Birth Cohort Study Identifies Candida albicans as a Risk Factor for Dental Caries

Abstract

Candida albicans has been implicated as a potential cariogenic microorganism, yet no prospective longitudinal studies have examined its role in severe early childhood caries (S-ECC). This study aimed to evaluate the association between oral C. albicans and the onset of S-ECC in a longitudinal setting. This prospective birth cohort study (2018 to 2023) enrolled 186 low-income pregnant women in their third trimester in Western New York, United States. Overall, 160 eligible infants born to these women were followed from birth to 2 y at 7 time points. Oral samples were collected to assess Candida species (C. albicans, Candida krusei, and Candida glabrata) and Streptococcus mutans. The primary outcome was the onset of S-ECC. Two-step LASSO (least absolute shrinkage and selection operator)–penalized logistic regression models were developed to identify predictive factors for S-ECC from 234 covariates grouped by their proximal association with S-ECC: infant oral microorganisms, biological-environmental factors, and maternal characteristics. Logistic regression was used to validate the association between C. albicans and S-ECC. Among the 118 children who completed the study, 28% developed S-ECC. The racial background of the cohort was 57% Black, 21% White, and 22% other. Oral C. albicans colonized in 12% of infants at 1 wk, peaking at 57% by 18 mo. Salivary C. albicans was associated with a 4.47-fold increased risk for S-ECC (odds ratio [OR]; 95% CI, 1.28 to 15.58; P = 0.02), in addition to other risk factors, including plaque score (OR, 5.19; 95% CI, 2.10 to 12.83) and salivary S. mutans (OR, 9.74; 95% CI, 2.27 to 41.79). C. albicans demonstrated strong time sensitivity in predicting S-ECC as early as 1 mo, contrasting with S. mutans, which exhibited predictive ability after 1 y. Oral C. albicans could serve as a novel biological marker for predicting ECC risk in infancy, shining a light on opportunities to develop innovative caries-predictive and preventive strategies targeting fungal contributions in pediatric care settings.

Introduction

Early childhood caries (ECC) remains the most common chronic childhood disease worldwide, with nearly 1.8 billion new cases each year (Dye et al. 2015). Severe ECC (S-ECC) occurs in children younger than 3 y and in children 4 to 6 y of age with elevated caries scores (Colak et al. 2013). S-ECC can develop quickly and often exhibits an aggressive and painful progression. Untreated S-ECC is associated with a higher risk of caries lesions in permanent teeth, diminished oral health–related quality of life, hospitalizations and emergency room visits due to systemic infection, and even death (American Academy of Pediatric Dentistry Council on Clinical 2005; Casamassimo et al. 2009). Treating S-ECC frequently involves total oral rehabilitation under general anesthesia with multiple tooth extractions and restorations/crowns at a cost of >$7000 per child (Rashewsky et al. 2012). In the United States, >$1.5 billion per year is spent on treatment. Hence, better identifying high-risk children and more effective preventive strategies is critically needed.

ECC is a biofilm (plaque)–dependent infectious disease. Although factors such as sugar intake, salivary function, and oral hygiene behavior are associated with ECC, caries is primarily the product of an orchestrated sequence of microbiological events that occur on tooth surfaces (Caufield and Griffen 2000; Tinanoff and Reisine 2009). While Streptococcus mutans, Scardovia wiggsiae, and certain lactobacilli have traditionally been considered the prime microbial risk markers and preventive targets for S-ECC (Caufield et al. 1993; Klein et al. 2004; Li et al. 2005; Kanasi et al. 2010; Slayton 2011; Tanner et al. 2011; Zhan et al. 2012; Klinke et al. 2014), recent research on the role of Candida albicans in S-ECC and its synergistic interaction with S. mutans has shed new light on potential fungus-focused approaches to early prediction and subsequent prevention of S-ECC (Koo and Bowen 2014; Koo et al. 2018; Xiao et al. 2018; Kim et al. 2021; Alkhars et al. 2022; Xiao et al. 2022; Xiang et al. 2023).

