Saliva proteomic patterns in patients with molar incisor hypomineralization

Molar incisor hypomineralization (MIH) is an endemic pediatric disease with an unclear pathogenesis. Considering that saliva controls enamel remineralization and that MIH is associated with higher saliva flow rate, we hypothesized that the protein composition of saliva is linked to disease. To test this, we enrolled 5 children aged 6–14 years with MIH showing at least one hypersensitive molar and 5 caries-free children without hypomineralization. Saliva samples were subjected to proteomic analysis followed by protein classification in to biological pathways. Among 618 salivary proteins identified with high confidence, 88 proteins were identified exclusively in MIH patients and 16 proteins in healthy controls only. Biological pathway analysis classified these 88 patient-only proteins to neutrophil-mediated adaptive immunity, the activation of the classical pathway of complement activation, extracellular matrix degradation, heme scavenging as well as glutathione -and drug metabolism. The 16 controls-only proteins were associated with adaptive immunity related to platelet degranulation and the lysosome. This report suggests that the proteaneous composition of saliva is affected in MIH patients, reflecting a catabolic environment which is linked to inflammation.

ethical aspects and saliva collection. Approval for this clinical investigation was obtained from the ethics committee of the local University Review Board (Medical University of Vienna; Approval: 1463/2016). Written statements of consent were read and signed by children and their guardians prior to their participation and all experiments were performed in accordance with relevant guidelines and regulations of the Ethics Committee of the Medical University of Vienna. Study subjects were instructed to refrain from all oral hygiene procedures, chewing gum and painkillers for 8 h and from eating, drinking and brushing for 1 h prior to examinations 41,42 . Saliva collection was performed using Salivette (Sarstedt, Germany). Each patient was instructed to chew the swab for 120 seconds to stimulate salivation. Afterwards, the swab with the absorbed saliva was returned to the Salivette 43 . Samples were centrifuged for 10 min, 10000 rpm, 4 °C, and stored at −20 °C until further use.
Mass spectrometry. Saliva proteins were precipitated using methanol/dichloromethane and digested with trypsin as described earlier 44 (For detail see Suppl. Methods 1). Precipitated proteins were dissolved in 0.1% Rapigest (Waters, Vienna, Austria), dissolved in 50 mM triethylammonium bicarbonate, and protein concentration was determined using the Bradford assay. Proteins were digested overnight at 37 °C using a trypsin:protein ratio of 1:50, digestion was stopped by acidification with trifluoroacetic acid (TFA). Following injection onto the trapping column (Acclaim C18 trap column, 300μm inner diameter × 5 mm), peptides were separated by nano-reverse-phase (Acclaim C18, 75μm inner diameter × 500 mm) using an UltiMate nano RSLC HPLC (Thermo Fisher, Germering, Germany) separation system, consisting of the autosampler, column switching unit, nano and loading pump and UV detector. Both, trap-and separation columns were operated at 60 °C and UV peptide detection at 214 nm served as quality control for HPLC separation. Samples were loaded onto the trap column using 0.1% TFA at 30 μl/min and precooled to 3 °C 45 , nano separation was performed in gradient mode at 300 nl/min. A user defined injection program was used for sample injection and additional injector and trap column wash. Every sample injection was followed by two blank runs with injections of 2,2,2-trifluoroethanol for removal of possible sample remains in the injector or on the trap column and prevention of carryover in the separation system. Mass spectrometry (MS) analysis was performed using the Q-Exactive plus mass spectrometer (Thermo Fisher Scientific) and the "top 20" method for MS/MS experiment; that is, the 20 most intensive ions from the MS scan were selected for tandem MS (MS/MS), single-charged ions were excluded from fragmentation, and detected ions were excluded for further fragmentation for 2 min after initial MS/MS fragmentation had been performed. Mass resolution of 70000 was selected for MS at AGC set to 3E6 ions, MS/MS resolution was set to 35000 and AGC set to 1E5 ions. Fragmentation was performed using the HCD approach at normalized collision energy of 30 eV. Data analysis (database search and label-free quantitation) was performed using MaxQuant (version 1.6.0.1) with following parameters: MS/MS data were searched against the Human Fasta Database (Uniprot, version September 2018), MS/MS tolerance was set to 20ppm, deamidation on N and Q, N-term acetylation and oxidation on M were selected as variable modifications. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD016126 46 . Statistical analysis and data visualization. Bioinformatic analyses of protein identifications were done in Perseus (version 1.5.5.3) and overlapping proteins were visualized in Venn-Diagrams in combination with the CNB-CSIC online tool Venny 47 . Gene-Ontology as well as biological pathway enrichment analysis were done using the "Enricher" interactive data-analysis tool 48,49 . Throughout this report we express enriched terms by enumerating the actual number of proteins in our dataset in relation to the total number of proteins consolidated for the respective pathway. Normalizations were done using the web-based tool NormalyzerDE 50 . Spearman-rank correlation analysis and visualization of log2-transformed label-free quantification (LFQ) values, unsupervised hierarchical clustering and heat-maps of median, quantile and rank-normalized protein abundance www.nature.com/scientificreports www.nature.com/scientificreports/ values were done in Perseus using default settings (euclidian-distance on average linkage, pre-processing with k-means with a maximum of 300 clusters and 10 iterations). For group-wise comparisons we employed FDR-based 1-Way-ANOVA with 250 permutations and multiple-testing correction of p-values were done in InstantClue (version 0.5.2) using non-log transformed p-values for 2-stage-set-up Benjamini-Krieger-Yekutieli, alpha-error of 0.05 as well as Benjamini-Hochberg correction methods 51 . Volcano-plots were prepared in Perseus and proteins with a corrected p-value <0.05 were annotated with the corresponding gene symbol, while proteins with a non-adjusted p-value <0.05 were color-indicated only. For rank-normalization, only proteins with a corrected p-value smaller than 0.015 were annotated. For proteins exclusively present in patients or controls, a (−)log10-p-value of zero was assigned after all statistical analysis to enable visualization in Volcano plots. Visualization of protein-protein associations was performed with STRING v11. Results patient characteristics. In the MIH group, 4 female and 1 male patient with a sum of 23 affected teeth were included (mean age 8.70 ± 2.36 years). The caries-free control group included 3 male and 2 females (mean age 10.73 ± 0.54 years). MIH patients showed three up to four affected first permanent molars with at least one tooth exhibiting hypersensitivity. Detailed study-cohort characteristics are shown in Supplementary Table 1. In total, 87.5% of the teeth were hypersensitive with a mean Schiff Score of 2.3 ± 0.48 and a VAS of 7.03 ± 2.14.
proteomic analysis. In our high-resolution shotgun proteomic analysis, 462 (602) and 411 (530) proteins, identified and quantified with at least 2 peptides, were present in all 5 subjects or in at least 3 out of 5 subjects (numbers in parenthesis) within the patients and control group, respectively (Fig. 1A). Analysis of overlapping proteins among patients and controls is shown in Fig. 1B. The vast majority of proteins (514, 83.2%) were present in both patients and controls, however, 88 (14.2%) proteins were exclusively present in patient's saliva and 16 (2.6%) in controls only (Fig. 1B, Table 1 and Table 2). Visualization of the protein-protein associations showed 81 nodes and 105 edges on the proteins exclusively detected in MIH saliva, whereas 14 nodes and only 2 edges were shown for saliva of healthy controls (Fig. 2).

Label-free quantification (LFQ).
Applying an LFQ algorithm on our proteomic data, we found good individual correlations among controls (mean r = 0.866 ± 0.036) and patients (mean r = 0.829 ± 0.0518), although patient heterogeneity was higher ( Supplementary Fig. 1A). We also noticed that LFQ-intensities of proteins within the patient´s group were generally higher, impeding comparative statistical analysis without prior normalization. To this end, we performed median and quantile-normalizations as well as rank-normalization, which achieves robustness to non-additive noise at the expense of losing parametric information on abundance values 50,52 . To ensure a highly reliable dataset, we omitted imputation of missing values which is applied in many proteomics studies.    Fig. 2A-C). Pertinent statistical evaluation of differentially expressed proteins obtained by FDR-based 1-way ANOVA following each

Discussion
Here, we performed a mapping of the proteome of MIH saliva and respective controls from healthy individuals. Our findings show that out of 618 proteins, 88 and 16 proteins were exclusively detected in MIH saliva and control saliva, respectively. Proteins present exclusively in patient´s saliva were functionally linked to "neutrophil degranulation" with the highest enrichment score. In line, enrichment for Biological Process revealed "leukocyte mediated immunity", "neutrophil mediated immunity" and "neutrophil activation". Together, these analysis are indicative of ongoing activation and neutrophil degranulation, and supportive of the observed subclinical pulpal inflammation 9 , enhanced emigration of neutrophils into the inflamed pulp 10,11 and increased numbers of degranulated neutrophils in periodontitis patients 14 . It is thus likely that "neutrophil degranulation" is a confounding element of the salivary protein signature of MIH patients, reflecting ongoing inflammation. Thus, the disease specific signature we identified provides insight into MIH disease pathophysiology and present a potential basis for therapeutic monitoring.
