**Gastrointestinal Cancers and Personalized Medicine**

Editors **Stefania Nobili Enrico Mini**

MDPI ' Basel ' Beijing ' Wuhan ' Barcelona ' Belgrade ' Manchester ' Tokyo ' Cluj ' Tianjin

*Editors* Stefania Nobili Department of Neurosciences, Imaging and Clinical Sciences "G. d'Annunzio" University of Chieti-Pescara Chieti Italy Enrico Mini Department of Health Sciences University of Florence Florence Italy

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Journal of Personalized Medicine* (ISSN 2075-4426) (available at: www.mdpi.com/journal/jpm/special issues/gastrointestinal medicine).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Volume Number*, Page Range.

**ISBN 978-3-0365-3808-2 (Hbk) ISBN 978-3-0365-3807-5 (PDF)**

© 2022 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**


## **About the Editors**

#### **Stefania Nobili**

Dr. Nobili is an Assistant Professor of Pharmacology at the Department of Neurosciences, Imaging and Clinical Sciences of the "G. d'Annunzio"University of Chieti-Pescara (Italy). She achieved her Pharm.D. degree at the University of Rome 'La Sapienza'in 1989, her Specialization in Applied Pharmacology at the University of Florence in 1997, and her Ph.D. in Chemotherapy at the University of Milan in 2001. In 2013, she was awarded the Alberico Benedicenti Award for Pharmacology and Toxicology from the Italian Society of Pharmacology. From 2007 to 2019, she worked as an Assistant Professor of Pharmacology at the University of Florence (Italy). Her research interests are focused on the preclinical and clinical pharmacology of anticancer drugs, in particular, cancer pharmacogenetics and pharmacogenomics and tumor drug resistance, with an emphasis on the drugs used in the treatment of gastrointestinal neoplasms.

#### **Enrico Mini**

Enrico Mini is a Professor of Medical Oncology at the School of Human Health Sciences and the Director of the Specialty School of Medical Oncology, University of Florence (Italy). He is the Director of the Clinical Program on 'Gastrointestinal Cancer Therapeutics'at the Section of Chemotherapy, Careggi University Hospital, Florence. He is also Head of the Laboratory of Anticancer Chemotherapy at the Department of Health Sciences, University of Florence.

Dr. Mini achieved his M.D. in 1977 and his Ph.D. in 1988 at the University of Florence. He completed his post-graduate training at Yale University School of Medicine (Departments of Pharmacology and Medicine) in the USA. He later became a Lecturer in Pharmacology at the Universities of Siena and Ferrara in Italy. He became an Associate Professor of Pharmacology in the Department of Pharmacology, School of Medicine at the University of Florence in 1992, and Professor of Pharmacology at the School of Medicine (University of Florence) in 2001. In recognition of his achievements, Dr. Mini received the Leukemia Society of America Special Fellow Award in 1982, the Lady Tata Memorial Trust Award and the Alberico Benedicenti Italian Society of Pharmacology Special Mention Award in 1988. Currently, he is a member of the Steering Committee of the European Organization for Research and Treatment of Cancer (EORTC), Pan-European Trials in Adjuvant Colon Cancer (PETACC), and is the Chairman of the Cancer Section of the International Society of Chemotherapy. He is the co-editor of the {Journal of Chemotherapy}. He was also a Visiting Professor at Yale in 2010.

Dr. Mini has authored more than 100 scientific publications. His research interests comprise the mechanisms of tumor drug resistance, in particular to antimetabolites and metal-based drugs, cancer pharmacogenomics and pharmacogenetics, and controlled clinical trials in gastrointestinal cancers.

