The extracellular matrix landscape in salivary gland carcinomas is defined by cellular differentiation via expression of three distinct protein modules.

The extracellular matrix (ECM) is an integral part of the tumor microenvironment of carcinomas. Even though salivary gland carcinomas (SGCs) display a range of tumor cell differentiation and distinct extracellular matrices, their ECM landscape has not been characterized in depth. The ECM composition of 89 SGC primaries, 14 metastases, and 25 normal salivary gland tissues was assessed using deep proteomic profiling. Machine learning algorithms and network analysis were used to detect tumor groups and protein modules that explain specific ECM landscapes. Multimodal in situ studies to validate exploratory findings and to infer a putative cellular origin of ECM components were applied. We revealed two fundamental SGC ECM classes which align with the presence or absence of myoepithelial tumor differentiation. We describe the SGC ECM through three biologically distinct protein modules that are differentially expressed across ECM classes and cell types. The modules have a distinct prognostic impact on different SGC types. Since targeted therapy is rarely available for SGC, we used the proteomic expression profile to identify putative therapeutic targets. In summary, we provide the first extensive inventory of ECM components in SGC, a difficult-to-treat disease that encompasses tumors with distinct cellular differentiation. © 2023 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

Salivary Pellicle Formed on Dental Composites Evaluated by Mass Spectrometry-An In Situ Study.

(1) Background: In the oral environment, sound enamel and dental restorative materials are immediately covered by a pellicle layer, which enables bacteria to attach. For the development of new materials with repellent surface functions, information on the formation and maturation of salivary pellicles is crucial. Therefore, the present in situ study aimed to investigate the proteomic profile of salivary pellicles formed on different dental composites. (2) Methods: Light-cured composite and bovine enamel samples (controls) were exposed to the oral cavity for 30, 90, and 120 min. All samples were subjected to optical and mechanical profilometry, as well as SEM surface evaluation. Acquired pellicles and unstimulated whole saliva samples were analyzed by SELDI-TOF-MS. The significance was determined by the generalized estimation equation and the post-hoc bonferroni adjustment. (3) Results: SEM revealed the formation of homogeneous pellicles on all test and control surfaces. Profilometry showed that composite surfaces tend to be of higher roughness compared to enamel. SELDI-TOF-MS detected up to 102 different proteins in the saliva samples and up to 46 proteins in the pellicle. Significant differences among 14 pellicle proteins were found between the composite materials and the controls. (4) Conclusions: Pellicle formation was material- and time-dependent. Proteins differed among the composites and to the control.

High-resolution MRI of the human palatine tonsil and its schematic anatomic 3D reconstruction.

The palatine tonsils form an important part of the human immune system. Together with the other lymphoid tonsils of Waldeyer's tonsillar ring, they act as the first line of defense against ingested or inhaled pathogens. Although histologically stained sections of the palatine tonsil are widely available, they represent the tissue only in two dimensions and do not provide reference to three-dimensional space. Such a representation of a tonsillar specimen based on imaging data as a 3D anatomical reconstruction is lacking both in scientific publications and especially in textbooks. As a first step in this direction, the objective of the present work was to image a resected tonsil specimen with high spatial resolution in a 9.4 T small-bore pre-clinical MRI and to combine these data with data from the completely sectioned and H&E stained same palatine tonsil. Based on the information from both image modalities, a 3D anatomical sketch was drawn by a scientific graphic artist. In perspective, such studies could help to overcome the difficulty of capturing the spatial extent and arrangement of anatomical structures from 2D images and to establish a link between three-dimensional anatomical preparations and two-dimensional sections or illustrations, as they have been found so far in common textbooks and anatomical atlases.

CD138 Is Expressed in Different Entities of Salivary Gland Cancer and Their Lymph Node Metastases and Therefore Represents a Potential Therapeutic Target.