Intriguingly, even though researchers (Pienihäkkinen 1988) proposed in the late 1980s that salivary Candida levels had better caries-predictive power than salivary Lactobacillus levels, the predictive power of oral Candida for caries initiation and progression has not been evaluated rigorously. Moreover, until today, to our best knowledge, only cross-sectional human studies have been performed thus far to assess the association between oral C. albicans and ECC (Khachatryan et al. 2024). Without prospective cohort studies through the age of S-ECC onset and further clarification of microbiological contributions, it remains unclear how Candida detection is linked with the disease process and whether this fungal organism could be a reliable marker for risk of onset and progression of S-ECC. Thus, we hypothesize that the detection of oral C. albicans could be used as a novel biological marker for S-ECC risk prediction. The objective of the study was to evaluate the association between oral C. albicans and the onset of S-ECC in a longitudinal setting.

Methods

Study Population

The study recruited 186 pregnant women from low-income backgrounds in their third trimester from patients visiting the University of Rochester Highland Family Medicine or the Eastman Institute for Oral Health Perinatal Dental Clinic. Following that, 160 eligible newborn infants were enrolled and followed for 2 y from 2018 to 2023. The inclusion and exclusion criteria are detailed in Appendix 1. This study adhered to the STROBE reporting guidelines (Strengthening the Reporting of Observational Studies in Epidemiology). The study protocol was approved by the University of Rochester Research Subject Review Board.

Comprehensive Oral Examination and Data Collection

A detailed protocol was used for clinical examination and sample collection, employing previously used methods established by our study team (Xiao et al. 2016; Xiao et al. 2019; Alkhars et al. 2022; Alkhars et al. 2023; Gilbert Klaczko et al. 2023; Alkhars et al. 2024). An oral examination, including caries score, plaque index, and oral candidiasis, was performed by 1 of 3 calibrated dentists on the pregnant women during the third trimester and their infants at 1, 2, 4, 6, 12, 18, and 24 mo (with a ±15-d window for 0- to 6-mo visits and a ±2-mo window for 12- to 24-mo visits) in a dedicated examination room at the University of Rochester clinics, using standard dental examination equipment. Methods are detailed in Appendix 1. Inter- and intraexaminer agreement for the evaluated criteria was calculated by kappa statistics and exceeded 83% at the calibration. Recalibration and intra-/interexaminer reliability were assessed annually.

Demographic data including race and ethnicity were self-reported. Medical background data were obtained via self-reporting and confirmation by electronic medical records, including disease types, medication usage, and smoking status (mothers). Data on socioeconomic status, oral hygiene practices, and feeding methods were collected by questionnaires. Sweet and nonsweet indices were created according to the foods and drinks consumed, as detailed in Appendix 1.

Oral samples (saliva/plaque/swab) were collected with a sterile device, stored on ice after collection, and transferred to the laboratory within 2 h for microbiological assays. Candida species (C. albicans, Candida krusei, and Candida glabrata) and S. mutans were identified and quantified by a culture-dependent method and validated by a colony polymerase chain reaction method, as described previously (Xiao et al. 2016; Xiao et al. 2019; Zeng et al. 2020) and detailed in Appendix 1.

Statistics Analysis

The primary outcome was the onset of S-ECC among infants by 2 y of age. The independent variables were classified into 4 platforms based on their proximity to S-ECC: maternal sociodemographic, behavioral, and environmental oral characteristics (77 variables); children’s care provider feeding and brushing characteristics (43 variables); children’s sociodemographic and biomedical factors (79 variables); and children’s oral cariogenic microbial carriage (35 variables).

A 2-step procedure was used to integrate the factors from the 4 platforms. First, a penalized logistic regression model was fit per the LASSO penalty (least absolute shrinkage and selection operator) within each platform, with the response variable being the binary indicator of whether the child had caries or not (1 = S-ECC, 0 = caries free). The LASSO penalty was used to implement variable selection, and the tuning parameter was chosen by 3-fold cross-validation. Three-fold cross-validation is a process by which the data set is divided into 3 sets of equal size. The model is fit by using 2 data sets and then checked with the 1 left out. This process is repeated, leaving out a different third of the data each time. The model selected is determined by the performance of the model on the set of “left out” data. Next, the selected variables from each of the 4 models were combined and used for the selection in the final LASSO-penalized logistic regression model with the same binary S-ECC outcome variable. Again, 3-fold cross-validation was used to determine the penalty parameter and therefore the number of variables selected for the final model. A solution path was created to show the order of the variables entering each model. The receiver operating characteristic curve for the final 2-step model was plotted, and the area under the curve was calculated to assess the discrimination ability for ECC prediction.

