCG and subsequently challenged with a high dose (500 Colony Forming Units (CFU)) of Mtb strain Erdman, previously described in part in Lubbers et al. 18 . These animals were monitored for 1 year after infectious challenge, or until reaching a humane endpoint due to progressive TB disease. In this cohort, prior BCG vaccination was efficacious and resulted in significant improvement of survival after Mtb challenge (p = 0.0031, Fig. 1a) and reduction of TB associated pathology (p = 0.0196, Fig. 1b).
Serum C1q levels did not differ between the two groups prior to infection (Supplemental Fig. 1a). Assessing changes over time, we observed most prominent upregulation in non-vaccinated controls (nv.ctrls) already from 3 weeks after infection with Mtb (Fig. 1c). In some individuals, an increase in C1q levels preceded a humane endpoint event by several weeks. By week 6 post-Mtb infection, group median serum C1q in non-vaccinated controls was significantly higher compared to BCG vaccinees (Fig. 1d, left panel). Despite reducing TB pathology by prior BCG vaccination (Fig. 1b), when comparing C1q levels at individuals' endpoints, non-vaccinated controls and BCG vaccinated animals showed no significant difference in serum C1q (Fig. 1d, right panel).
To determine the potential of serum C1q as a prognostic marker regardless of prophylactic treatment and prior to the incidence of humane endpoints, we divided all animals in two groups irrespective of treatment: those exhibiting week 6 C1q levels below the median value (C1q.lo, 2 nv.ctrl and 9 BCG) and animals with week 6 levels above the median (C1q.hi, 9 nv.ctrl and 2 BCG). When comparing Kaplan-Meier curves for time-to-endpoint, animals with the highest C1q levels displayed significantly reduced survival (p = 0.0081, Fig. 1e). Furthermore, we found a strong statistical correlation between the total TB pathology score at necropsy and serum C1q levels at 6 weeks post-infection (Spearman's rho = 0.772, p < 0.0001, Fig. 1f, left panel). Individual C1q levels at endpoint correlated significantly with the total amount of pathology (Spearman's rho = 0.591, p = 0.003, Fig. 1f, right panel). Taken together, these observations suggest that C1q is an early marker of progressive disease after experimental infection of rhesus macaques.
C1q is a marker of differential disease severity in distinctive rhesus cohorts. We next sought to confirm the correlation of C1q upregulation with disease severity in an independent cohort of Mtb infected NHP. As a species, rhesus macaques can be divided into distinct subpopulations based on genetic variation that is associated with their geographical distribution 28 . When comparing Chinese versus Indian type rhesus macaques head-to-head, we found the latter to exhibit a more severe TB disease phenotype, 12 weeks after single high dose (500 CFU) Mtb challenge ( 23 & Fig. 2a, non-vaccinated controls). Prior intradermal BCG vaccination reduced TB associated pathology in Indian but not Chinese type rhesus macaques. We measured C1q in serum collected at 3 weekly intervals and assessed the association of C1q levels with TB pathology.
As observed previously, serum C1q levels increased during experimental infection with Mtb, from 3 weeks post-infection onward (Fig. 2b). The strongest increase in serum C1q was again observed in the animals with the highest disease severity, the non-vaccinated Indian rhesus macaques (Fig. 2b). Reduced TB pathology, as seen in Chinese type rhesus macaques or after prior BCG vaccination of Indian type rhesus macaques, is reflected in the limited upregulation of C1q in these groups (Fig. 2b). Since we observed a trend of higher C1q levels prior to infection in BCG vaccinated Chinese rhesus macaques in particular (Supplemental Fig. 1b), we used individual fold-increase of C1q (a relative measure of C1q) in subsequent analyses. When comparing values 6 weeks post-infection, relative C1q values were significantly higher in unvaccinated Indian rhesus macaques compared to BCG vaccinated Indian rhesus macaques or unvaccinated Chinese type macaques (Fig. 2c). Non-parametric Spearman's analysis of the relative C1q levels at week 6 versus total TB pathology scores revealed a strong correlation (rho = 0.7014, p = <0.0001, Fig. 2d), corroborating the association between C1q upregulation and tuberculosis disease severity.