Advanced salivary gland carcinomas (SGC) often lack therapeutic options. Agents targeting CD138 have recently shown promising results in clinical trials for multiple myeloma and a preclinical trial for triple-negative breast cancer. Immunohistochemistry for CD138 was performed for all patients who had undergone primary surgery for SGC with curative intent. Findings were validated using matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) imaging. Overall, 111 primary SGC and 13 lymph node metastases from salivary duct carcinomas (SaDu) were evaluated. CD138 expression was found in 60% of all SGC with differing expression across entities (p < 0.01). A mean of 25.2% of the tumor cells in mucoepidermoid carcinoma (MuEp) were positive, followed by epithelial-myoepithelial carcinoma (20.9%), acinic cell carcinoma (16.0%), and SaDu (15.2%). High-/intermediate-grade MuEp showed CD138 expression in a mean of 34.8% of tumor cells. For SaDu, lymph node metastases showed CD138 expression in a mean of 31.2% of tumor cells which correlated with CD138 expression in their primaries (p = 0.01; Spearman's ρ = 0.71). MALDI-MS imaging confirmed the presence of the CD138 protein in SGC. No significant association was found between clinicopathological data, including progression-free survival (p = 0.50) and CD138 expression. CD138 is expressed in the cell membrane of different entities of SGC and SaDu lymph node metastases and therefore represents a potential target for CD138 targeting drugs.

Tissue-resident macrophages mediate neutrophil recruitment and kidney injury in shiga toxin-induced hemolytic uremic syndrome.

Enterohaemorrhagic E. coli cause major epidemics worldwide with significant organ damage and very high percentages of death. Due to the ability of enterohaemorrhagic E. coli to produce shiga toxin these bacteria damage the kidney leading to the hemolytic uremic syndrome. A therapy against this serious kidney disease has not been developed yet and the impact and mechanism of leukocyte activation and recruitment are unclear. Tissue-resident macrophages represent the main leukocyte population in the healthy kidney, but the role of this important cell population in shiga toxin-producing E. coli-hemolytic uremic syndrome is incompletely understood. Using state of the art microscopy and mass spectrometry imaging, our preclinical study demonstrated a phenotypic and functional switch of tissue-resident macrophages after disease induction in mice. Kidney macrophages produced the inflammatory molecule TNFα and depletion of tissue-resident macrophages via the CSF1 receptor abolished TNFα levels in the kidney and significantly diminished disease severity. Furthermore, macrophage depletion did not only attenuate endothelial damage and thrombocytopenia, but also activation of thrombocytes and neutrophils. Moreover, we observed that neutrophils infiltrated the kidney cortex and depletion of macrophages significantly reduced the recruitment of neutrophils and expression of the neutrophil-attracting chemokines CXCL1 and CXCL2. Intravital microscopy revealed that inhibition of CXCR2, the receptor for CXCL1 and CXCL2, significantly reduced the infiltration of neutrophils and reduced kidney injury. Thus, our study shows activation of tissue-resident macrophages during shiga toxin-producing E. coli-hemolytic uremic syndrome leading to the production of disease-promoting TNFα and CXCR2-dependent recruitment of neutrophils.

Microdissection-An Essential Prerequisite for Spatial Cancer Omics.

The problem with cancer tissue is that its intratumoral heterogeneity and its complexity is extremely high as cells possess, depending on their location and function, different mutations, different mRNA expression and the highest intricacy in the protein pattern. Prior to genomic and proteomic analyses, it is therefore indispensable to identify the exact part of the tissue or even the exact cell. Laser-based microdissection is a tried and tested technique able to produce pure and well-defined cell material for further analysis with proteomic and genomic techniques. It sheds light on the heterogeneity of cancer or other complex diseases and enables the identification of biomarkers. This review aims to raise awareness for the reconsideration of laser-based microdissection and seeks to present current state-of-the-art combinations with omic techniques.

Integration of 3D multimodal imaging data of a head and neck cancer and advanced feature recognition.

In the last years, matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI) became an imaging technique which has the potential to characterize complex tumor tissue. The combination with other modalities and with standard histology techniques was achieved by the use of image registration methods and enhances analysis possibilities. We analyzed an oral squamous cell carcinoma with up to 162 consecutive sections with MALDI MSI, hematoxylin and eosin (H&E) staining and immunohistochemistry (IHC) against CD31. Spatial segmentation maps of the MALDI MSI data were generated by similarity-based clustering of spectra. Next, the maps were overlaid with the H&E microscopy images and the results were interpreted by an experienced pathologist. Image registration was used to fuse both modalities and to build a three-dimensional (3D) model. To visualize structures below resolution of MALDI MSI, IHC was carried out for CD31 and results were embedded additionally. The integration of 3D MALDI MSI data with H&E and IHC images allows a correlation between histological and molecular information leading to a better understanding of the functional heterogeneity of tumors. This article is part of a Special Issue entitled: MALDI Imaging, edited by Dr. Corinna Henkel and Prof. Peter Hoffmann.