Furthermore, we used multiple logistic regression to validate the effect of C. albicans colonization and other known variables on S-ECC onset. The variables included were infant gender, ethnicity, race, child oral hygiene status (plaque score), feeding method (exclusive breastfeeding, solid food introduction before age 1 y), sugar consumption (sweet and nonsweet index), salivary S. mutans and C. albicans carriage, and the number of erupted teeth.

All statistical tests were 2-sided at a 0.05 significance level. For sample size consideration, see Appendix 1.

Results

The racial background of the infants was 57% Black, 21% White, and 22% other. A total of 118 infants completed the study, resembling the initially enrolled 160 infants (Appendix Table 1). Appendix Table 1 and Appendix Figure 2 show the demographics, socioeconomics, oral conditions, feeding patterns, and caregiving of infants. Appendix Figure 1 shows the enrollment and retention of the study cohort.

S-ECC Outcomes

Among the 118 infants who completed the study, S-ECC was observed in 3% at 12 mo, 9% at 18 mo, and 15% at 24 mo (Fig. 1A). By 24 mo, 28% of the infants (33 cases) had developed S-ECC (Fig. 1B). Notably, the incidence of S-ECC was higher among Black infants (34%) as compared with their White counterparts (20%; Fig. 1C). Examination of decayed (cavitated) teeth, stratified by race, revealed that Black infants had a higher mean number of cavitated teeth at 24 mo, averaging 4.8 teeth (Fig. 1D). S-ECC progressed rapidly, as shown in Figure 1E, with a white spot lesion at 12 mo advancing to cavitation within 6 mo.

Figure 1.

Severe early childhood caries (S-ECC) outcomes among the US underserved infants. (A) S-ECC new cases from 12 to 24 mo after birth. The incidence of S-ECC increased from 3% at 12 mo to 15% at 24 mo. (B) The accumulated S-ECC cases were 28% of the infant cohort at 24 mo. (C) S-ECC incidence stratified by race. The incidence of S-ECC was higher among Black infants (34%) versus their White counterparts (20%). (D) Number (mean ± SD) of decayed (cavitated) teeth at 18 and 24 mo, stratified by race. Black infants had a higher number of decayed (cavitated) teeth at 24 mo, with a mean 4.8 teeth. (E) An example of S-ECC progression from white spot lesion at 12 mo to cavitation at 18 mo. Blue arrow: white spot. Yellow arrow: cavitated lesion.

Oral Colonization of Candida and S. mutans

Figure 2A (saliva) and Figure 2B (plaque) illustrate oral colonization of Candida spp. and S. mutans in infants from birth to 24 mo. Oral Candida was detected in 12% of infants as early as 1 wk, which increased to 39% by 2 mo, peaked at 57% by 18 mo, and then declined to 39% by 24 mo. C. albicans was the most prevalent species, followed by C. krusei and C. glabrata in saliva and plaque. Salivary S. mutans increased from 1% at 1 mo to 49% by 24 mo, while plaque S. mutans rose from 22% at 12 mo to 56% at 24 mo.

Figure 2.

Oral colonization of Candida and Streptococcus mutans among the US underserved infants. (A) Salivary detection of Candida species and S. mutans from birth to 24 mo. S. mutans detection rate increased from 1% at 1 mo to 49% at 24 mo. C. albicans was the most frequently detected Candida species. (B) Detection of Candida species and S. mutans in dental plaque from 12 to 24 mo. (C) Detection of Candida species at different sites of the infant’s oral cavity. No differences were revealed among oral sites in terms of the detection rate. (D) Stability of oral Candida in the oral cavity. Stability is defined as those infants who had positive Candida and remained positive until 24 mo. C. albicans remains with high stability (63% to 70%) once colonized on soft tissues, such as the tonsil, tongue, palate, cheek mucosa, and lip, as well as in dental plaque. (E) Candida species diversity in salivary samples. The oral Candida species composition was more diverse when infants got older. (F) Salivary carriage of Candida species and S. mutans. Notably, a significant portion of infants had >400 CFU/mL of C. albicans in saliva, which could be diagnosed with oral candidiasis by using the laboratory standard established by Epstein et al. (1980). Line and error bar indicate mean ± SD. Ca, Candida albicans; CFU, colony-forming unit; Cg, Candida glabrata; Ck, Candida krusei.