Molecular Function analysis revealed significant enrichment of "catalytic activity" and "hydrolase activity" involving 43 and 26 proteins, covering 50% of the identified proteins in MIH saliva. Catalytic and hydrolase activities are associated with inflammatory processes including neutrophil degranulation, which is linked to tissue degeneration. In this regard, for example, prolyl endopeptidase (PREP), which is produced by neutrophils and cleaves collagen, thereby generating a neutrophil chemoattractant environment, may serve as a valuable biomarker and therapeutic target for diseases caused by chronic, neutrophilic inflammation 53 . Concordantly, interfering with proteolytic activities of the non-lysosomal thiol protease calpain-2 (CAPN2), present exclusively in MIH saliva, could potentially limit the ongoing tissue/bone degradation as calpain-2 inhibitor(s) reportedly reduce colitis and colitis-associated cancer through limiting macrophage activation and inhibiting growth of cancer cells 54 . We identified several proteins in MIH saliva associated with skin-abnormalities caused by chronic inflammation. For example, FUCA1 is a carbohydrate degrading enzyme and FUCA1 gene-mutations are linked to fucosidosis that causes severe skin abnormalities due to disturbed carbohydrate metabolism 55 . The human kallikrein 8 protein (KLK8) is expressed in many normal tissues including the salivary gland 56 . KLK8 serum levels are increased in psoriatic arthritis patients 57 and in the stratum spinosum during murine skin inflammation 58 . Notably, we also found a protein belonging to the peptidoglycan recognition proteins (PGLYRP2) which recognize bacterial peptidoglycan and functions in antibacterial immunity and inflammation. PGLYRP2 is reportedly produced by salivary glands 59 and its expression is upregulated by oral epithelial cells derived IL-36 cytokines in Scientific RepoRtS | (2020) 10:7560 | https://doi.org/10.1038/s41598-020-64614-z www.nature.com/scientificreports www.nature.com/scientificreports/ response to Porphyromonas gingivalis infections 60 . Though we did not detect the PGLYRP2 activating cytokine IL-36 in MIH-saliva, we found an antagonist of this signaling pathway (IL36RN), suggesting counterbalancing feedback mechanisms of this pathway at the receptor-ligand level 60 .
Negative feedback mechanisms limiting inflammation might also operate at the level of the proteasome as we identified proteasome subunits including PSMA2, functionally linked to inflammatory bowel disease 61 and PSMB1, described to suppresses innate antiviral immunity 62 . Additionally, we identified proteins exerting both pro-and anti-inflammatory properties in different cell types such as the GTPases RAB6A, RAP1B and SAR1B that regulate intracellular protein transport and secretion. While RAB6A facilitates TNF secretion following LPS stimulation of macrophages 63 , RAP1B limits neutrophil tissue infiltration in mice 64 . SAR1B reportedly protects intestinal cells from disorders of lipid homeostasis, oxidative stress, and inflammation 65 . Importantly, we found a remarkable accumulation of immunoglobulins in MIH saliva, a cardinal sign of inflammation. Summarizing, the protein signature of MIH patients is characteristic of other oral inflammatory diseases reflecting an overall principle rather than a disease specific pattern.
Among the chemokines and cytokines, CCL28, IL18 and IL36RN were exclusively identified in the saliva of MIH patients. CCL28 is produced by the salivary gland and displays strong homing capabilities for lymphocytes at mucosal and epithelial sites 66 . CCL28 is not detected in salivary glands of primary Sjögren's syndrome patients 67 . The stress induced cytokine IL-18 is also produced in salivary glands 68 and is prominent in the saliva of patients with oral lichen planus 69 and periodontitis 70 . Proteomic analysis has identified IL-18 as a biomarker in the saliva of burning mouth syndrome, a chronic pain disorder defined by a severe burning sensation in normal looking oral mucosa 71 . IL-36 can act on keratinocytes and immune cells to induce a robust inflammatory response and has been implicated in psoriatic disorders 72 as well as in inflammatory activation of oral epithelial cells 60 . Supportive of our hypothesis that increased expression of negative regulators mirrors active and ongoing inflammation and in line with our data, IL-36Ra/IL-36RN levels are reported to be higher in active versus inactive ulcerative colitis 73 . Together, proteomic profiles of MIH saliva point towards oral inflammation which is driven by neutrophil activation and degranulation.