Multigrid MALDI mass spectrometry imaging (mMALDI MSI).

Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI) is an important technique for the spatially resolved molecular analysis of tissue sections. The selection of matrices influences the resulting mass spectra to a high degree. For extensive and simultaneous analysis, the application of different matrices to one tissue sample is desirable. To date, only a single matrix could be applied to a tissue section per experiment. However, repetitive removal of the matrix makes this approach time-consuming and damaging to tissue samples. To overcome these drawbacks, we developed a multigrid MALDI MSI technique (mMALDI MSI) that relies on automated inkjet printing to place differing matrices onto predefined dot grids. We used a cooled printhead to prevent cavitation of low viscosity solvents in the printhead nozzle. Improved spatial resolution of the dot grids was achieved by using a triple-pulse procedure that reduced droplet volume. The matrices can either be applied directly to the thaw-mounted tissue sample or by precoating the slide followed by mounting of the tissue sample. During the MALDI imaging process, we were able to precisely target different matrix point grids with the laser to simultaneously produce distinct mass spectra. Unlike the standard method, the prespotting approach optimizes the spectra quality, avoids analyte delocalization, and enables subsequent hematoxylin and eosin (H&E) staining. Graphical Abstract Scheme of the pre-spotted multigrid MALDI MSI workflow.

Novel workflow for combining Raman spectroscopy and MALDI-MSI for tissue based studies.

Molecular heterogeneity of cancer is a major obstacle in tumor diagnosis and treatment. To deal with this heterogeneity, a multidisciplinary combination of different analysis techniques is of urgent need because a combination enables the creation of a multimodal image of a tumor. Here, we develop a computational workflow in order to combine matrix-assisted laser desorption/ionization mass spectrometric (MALDI-MS) imaging and Raman microspectroscopic imaging for tissue based studies. The computational workflow can be used to confirm a spectral histopathology (SHP) based on one technique with another technique. In this contribution, we confirmed a Raman spectroscopic based SHP with MALDI-imaging. Owing to this combination, we could demonstrate, for a larynx carcinoma sample, that tissue types and different metabolic states could be extracted from the Raman spectra. Further investigations with the help of MALDI spectra yield a better characterization of variable epithelial differentiation and a better understanding of ongoing dysplastic alterations.

Transthyretin is dysregulated in preeclampsia, and its native form prevents the onset of disease in a preclinical mouse model.

Preeclampsia is a major pregnancy complication with potential short- and long-term consequences for both mother and fetus. Understanding its pathogenesis and causative biomarkers is likely to yield insights for prediction and treatment. Herein, we provide evidence that transthyretin, a transporter of thyroxine and retinol, is aggregated in preeclampsia and is present at reduced levels in sera of preeclamptic women, as detected by proteomic screen. We demonstrate that transthyretin aggregates form deposits in preeclampsia placental tissue and cause apoptosis. By using in vitro approaches and a humanized mouse model, we provide evidence for a causal link between dysregulated transthyretin and preeclampsia. Native transthyretin inhibits all preeclampsia-like features in the humanized mouse model, including new-onset proteinuria, increased blood pressure, glomerular endotheliosis, and production of anti-angiogenic factors. Our findings suggest that a focus on transthyretin structure and function is a novel strategy to understand and combat preeclampsia.

Urine protein profiling identified alpha-1-microglobulin and haptoglobin as biomarkers for early diagnosis of acute allograft rejection following kidney transplantation.

PURPOSE: Early diagnosis of acute rejection and effective immunosuppressive therapy lead to improvement in graft survival following kidney transplantation. In this study, we aimed to establish a urinary protein profile suitable to distinguish between patients with rejection and stable graft function and to predict acute rejection based on postoperatively collected urine samples. A further objective was to identify candidate proteins for the use as biomarkers in clinical practice. METHODS: Urine samples of 116 kidney recipients were included. Rejection was proven by biopsy (n = 58), and stable transplant function was monitored for at least 2 years (n = 58). Postoperative urine samples were collected between 3rd and 10th day following transplantation. Urinary protein profiles were obtained by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry. Protein identification and validation were performed using multiplex fluorescence 2DE, peptide mass fingerprinting and enzyme-linked immunosorbent assay. RESULTS: A protein profile including four mass peaks differentiated acute rejection from stable transplants at the time point of rejection and at the postoperative state with 73 % sensitivity and 88 % specificity. Alpha-1-microglobulin (A1MG) and Haptoglobin (Hp) were identified as putative rejection biomarkers. Protein levels were significantly higher in postoperative urine from patients with rejection (A1MG 29.13 vs. 22.06 µg/ml, p = 0.001; Hp 628.34 vs. 248.57 ng/ml, p = 0.003). The combination of both proteins enabled the diagnosis of early rejection with 85 % sensitivity and 80 % specificity. CONCLUSION: Protein profiling using mass spectrometry is suitable for noninvasive detection of rejection-specific changes following kidney transplantation. A specific protein profile enables the prediction of early acute allograft rejection in the immediate postoperative period. A1MG and Hp appear to be reliable rejection biomarkers.