We assessed Candida detection at the tonsil, tongue, palate, inner lip, and cheek (Fig. 2C). C. albicans detection ranged from 33% to 38% across sites, with no significant differences (P > 0.05). The tongue was the most common site for C. krusei (25%) and C. glabrata (13%), followed by the tonsil, palate, cheek, and inner lip. Candida detection remained stable over time (Fig. 2D). Stability was defined as infants who had positive Candida and remained positive for the remaining study period. Once colonized on soft tissues, C. albicans showed high stability (63% to 70%), whereas C. krusei and C. glabrata were less stable (31% to 40% and 19% to 33%, respectively).

We investigated the diversity of oral Candida in infants (Fig. 2E). Most infants initially carried only a single Candida species (88% at 1 mo, 76% at 4 mo), but their oral Candida composition became more diverse over time. Figure 2F shows the salivary carriage of Candida and S. mutans. Although few infants were clinically diagnosed with oral candidiasis, many had C. albicans levels exceeding 400 CFU/mL in saliva, meeting the Epstein et al. (1980) laboratory standard for oral candidiasis diagnosis.

S-ECC Prediction Models

In step 1 of the 2-step model-building approach, we fit 4 separate LASSO-penalized logistic regression models. The full list of variables is in Appendix Table 2. From platform 1, 7 maternal factors were identified (Fig. 3A). Maternal perinatal dental treatment was negatively associated with S-ECC, whereas the other 6 factors were associated with an increased risk for S-ECC: postpartum Gardnerella vaginitis, prenatal Candida vaginitis, postpartum oral Candida carriage, mothers having ≥3 missing teeth, the number of children, and maternal missing tooth surfaces. From platform 2, 6 child environmental factors were identified (Fig. 3B), with plaque score being the only factor positively linked to S-ECC. The remaining 5 protective factors for S-ECC were chips consumption, erupted teeth number, paternal care (dad involved as care provider), nonsweet index, and exclusive breastfeeding. Furthermore, platform 3 revealed 4 child sociodemographic-biomedical factors that increased the risk for S-ECC (Fig. 3C): child oral thrush, intake of fluconazole and nystatin, and Black race. From platform 4, the detection of S. mutans in saliva and C. albicans and S. mutans in plaque increased the risk for S-ECC (Fig. 3D). The predictive models for each platform are detailed in Appendix 2.

Figure 3.

Identified factors associated with severe early childhood caries (S-ECC) onset via a 2-step model. LASSO-penalized logistic regression modeling was used for caries predictor selection based on 234 variables from maternal, provider-feeding, child demographic-biological, and child microbial factors. (A–D) Variables from 4 platforms were identified in association with S-ECC onset: maternal factors, child environmental factors, child sociodemographic-biomedical factors, and child microbial factors. The final 2-step model incorporated variables identified from the aforementioned 4 individual platforms, and the LASSO solution path for the final model is shown in panel E. Panels A to E indicate how the models are built sequentially by adding 1 variable at a time to the active set. (F) The rationale of grouping variables into the 4 platforms is based on consideration of the distal and proximal determinants of S-ECC. LASSO, least absolute shrinkage and selection operator. C, child factor; M, maternal factor.

These 20 variables from step 1 were then used as the candidate variables in step 2. The final model selected 12 variables. Their order of being selected into the model is shown in the solution path (Fig. 3E). There were 9 risk factors for S-ECC: child plaque carriage of C. albicans and child plaque score, which entered the model first with high coefficients (2.20 and 1.75, respectively), followed by child oral thrush, child plaque S. mutans, and child salivary S. mutans with coefficients of 1.68, 2.33, and 0.83 and, finally, mothers’ prenatal Candida vaginitis and postpartum Gardnerella vaginitis. The S-ECC protective factors identified were erupted teeth number, nonsweet index, paternal care, chips consumption, and mothers’ perinatal dental care. Notably, paternal care had a significant impact on lowering caries risk with a coefficient of −1.01.