Increased C1q is observed in the pulmonary space after Mtb infection. Over time we have moved away from high dose infection studies, since macaques appeared highly susceptible to infection and natural exposure to Mtb likely occurs at much lower doses than the 500 CFU applied in the studies described above 29 . To investigate whether increasing serum C1q still associates with disease severity after low dose Mtb infection, we assessed serum C1q dynamics in a cohort consisting of rhesus and cynomolgus macaques (Macaca mulatta and Macaca fascicularis, respectively) that were infected with a low dose (<10 CFU) of Mtb. The two species differ in the extent of TB pathology development after experimental infection; rhesus macaques typically tend to develop more severe TB pathology (Fig. 3a, 30,31 ). These animals were sacrificed either 6 or 12 weeks after Mtb infection, and displayed significantly less TB pathology compared to the animals from the studies described in Figs. 1 and 2 (Supplemental Fig. 2).
Baseline pre-infection levels of serum C1q did not differ between the two species (Supplemental Fig. 1c). After infection with low dose Mtb, overall we observed a modest increase in serum C1q only, most apparent between 3 and 6 weeks post-infection (with the exception of one rhesus showing a progressive increase) (Fig. 3b). However, despite increased TB disease levels in rhesus over cynomolgus macaques, there was no significant difference in  www.nature.com/scientificreports www.nature.com/scientificreports/ C1q levels between the two species, neither when comparing C1q levels at week 6 post-infection nor at endpoint (that is pooling endpoint values of animals sacrificed at week 6 or 12) (Fig. 3c, left and right panel, respectively).
Considering the relatively low serum C1q levels in these animals after low dose Mtb challenge, we interrogated C1q at the site of infection by measuring C1q in BAL fluids collected at various timepoints post-infection. Prior to Mtb infection, C1q was virtually undetectable in BAL fluids from either species (Supplemental Fig. 1c, right panel). However, from 3 to 6 weeks post-infection we observed a marked increase in C1q in the BALs of rhesus macaques (Fig. 3d), though C1q levels dropped between week 6 and week 12. In contrast, this marked increase in local C1q was absent in cynomolgus macaques (except for one animal with an (unexplained) outlier measurement at week 3). When comparing C1q values at 6 weeks post-infection and at study endpoint, increased C1q levels were observed in the rhesus but not the cynomolgus cohort (Fig. 3e). As expected from the above findings, serum C1q levels did not correlate with TB disease scores (Fig. 3f). But, when analyzing C1q expression in BAL 6 weeks post-challenge, we did find a statistically significant correlation with TB pathology (Spearman's rho = 0.599, p = 0.005, Fig. 3g). Thus, also after low dose Mtb challenge and subsequent mild(er) TB disease, increasing C1q levels remain associated with disease progression and pathological involvement, albeit locally rather than peripherally.
C1q upregulation does not precede IGRA conversion. Having established that C1q is upregulated either peripherally and/or locally in association with various TB (challenge and) disease levels, we subsequently wanted to investigate if this increase in C1q could serve as a diagnostic marker of Mtb infection, with superior sensitivity to a conventional IFNγ Release Assay (IGRA). To address the diagnostic potential of C1q measurement, we profiled peripheral and local C1q levels in a challenge study in rhesus macaques in which we deployed a novel repeated limiting dose (RLD) Mtb challenge modality. In this RLD challenge model, we have demonstrated prevention of infection and disease after mucosal BCG vaccination (Fig. 4a,b, previously published and further detailed in Dijkman et al. 26 , and depicted here only to provide relevant background).
As observed after single low dose infection, RLD Mtb challenge resulted in a modest increase in serum C1q over the course of the infection phase, in particular in non-vaccinated control animals, which, expectedly displayed the highest tuberculosis pathology levels at study end-point (Fig. 4c). Animals protected from severe TB disease by prior BCG vaccination (by either standard intradermal or pulmonary mucosal route of administration) did not show any increase in relative serum C1q. Similarly, C1q is markedly increased in BAL fluids of the non-vaccinated control group, but not after prior BCG vaccination (with the exception of a few intradermally BCG vaccinated animals) (Fig. 4d). Also in this study, we found BAL as well as serum C1q levels to statistically correlate with the level of TB pathology at endpoint, once more confirming C1q upregulation to be associated with disease development and severity rather than infection (Fig. 4e,f). Along Mtb infection (rather than post mortem), disease severity can be assessed by means of PET-CT imaging, in which inflammation and metabolically active granulomas can be visualized using 18 F-fluorodeoxyglucose (FDG) as a PET tracer 32 . To investigate whether C1q levels correlate with TB pathology as measured by PET-CT, we plotted summed lung and bronchoalveolar Standard Uptake Values (SUVs) obtained 8 weeks after initial challenge against C1q levels measured in serum and BAL collected at that same timepoint. Again, we found both serum and BAL fluid C1q concentrations to correlate significantly (p < 0.05) with PET-CT signals of TB pathology (Spearman's rho 0.554 and 0.709 respectively, Fig. 4g,h).