Hyperspectral imaging and evaluation of surgical margins: where do we stand?

PURPOSE OF REVIEW: To highlight the recent literature on the use of hyperspectral imaging (HSI) for cancer margin evaluation ex vivo, for head and neck cancer pathology and in vivo during head and neck cancer surgery. RECENT FINDINGS: HSI can be used ex vivo on unstained and stained tissue sections to analyze head and neck tissue and tumor cells in combination with machine learning approaches to analyze head and neck cancer cell characteristics and to discriminate the tumor border from normal tissue. Data on in vivo applications during head and neck cancer surgery are preliminary and limited. Even now an accuracy of 80% for tumor versus nonneoplastic tissue classification can be achieved for certain tasks, within the current in vivo settings. SUMMARY: Significant progress has been made to introduce HSI for ex vivo head and neck cancer pathology evaluation and for an intraoperative use to define the tumor margins. To optimize the accuracy for in vivo use, larger HSI databases with annotations for head and neck cancer are needed.

Disruption of ALX1 causes extreme microphthalmia and severe facial clefting: expanding the spectrum of autosomal-recessive ALX-related frontonasal dysplasia.

We present an autosomal-recessive frontonasal dysplasia (FND) characterized by bilateral extreme microphthalmia, bilateral oblique facial cleft, complete cleft palate, hypertelorism, wide nasal bridge with hypoplasia of the ala nasi, and low-set, posteriorly rotated ears in two distinct families. Using Affymetrix 250K SNP array genotyping and homozygosity mapping, we mapped this clinical entity to chromosome 12q21. In one of the families, three siblings were affected, and CNV analysis of the critical region showed a homozygous 3.7 Mb deletion containing the ALX1 (CART1) gene, which encodes the aristaless-like homeobox 1 transcription factor. In the second family we identified a homozygous donor-splice-site mutation (c.531+1G > A) in the ALX1 gene, providing evidence that complete loss of function of ALX1 protein causes severe disruption of early craniofacial development. Unlike loss of its murine ortholog, loss of human ALX1 does not result in neural-tube defects; however, it does severely affect the orchestrated fusion between frontonasal, nasomedial, nasolateral, and maxillary processes during early-stage embryogenesis. This study further expands the spectrum of the recently recognized autosomal-recessive ALX-related FND phenotype in humans.

Analysis and interpretation of imaging mass spectrometry data by clustering mass-to-charge images according to their spatial similarity.

Imaging mass spectrometry (imaging MS) has emerged in the past decade as a label-free, spatially resolved, and multipurpose bioanalytical technique for direct analysis of biological samples from animal tissue, plant tissue, biofilms, and polymer films. Imaging MS has been successfully incorporated into many biomedical pipelines where it is usually applied in the so-called untargeted mode-capturing spatial localization of a multitude of ions from a wide mass range.3 An imaging MS data set usually comprises thousands of spectra and tens to hundreds of thousands of mass-to-charge (m/z) images and can be as large as several gigabytes. Unsupervised analysis of an imaging MS data set aims at finding hidden structures in the data with no a priori information used and is often exploited as the first step of imaging MS data analysis. We propose a novel, easy-to-use and easy-to-implement approach to answer one of the key questions of unsupervised analysis of imaging MS data: what do all m/z images look like? The key idea of the approach is to cluster all m/z images according to their spatial similarity so that each cluster contains spatially similar m/z images. We propose a visualization of both spatial and spectral information obtained using clustering that provides an easy way to understand what all m/z images look like. We evaluated the proposed approach on matrix-assisted laser desorption ionization imaging MS data sets of a rat brain coronal section and human larynx carcinoma and discussed several scenarios of data analysis.

Urine screening by Seldi-Tof, followed by biomarker identification, in a Brazilian cohort of patients with renal cell carcinoma (RCC).