Among the 16 control-only proteins enriched in healthy saliva, we identified chitotriosidase (CHIT1) an enzyme with the capacity to hydrolyse chitin, a structural component of fungi, parasitic nematodes, and insects 74 . CHIT1 might have a protective role against chitin-containing pathogens that is absent in MIH saliva. Lysosomal serine carboxypeptidase cathepsin A (CTSA) facilitates the activation of beta-galactosidase and alpha-neuraminidase. Further, CTSA plays a role in the inactivation of bioactive peptides including bradykinin, substances P, oxytocin, angiotensin I and endothelin-I 75 . Alpha-N-acetylgalactosaminidase enzyme (α-NAGA) belongs to the glycoside hydrolase family 27 that breaks down its substrates via the cleavage of their terminal N-acetylgalactosamine residues 76 . Thus, these aforementioned hydrolases might play a role in maintaining innate immune function in normal saliva. Intuitively, their absence in MIH patients makes sense.
In line with this concept, KEGG enrichment revealed HCLS1 and TUBA4A to be enriched for Pathogenic Escherichia coli infection (hsa05130; 2/53) and Tight junction (hsa04530; 2/167). For example, hematopoietic cell-specific protein-1 (HCLS1) regulates leukocyte actin remodeling and thereby their recruitment to sites of inflammation 77 . Further indicative of cytoskeletal effects, TUBA4A was enriched, however its role in oral health remains unclear 78 . CD63, a member of the transmembrane 4 superfamily, is a cell-surface protein often used as a marker for multivesicular bodies such as endosomes, lysosomes and exosomes 79,80 , and is interestingly only present in normal saliva but not MIH saliva. Studying exosome in saliva also would apply for MIH patients as proposed for periodontitis 81 , head and neck squamous cell carcinoma 82 and as liquid biopsy in cancer detection and therapy response prediction 83 .
Our study has limitations. Firstly, we cannot identify the cellular origin of the proteins specifically observed in the MIH saliva nor can we ascribe if the MIH specific protein signature is a cause or consequence of the disease. Hypothesizing, it is plausible that the saliva of MIH patients changes before the disease onset and is caused by an altered physicochemical environment including flow rates and pH-buffering capacities consequent to an accumulation and/or activation of neutrophils that perpetuate MIH-pathogenesis. Another limitation is that we have focused on proteins that are exclusively present in MIH saliva. Differentially expressed proteins would have refined the overall picture of MIH versus healthy patients' saliva. This kind of analyses is subject of our further investigations and will be reported at a later timepoint. Finally, we are aware of the limitations associated with the sample collection method used. Our study was based on the Salivette system 43 and there are possible variances when using passive drooling, paraffin gum and Salivette based collection methods 84 . There are also other systems available for saliva collection that may affect result outcome such as Pure·SAL from Oasis Diagnostics 41 or Sorbette from Becton Dickinson 85 . Indeed, there is the concern of selective absorption of proteins to cotton or cellulose-based collection devices such as eosinophil cationic protein as detection of this particular protein was reported to be higher in passive drool compared to saliva collected using these aforementioned methods 85 . Thus, comparability of our results to other studies might be partially affected not only by the intrinsic heterogeneity of individual clinical subjects but also by the saliva collection method utilized. Future work may employ the analysis of endogenous peptides from saliva isolated from the saliva prior to protein precipitation and extraction, enabling a deeper insight in processes involving biologically active peptides. Given the limitations mentioned above, the use of passive drool collected saliva would be an interesting aspect to consider in the future.
Our work paves the path for future studies which we suggest should focus on possible similarities on the MIH saliva proteins signature identified herein with more established chronic inflammatory diseases such as mucositis, periodontitis, pulpitis and also peri-implantitis. It is possible that in these diseases similar neutrophil related proteins are identified. The present research is also a primer for understanding the role of neutrophils in MIH -in particular, if the neutrophils are only a consequence of the disease and mainly reflect chronic inflammation -or possibly also contribute to its pathogenesis. Although, theoretically, local enhanced neutrophil activation might impair the mineralization of the ameloblasts it is unlikely that neutrophils can affect enamel mineralization. It will Scientific RepoRtS | (2020) 10:7560 | https://doi.org/10.1038/s41598-020-64614-z www.nature.com/scientificreports www.nature.com/scientificreports/ be further interesting to exploit novel proteogenomics-based bioinformatic tools and investigate to which extent the diversity of the oral microbiome is changed in MIH saliva versus control subjects.
In conclusion, among 618 salivary proteins identified with high confidence, 88 proteins were identified exclusively in MIH patients. Enrichment analysis revealed neutrophil-mediated adaptive immunity, the activation of the classical pathway of complement activation, extracellular matrix degradation, and other processes linked to inflammation. Whether or not this catabolic environment is a consequence of the inflammation that goes along with MIH warrants further analysis.