We next set out to investigate if C1q upregulation preceded detection of Mtb infection by IGRA conversion and therefore aligned the C1q responses in the non-vaccinated control animals to their time point of IGRA conversion. As can be expected for a marker associated with progressive disease, we did not observe an increase in C1q prior to IGRA conversion, neither in serum nor in BAL (Fig. 4i,j). Not earlier than two weeks post-IGRA conversion local C1q levels were elevated over baseline. This observation further asserts the association of C1q with progressive, rather than latent or incipient, tuberculosis disease. Finally, to further assert the relation between serum C1q and disease progression, we pooled the data of all studies described and correlated serum C1q levels with the amount of "local" pathology in the lung and with the amount of disseminated disease, i.e. lung draining lymph node pathology and extra-thoracic pathology. As expected, serum C1q levels determined at end-point correlated most strongly with disseminated disease (for lung draining lymph node pathology: Spearman's rho = 0.697, P = < 0.0001, for extra thoracic pathology: Spearman's rho = 0.635, P = < 0.0001). Correlation of serum C1q with the amount of lung pathology was significant, but weak (Spearman's rho = 0.311, P = < 0.0001).

Pulmonary BCG vaccination increases local C1q production.
C1q is considered to contribute to control of infection by enhancing phagocytosis of pathogens, either by opsonization and specific uptake or by activating phagocytes through engagement of C1q-binding receptors 33 . As we have previously found mucosal BCG to be superior over intradermal BCG in preventing TB infection and disease, we here investigated BCG's capacity to induce upregulation of C1q. We determined C1q in serum and BAL fluid from intradermal and mucosal BCG vaccinated animals from the repeated low dose challenge study described above and in Fig. 4. Similar to what has been described for humans, BCG vaccination in rhesus macaques by either route of administration, does not lead to upregulation of serum C1q (Fig. 5a,b). Interestingly, pulmonary, but not intradermal, BCG vaccination resulted in a marked increase in local C1q either 3 or 8 weeks after vaccination, to return to baseline levels by week 12 after vaccination (Fig. 5c,d).
As we observed a local, but not peripheral, elevation of C1q levels, we hypothesized that local production of C1q could underlie the increase observed in the mucosally vaccinated animals. We sought to verify the possibility of local C1q production by in vitro stimulation of BAL cells versus Peripheral Blood Mononuclear Cells (PBMCs) from BCG vaccinated rhesus macaques (intradermal as well as mucosal, at 8 weeks after vaccination), and from low dose Mtb-infected animals (at 11 weeks after infection) and untreated controls. Cells were incubated either www.nature.com/scientificreports www.nature.com/scientificreports/ with culture medium to assess ex vivo C1q production or with dexamethasone plus IFNγ (Dex/IFNγ), a positive control stimulus known to induce C1q release 17 , to assess the potency of these cells to produce C1q.
In PBMCs, a trend towards higher production by unstimulated PBMCs from vaccinated or Mtb infected animals (Fig. 5e, left panel) was observed when compared to unvaccinated/uninfected control animals. Prior BCG vaccination also seemed to potentiate C1q secretion of PBMCs, as reflected by the increase in C1q production in response to Dex/IFNγ stimulation in the vaccinees (Fig. 5e, right panel). Locally, ex vivo C1q production was highest in unstimulated BAL cells of animals that were exposed to either mucosal BCG or Mtb (Fig. 5f, left  panel). After Dex/IFNγ stimulation, lower C1q production was observed in BCG vaccinated animals compared to non-vaccinated controls, potentially reflecting alterations in the cellular composition of the BAL (Fig. 5f, right panel) 26 . Collectively, our data show that pulmonary mycobacterial exposure can induce the production of local C1q, likely by resident alveolar macrophages.

Discussion
In this work, we show that increasing C1q levels in serum and BAL are associated with TB disease severity in various non-human primate TB studies. High C1q levels did not precede detection of Mtb infection by IGRA conversion, but did correlate with TB pathology after both high (500 CFU) and low (<10 CFU) dose Mtb challenge in different macaque species. Likewise, in a long-term follow up (up to 1 year) setting, we found increasing C1q levels to be associated with reduced survival. Contrarily to what has been described for serum C1q, increasing www.nature.com/scientificreports www.nature.com/scientificreports/ pulmonary C1q was not exclusive to infection with Mtb; pulmonary vaccination with BCG also resulted in a temporal increase of C1q in BAL fluid. Whether this local C1q production plays a role in protection against Mtb infection and disease remains to be investigated.