PURPOSE: To screen proteins/peptides in urine of Renal Cell Carcinoma (RCC) patients by SELDI-TOF (Surface Enhanced Laser Desorption Ionization - Time of Flight) in search of possible biomarkers. MATERIAL AND METHODS: Sixty-one urines samples from Clear Cell RCC and Papillary RCC were compared to 29 samples of control urine on CM10 chip. Mass analysis was performed in a ProteinChip Reader PCS 4,000 (Ciphergen Biosystems, Fremont, CA) with the software Ciphergen Express 3.0. All chips were read at low and at high laser energy. For statistical analysis the urine samples were clustered according to the histological classification (Clear Cell and Papillary Carcinoma). For identification urine was loaded on a SDS PAGE gel and bands of most interest were excised, trypsinized and identified by MS/MS. Databank searches were performed in Swiss-Prot database using the MASCOT search algorithm and in Profound. RESULTS: Proteins that were identified from urine of controls included immunoglobulin light chains, albumin, secreted and transmembrane 1 precursor (protein K12), mannan-binding lectin-associated serine protease-2 (MASP-2) and vitelline membrane outer layer 1 isoform 1. Identification of immunoglobulins and isoforms of albumin are quite common by proteomics and therefore cannot be considered as possible molecular markers. K12 and MASP-2 play important physiological roles, while vitellite membrane outer layer 1 role is unknown since it was never purified in humans. CONCLUSIONS: The down expression of Protein K-12 and MASP-2 make them good candidates for RCC urine marker and should be validated in a bigger cohort including the other less common histological RCC subtypes.

Differential vascular expression and regulation of oncofetal tenascin-C and fibronectin variants in renal cell carcinoma (RCC): implications for an individualized angiogenesis-related targeted drug delivery.

The study was aimed at determining the vascular expression of oncofetal fibronectin (oncfFn) and tenascin-C (oncfTn-C) isoforms in renal cell carcinoma (RCC) and its metastases which are well-known targets for antibody-based pharmacodelivery. Furthermore, the influence of tumour cells on endothelial mRNA expression of these molecules was investigated. Evaluation of vascular ED-A(+) and ED-B(+) Fn as well as A1(+) and C(+) Tn-C was performed after immunofluorescence double and triple staining using human recombinant antibodies on clear cell, papillary and chromophobe primary RCC and metastases. The influence of hypoxic RCC-conditioned medium on oncfFn and oncfTn-C mRNA expression was examined in human umbilical vein endothelial cells (HUVEC) by real time RT-PCR. There are RCC subtype specific expression profiles of vascular oncfFn and oncfTn-C and corresponding patterns when comparing primary tumours and metastases. Within one tumour, there are different vessel populations with regard to the incorporation of oncfTn-C and oncfFn into the vessel wall. In vitro tumour-derived soluble mediators induce an up regulation of oncfTn-C and oncfFn mRNA in HUVEC which can be blocked by Avastin(®). Vascular expression of oncFn and oncTn-C variants depends on RCC subtype and may reflect an individual tumour stroma interaction or different stages of vessel development. Therefore, oncFn or oncTn-C variants can be suggested as molecular targets for individualized antibody based therapy strategies in RCC. Tumour-derived VEGF could be shown to regulate target expression.

A novel multiplex-protein array for serum diagnostics of colon cancer: a case-control study.

BACKGROUND: More than 1.2 million new cases of colorectal cancer are reported each year worldwide. Despite actual screening programs, about 50% of the patients are diagnosed at advanced tumor stages presenting poor prognosis. Innovative screening tools could aid the detection at early stages and allow curative treatment interventions. METHODS: A nine target multiplex serum protein biochip was generated and evaluated using a training- and validation-set of 317 highly standardized, liquid nitrogen preserved serum samples comprising controls, adenomas, and colon cancers. RESULTS: Serum levels of CEA, IL-8, VEGF, S100A11, MCSF, C3adesArg, CD26, and CRP showed significant differences between cases and controls. The largest areas under the receiver operating characteristics curve were observed for CEA, IL-8, and CRP. At threshold levels yielding 90% specificity, sensitivities for CEA, IL-8 and CRP were 26%, 22%, and 17%, respectively. The most promising marker combinations were CEA + IL-8 reaching 37% sensitivity at 83% specificity and CEA + CRP with 35% sensitivity at 81% specificity. In an independent validation set CEA + IL-8 reached 47% sensitivity at 86% specificity while CEA + CRP obtained 39% sensitivity at 86% specificity. Early carcinomas were detected with 33% sensitivity for CEA + IL-8 and 28% for CEA + CRP. CONCLUSIONS: Apart from CEA, IL-8, and CRP, the screening value of additional blood markers and the potential advantage of combining serum biochip testing with fecal occult blood testing needs to be studied. Multiplex biochip array technology utilizing serum samples offers an innovative approach to colorectal cancer screening.