The LASSO 2-step final model, with an area under the curve of 0.95, demonstrated strong discriminative capabilities for S-ECC within our birth cohort. The final predictive model is the following:

$$\begin{array}{l}\mathrm{logit(p)}=\times \beta =-0.60+2.33\mathrm{plaque}S.mutans+2.20\mathrm{plaque}\\ C.albicans+0.83\mathrm{salivary}S.mutans+1.75\mathrm{plaque}\mathrm{score}\\ +1.68\mathrm{oral}\mathrm{thrush}+1.01\mathrm{maternal}\mathrm{prenatal}Candida\mathrm{vaginitis}\\ +0.48\mathrm{postpartum}Gardnerella\mathrm{vaginitis}-0.56\mathrm{nonsweet}\mathrm{index}\\ -0.37\mathrm{erupted}\mathrm{teeth}\mathrm{number}-1.35\mathrm{chips}-2.31\mathrm{perinatal}\\ \mathrm{dental}\mathrm{treatment}-1.01\mathrm{dad}\mathrm{care}\mathrm{provider}.\end{array}$$

Oral C. albicans and S-ECC Onset

We validated the association between C. albicans and S-ECC through multiple logistic regression. This analysis included known risk and protective factors for S-ECC (Fig. 4). The presence of salivary C. albicans was associated with a 4.47-fold increased risk for S-ECC (odds ratio [OR]; 95% CI, 1.28 to 15.58; P = 0.02), in addition to risk factors such as plaque score (OR, 5.19; 95% CI, 2.10 to 12.83; P = 0.0004) and salivary S. mutans (OR, 9.74; 95% CI, 2.27 to 41.79; P = 0.002). Conversely, children who consumed more food categorized with a higher nonsweet index (OR, 0.48; 95% CI, 0.28 to 0.77; P = 0.004) and children who had a greater number of erupted teeth (OR, 0.71; 95% CI, 0.51 to 0.92; P = 0.02) had a decreased risk for S-ECC by 2 y of age.

Figure 4.

Factors associated with severe early childhood caries (S-ECC) incidence from the multivariate logistic model. The multivariate logistic regression model indicated analyzed factors associated with S-ECC incidence by 2 y of age. The factors included in the model were demographics, care provider, feeding, sweets consumption, brushing (plaque score), and microbial factors. Nonsweet index and erupted teeth number provide a protective role against S-ECC, lowering S-ECC risk. Children with a high plaque score had a 5.19-times higher risk for S-ECC than children with a low plaque score. Salivary S. mutans was positively associated with S-ECC incidence, as well as salivary C. albicans. OR, odds ratio.

Furthermore, we developed time-specific S-ECC prediction models for cross-sectional time points using oral microbial platform (Fig. 5). Intriguingly, C. albicans demonstrated strong time-specific sensitivity in predicting S-ECC as early as 1 mo after birth, contrasting with S. mutans, which exhibited discriminative ability at a later stage, typically after the first year of infancy.

Figure 5.

Factors identified with predictive power for severe early childhood caries (S-ECC) at cross-sectional time point. Variables associated with S-ECC were selected from the child microbial platform at various points in early life. The color bar represents relative risk. Numbers highlighted in red indicate an increased risk for S-ECC (>0), while numbers highlighted in green indicate a reduced risk for S-ECC (<0).

Discussion

To our best knowledge, this is the first birth cohort study that longitudinally evaluated the association between oral Candida and S-ECC onset among US infants. Our findings indicate that the detection of C. albicans in the oral cavity of infants could serve as a novel biological marker for predicting S-ECC risk prior to tooth eruption and could be modified with proper screening and intervention. Based on our findings and scientific premises, we suggest a transformative hypothesis regarding the role of C. albicans in ECC: early colonization of the oral cavity by C. albicans might promote the subsequent colonization by S. mutans, a well-known culprit for dental caries, as shown by our previous findings (Alkhars et al. 2022) that infants with early oral Candida colonization had a 3.5-fold higher risk of S. mutans emergence by year 1 than those without Candida. The presence of oral Candida could also modulate oral microbiota development toward a state conducive to caries (Montelongo-Jauregui and Lopez-Ribot 2018; Xiao et al. 2018; Gilbert Klaczko et al. 2023) during crucial stages of dental development in early infancy. These chain events could lead to early onset and a more severe form of S-ECC.

In fact, robust scientific evidence supports the cariogenic potential of C. albicans. First, C. albicans is acidogenic and aciduric, and it is capable of dissolving hydroxyapatite, a major tooth component, at a rate approximately 20 times higher than S. mutans (Nikawa et al. 2003). C. albicans has also been shown to form a complex extracellular polysaccharide-mediated biofilm that facilitates cohabitation with S. mutans in the oral cavity (Gregoire et al. 2011; Falsetta et al. 2014; Koo et al. 2018; Kim et al. 2021; Xiang et al. 2023). Furthermore, C. albicans is capable of enhancing virulence and causing more severe caries when coinfected with S. mutans in a rodent model (Falsetta et al. 2014). The symbiotic relationship between C. albicans and S. mutans, as modulated by shared carbohydrate metabolism and environmental acidification, plays a crucial role in the pathogenesis of ECC (Jin et al. 2024).