Our data demonstrates that, in addition to being a marker of active TB in humans, C1q can also serve as a marker of progressive disease in experimental Mtb infection studies in NHP. C1q levels measured after Mtb infection correlated with disease severity measured post-mortem, but also with disease severity measured during infection by PET-CT. As C1q can be readily measured over the course of Mtb infection, it could therefore be applied to monitor TB disease progression in a resource-limited setting. Especially in cynomolgus macaques, which are known to develop latent TB infection 24 , measuring C1q at regular intervals after Mtb infection would be informative when assessing the occurrence and reactivation of LTBI. While the cohort of cynomolgus macaques measured in this study was not followed long enough to confirm the establishment of such a latent infection, the lack of C1q upregulation in the majority of this group might be indicative of TB latency development in these animals. As no post-infection increase in serum C1q could be observed after protective BCG vaccination, C1q can be used to assess efficacy along the infection phase when evaluating new drug or vaccine regimens in the NHP model.
Next to being a marker of progressive disease, it is tempting to speculate that C1q might also play a role in shaping the protective immune response observed after mucosal vaccination with BCG. Macrophages, including alveolar macrophages, express receptors, such as gC1qR and cC1qR (calreticulin), that can interact with either the globular head or collagen-like tail of C1q 11 . Not only does binding of (pathogen-bound) C1q to these receptors improve pathogen uptake 33 , C1q-receptor interaction can also modulate macrophage inflammatory status and cytokine responses. Monocytes incubated with C1q have been described to produce higher amounts of anti-inflammatory IL-10 34 , and it has been shown that C1q can polarize macrophages to an anti-inflammatory M2 phenotype, associated with the resolving of inflammation and tissue repair 35 . After mucosal BCG vaccination we observed an increase in pulmonary C1q (Fig. 5b), which coincides with increased PPD-specific IL-10 production by BAL cells 26 . This increased IL-10 production in response to PPD is sustained up to several weeks after Mtb challenge, and might be required to counterbalance the inflammatory response induced by BCG/Mtb, to prevent inflammation-induced damage 36 . However, both IL10 production and M2 polarization have also been identified as detrimental in the context of Mtb infection 37,38 , so the exact contribution of C1q to TB disease development remains to be elucidated. Lastly, C1q has the capacity to reduce production of IFNα 39,40 , which has been implied as detrimental in the (late phase) host immune response to Mtb 41,42 . In addition to modulating innate immune responses, C1q has also been described in the regulation of adaptive immune responses, either indirectly through regulating APC function or through direct interaction with C1q-receptors expressed on T-cells. However, reports on the capacity of C1q to directly activate T-cells are contradictory and need further investigation 43,44 . To fully appreciate the mechanistic role of C1q in shaping the protective immune response after mucosal BCG vaccination further research is required.
In summary, this work corroborates and extends on findings in human TB patients and links increased C1q levels with TB disease severity. It demonstrates that macaques can serve as informative model animals to study the role of (pulmonary) C1q in TB pathogenesis and/or protective immunity.

Materials and Methods
Animals & ethics. All housing and animal care procedures took place at the Biomedical Primate Research Centre (BPRC) in Rijswijk, the Netherlands. The BPRC is accredited by the American Association for Accreditation of Laboratory Animal Care (AAALAC) and is compliant with European directive 2010/63/EU as well as the "Standard for Humane Care and Use of Laboratory Animals by Foreign Institutions" provided by the Department of Health and Human Services of the US National Institutes of Health (NIH, identification number A5539-01). Before the start of each study ethical approval was obtained from the independent animal ethics committee (in Dutch: Dierexperimentencommissie, DEC), as well as BPRC's institutional animal welfare body (in Dutch: Instantie voor Dierwelzijn, IvD).
Animals included in each study were screened negative for prior exposure to mycobacteria by means of tuberculin skin testing with Old Tuberculin (Synbiotics Corporation, San Diego, CA) and an IFNγ ELISPOT against Purified Protein Derivative (PPD) from Mycobacterium bovis, Mycobacterium avium (both Fisher Scientific, USA) or Mycobacterium tuberculosis (Statens Serum Institute, Copenhagen, Denmark).