Homozygous CFTR mutation M348K in a boy with respiratory symptoms and failure to thrive. Disease-causing mutation or benign alteration?

UNLABELLED: We report on a 6-month-old premature boy from consanguineous parents. He presented with respiratory distress, necrotizing enterocolitis and hyperbilirubinemia shortly after birth. Persisting respiratory symptoms and failure to thrive prompted cystic fibrosis diagnostics, which showed the lack of wild-type signal for the mutation R347P suggesting a homozygous deletion or an alteration different from the known mutation at this position. Sequencing of this region revealed the homozygous substitution 1175 T > A (HGVS: c.1043 T > A) in exon 7 resulting in the homozygous amino acid change M348K. This mutation has never been reported in homozygosity before. Computational analysis tools classified M348K as 'presumably disease causing.' In our patient, sweat testing and electrophysiological assessment of CFTR function in native rectal epithelium demonstrated normal Cl(-) secretion. CONCLUSION: We assume that the homozygous alteration M348K is a harmless variant rather than a CF-causing mutation.

Mutually Exclusive Expression of COL11A1 by CAFs and Tumour Cells in a Large panCancer and a Salivary Gland Carcinoma Cohort.

Procollagen 11A1 (COL11A1) is a central component of the extracellular matrix in many carcinomas, which is considered to be mainly produced by cancer associated fibroblasts (CAFs). As COL11A1 expression correlates with adverse prognosis and is implicated in chemoresistance, it is a promising putative target. For the first time, we used RNA in-situ hybridization to systematically identify the cells that produce COL11A1 in the ten most prevalent carcinoma types, lymphomas (n = 275) and corresponding normal tissue (n = 55; panCancer cohort). Moreover, as most salivary gland carcinomas (SGC) display distinct stromal architectures, we also analysed 110 SGC. The corresponding protein formation of COL11A1 was determined by MALDI-TOF-MS-Imaging. We report that colon, breast and salivary duct carcinomas are highly infiltrated by COL11A1 positive CAFs (CAFsCOL11A1) and might thus be promising candidates for antidesmoplastic or COL11A1-targeted therapies. The amount of CAFsCOL11A1 correlated significantly with tumour grade, tumour stage and nodal spread in the panCancer cohort. Significant associations between CAFsCOL11A1 and vascular invasion, perineural spread and nodal spread were observed in the SGC cohort. Also, we discovered that tumour cells of intercalated duct derived SGC and CAFs produce COL11A1 in a mutually exclusive manner. Our findings represent a novel mode of extracellular matrix production in carcinomas and could be highly relevant in the future. Our findings elucidate the mode of COL11A1 expression in very different carcinoma types and may aid to categorise tumours in the setting of possible future COL11A1-related therapies.

Multi-Class Cancer Subtyping in Salivary Gland Carcinomas with MALDI Imaging and Deep Learning.

Salivary gland carcinomas (SGC) are a heterogeneous group of tumors. The prognosis varies strongly according to its type, and even the distinction between benign and malign tumor is challenging. Adenoid cystic carcinoma (AdCy) is one subgroup of SGCs that is prone to late metastasis. This makes accurate tumor subtyping an important task. Matrix-assisted laser desorption/ionization (MALDI) imaging is a label-free technique capable of providing spatially resolved information about the abundance of biomolecules according to their mass-to-charge ratio. We analyzed tissue micro arrays (TMAs) of 25 patients (including six different SGC subtypes and a healthy control group of six patients) with high mass resolution MALDI imaging using a 12-Tesla magnetic resonance mass spectrometer. The high mass resolution allowed us to accurately detect single masses, with strong contributions to each class prediction. To address the added complexity created by the high mass resolution and multiple classes, we propose a deep-learning model. We showed that our deep-learning model provides a per-class classification accuracy of greater than 80% with little preprocessing. Based on this classification, we employed methods of explainable artificial intelligence (AI) to gain further insights into the spectrometric features of AdCys.