Given the aforementioned studies confirming the potential cariogenicity of C. albicans, our further question is, where did the infants acquire oral C. albicans? Our study concurrently assessed the genetic relatedness between C. albicans isolates from mothers and infants within the same cohort, revealing a significant maternal influence on the acquisition of oral C. albicans in early infancy (Alkhars et al. 2024). These results underscore the potential benefits of screening for oral fungal carriage and administering effective antifungal treatments to mothers and infants during the perinatal period to improve existing ECC prevention strategies.

Furthermore, findings from this longitudinal birth cohort study suggest a potential shift in clinical practice that could improve children’s oral health: 1) implement early caries risk assessments that incorporate fungal predictors alongside traditional risk factors for ECC, starting at birth; 2) reduce ECC risk by introducing antifungal strategies during infancy, prior to the eruption of teeth; and 3) enhance the prevention of ECC relapse by incorporating antifungal regimens following conventional restorative treatments for children.

Additionally, since Candida identification relies on a well-established medical laboratory test that can be ordered by physicians, the availability of a fungal-aided ECC prediction and prevention strategy could facilitate a shift from the current dentist-centric care model to a collaborative model involving health care practitioners and dentists. This paradigm shift could be particularly beneficial for underserved children who have limited access to dental care, as it would allow for the initiation of ECC prevention outside of dental clinics.

Limitations

When our study results are being interpreted, several limitations should be considered. First, the demographic and socioeconomic characteristics of the infants are specific to Western New York, limiting generalizability. While the study suggests that C. albicans could serve as an early biomarker for caries prediction, independent validation with external data sets is necessary before recommending screening or treatment. Second, feeding data were collected via questionnaire, introducing recall bias. We did not collect data on specific brands of snacks and drinks, posing challenges in interpreting when chips were identified as S-ECC protective. Our research focused on the link between C. albicans and S-ECC. Further research is needed to assess the impact of early-life C. albicans colonization on the establishment of the early-life oral microbiome, which is an ongoing project by our team. Additionally, there may be unmeasured confounders that could influence C. albicans and S. mutans colonization and S-ECC onset, such as host genetics, immune factors, or other microbial species. Several longitudinal studies have shown classification of ECC status by other bacterial species before colonization with S. mutans in early life (Dashper et al. 2019; Blostein et al. 2023). Candida dubliniensis was identified in the saliva of caries-affected children (Lozano Moraga et al. 2017). A cross-sectional study (Khan et al. 2024) examining the fungal profiles of preschool children found significant enrichment of C. dubliniensis in dental plaque from children with ECC, highlighting its potential role in caries. Our study assessed the common Candida species (C. albicans, C. krusei, and C. glabrata) by using CHROMagar Candida, which intends to differentiate the aforementioned common Candida species by colony color but not C. dubliniensis. To capture the full mycobiome profile that includes emerging species such as C. dubliniensis, future study should comprehensively assess oral fungal profiling in infancy with methods such as internal transcribed spacer or metagenomic sequencing. Only well-designed longitudinal studies or randomized controlled trials of antifungal treatments can definitively clarify the role of C. albicans in S-ECC.

Conclusions

Our study demonstrates that early oral colonization by C. albicans strongly predicts S-ECC in infants, especially among high-risk children. While previous cross-sectional studies suggested a link between C. albicans and ECC, our longitudinal cohort confirms its association with an elevated risk of S-ECC. These findings, with future large-scale cohort studies, could inform the development of novel fungal-based strategies for caries prediction and prevention.

Author Contributions

N. Alkhars, S. Manning, Y. Zeng, acquisition, analysis and interpretation, drafted and critically revised the manuscript; N. Al Jallad, acquisition, analysis, critically revised the manuscript; T.T. Wu, analysis and interpretation, drafted and critically revised the manuscript; C. Fogarty, M. Mendoza, K. Fiscella, acquisition, critically revised the manuscript; E. van Wijngaarden, D.T. Kopycka-Kedzierawski, R. Billings, H. Koo, conception and design, acquisition, critically revised the manuscript; J. Xiao, conception and design, acquisition, analysis and interpretation, drafted and critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.