For the duration of the study animals were socially housed (pair-wise) at animal biosafety level 3. Animal welfare was monitored daily. Macaques were provided with enrichment in the form of food and non-food items on a daily basis. Animal weight was recorded prior to each blood collection event. To limit possible discomfort due to severe TB disease humane endpoints were predefined. All animal handling and biosampling was performed under ketamine sedation (10 mg/kg, by intra-muscular injection). When performing endobronchial challenge with Mtb or BAL, ketamine sedation was supplemented with intramuscular medetomidine (0.04 mg/kg) and an analgesic sprayed into the larynx. At the end of the study or when reaching a humane endpoint, animals were euthanized by intravenous injection of pentobarbital (200 mg/kg) under ketamine sedation. Veterinary staff and animal care-takers were blinded to animal treatment. www.nature.com/scientificreports www.nature.com/scientificreports/ Serum and BAL-fluid collection. Serum and BAL fluid were collected at various timepoints after Mtb challenge or BCG vaccination, as indicated in the figure axes. For serum, peripheral blood was collected in serum separator tubes by means of venipuncture. Tubes were spun for 10 minutes at 1000 g to harvest cell-free serum, which was stored at −80 °C pending further analysis. BAL was performed by targeting the lower left lung lobe by bronchoscope, followed by instillation and recovery of three times 20 mL prewarmed 0.9% saline solution. BALs were first passed over a 100 μm filter to remove mucus and debris, followed by centrifugation for 10 minutes at 400 g. Supernatant was decanted and stored at −80 °C pending further analysis. Serum and BAL fluid were filter-sterilized by centrifugation through 0.2 μm PVDF membrane plates (Fisher Scientific) before analysis. post-mortem pathology scoring. Post-mortem tuberculosis pathology was scored by a semi-quantitative grading system (adapted from 45 ) based on lesion size, manifestation and frequency, and lymph node involvement. Over time, from one to the other study, the scoring system has been adjusted slightly on minor details to more accurately describe disease manifestation, but the overall algorithm, as outlined in the paragraph below, has remained the same. When comparing disease severity between studies, pathology scores have been expressed as a percentage of the maximal possible score.
After euthanasia, the thoracic cavity, including the heart, ribcage, vertebrae and diaphragm were all macroscopically scored for the presence of granulomas and pleural adhesions. Lungs were isolated and lobes were separated from the trachea. Subsequently, lung lobes were cut in 5 mm thick slices and scored for the amount of pathology. Lung draining lymph nodes were removed from the trachea and scored for size and extent of involvement. Extra-thoracic organs such as kidneys, spleen, pancreas and liver were macroscopically assessed for the presence of lesions. The "Total Pathology score" that is depicted throughout represents the summed score of all these organs.

C1q production by PBMC and BAL cells.
To assess production of C1q by PBMC and BAL cells, freshly isolated cells were taken up in Roswell Park Memorial Institution 1640 medium (RPMI), supplemented with 10% Fetal Calf Serum (FCS), glutamine and penicillin/streptomycin. Cells were seeded at 200.000 cells per well in triplicate in 96-well round bottom plates and stimulated with 4 μg/mL dexamethasone (Merck) +200 U/mL IFNγ (Peprotech, UK), or left untreated. Supernatants were harvested after 96 hours, pooled and filter-sterilized by centrifugation through 0.2 μm PVDF membrane plates (Fisher Scientific) before storage at −80 °C and subsequent analysis by ELISA.
C1q ELISA. C1q levels in sera, BAL fluid and culture supernatants were measured using an in-house developed ELISA, as described previously 18 . In brief, 96-well Maxisorp plates (Nunc) were coated overnight with mouse anti-human C1q (Nephrology department, LUMC) in coating buffer (0.1 M Na 2 CO 3 , 0.1 M NaHCO 3 , pH9.6). The next day, plates were washed with PBS/0.05% Tween and blocked with PBS/1%BSA. After subsequent washing, samples (serum diluted 1:4000, BAL fluid 1:1, culture supernatant 1:1) were added to the plates. A serially diluted pool of normal human serum (NHS) was taken along as a standard. Bound C1q was detected by incubation with rabbit anti-human C1q, followed by goat anti-rabbit HRP (both Dako) and ABTS substrate. C1q concentration is calculated in μg/mL from a human C1q standard.
Statistics. Statistical analyses were performed using Graphpad Prism, software version 7. Significance of differences between groups was calculated by two-sided Mann-Whitney testing. All correlations were calculated by Spearman's log rank testing.

Data availability
The datasets generated and analyzed are available from the corresponding authors upon reasonable request. Likewise, biomaterials remaining from the studies described could be shared for further research.