### *Editorial* **Special Issue: "Gastrointestinal Cancers and Personalized Medicine"**

**Stefania Nobili 1,2, \* and Enrico Mini 3,4**


Gastrointestinal cancers represent more than 25% of all diagnosed cancers and more than 36% of cancer-related deaths worldwide [1]. Unfortunately, screening strategies are still limited. They are, in fact, available only for colorectal, gastric and esophageal cancers [2]. However, despite these early diagnostic opportunities, gastrointestinal cancers, including pancreatic, hepatobiliary, small bowel carcinomas and other uncommon cancers, such as anal canal cancer, neuroendocrine tumors of the gastrointestinal tract, primary gastric and intestinal lymphomas, and gastrointestinal stromal tumors (GISTs), are frequently diagnosed at an advanced stage, when treatment options are limited and cure is not possible. Moreover, a very high percentage of patients, about 50%, diagnosed with early potentially curable gastrointestinal cancers, will develop recurrent disease despite surgery, radiation therapy, and pharmacological treatment during the course of the disease [3]. Taken together, these conditions are responsible for poor prognosis in these tumors. Thus, despite current available knowledge on molecular determinants involved in the initiation and progression of cancer [4], there is an urgent clinical need to further improve our biological knowledge of gastrointestinal evolutionary processes toward increased dysregulation, heterogeneity, and the escape from immunosurveillance as well as from pharmacological treatment control [5,6].

Such complex processes substantially involve all types of molecules (e.g., nucleic acids, proteins, metabolites) and involve several cell types, such as transformed epithelial or mesenchymal cells, or other tumor microenvironment cells, including immune cells.

The development and availability of the newest biotechnologies that add knowledge to the field of cancer research strongly contribute to new cancer achievements aimed at discovering and validating novel molecular biomarkers predictive of prognosis and drug response (efficacy/toxicity) in gastrointestinal cancers. A number of cancer biomarkers, mainly represented by somatic alterations in tumor cells (e.g., in the *RAS*, *RAF*, *MMR*, *HER-2* and *KIT* genes), have been identified and validated as clinically useful biomarkers to predict patient prognosis and drug response in gastrointestinal cancers such as colorectal cancer, gastric cancer and GISTs, thus directly contributing to therapeutic decisions.

However, due to the high level of tumor heterogeneity, not only among patients but also among tumor sites in the same patient, the possibility of employing effective personalized medicine for all patients still represents a relevant challenge. Currently, a plethora of potential biomarkers predictive of prognosis or drug response have been suggested [7]. Their detection in tissue and/or in bio-fluidic samples has the potential to improve clinical oncology practice.

In addition, a pharmacogenetic approach, represented by the analysis of germline polymorphisms in genes that play a main role in the ADME of anticancer drugs, has also been progressively introduced into clinical practice for the prediction of the risk of toxicity

**Citation:** Nobili, S.; Mini, E. Special Issue: "Gastrointestinal Cancers and Personalized Medicine". *J. Pers. Med.* **2022**, *12*, 338. https://doi.org/ 10.3390/jpm12030338

Received: 21 February 2022 Accepted: 22 February 2022 Published: 24 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

related to these drugs. However, few polymorphisms in pharmacogenes have been shown to be responsible for drug toxicity. Thus, this aspect still represents a major issue in cancer care.

This Special Issue is designed to provide information on new biomarker research in the area of gastrointestinal tumors that could be useful for innovative personalized management and precision medicine modalities for individualized care.

Gastrointestinal cancers are often diagnosed at advanced stages when therapeutic options are limited. Liquid biopsy, a non-invasive procedure, widely investigated in recent years and already applied to monitor cancer progression and drug resistance, mainly in lung cancer, could also be successfully used to diagnose cancers at early stages. Liquid biopsy, usually performed in blood serum, could be obtained by several different body fluids [8]. Among gastrointestinal cancers, pancreatic cancer could greatly benefit from this opportunity. In fact, although it is only the 12th most common cancer, it is the 6th most common cause of cancer death [1]. Thus, it would be crucial to identify a strategy able to diagnose pancreatic cancer in advance, before tumor development. The incidence of pancreatic cysts is about 2% in adults and neoplastic cysts account for 10–15% of all pancreatic cystic lesions. Although their risk to change in malignant lesions is low, if this occurs, the patient prognosis will be very poor [9].

Hermoso-Durán et al. [10] investigated cyst liquid samples from patients, and the proteomic differences between pancreatic benign and premalignant cysts. To perform such an evaluation the authors used an approach previously used on serum or plasma, named "thermal liquid biopsy" (TLB), which they adapted to cyst liquid samples. Based on the TLB thermograms, cyst profiles were clustered according to their clinical assessment. The authors also elaborated a new TLB serum score based on the specific parameters reflecting differences between cysts. Results were encouraging although the number of analyzed samples was small. The availability of a dedicated TLB as a diagnostic tool for serum samples from patients with pancreatic cysts of which the nature is unknown could represent a relevant advantage in the diagnosis of premalignant lesions of pancreatic cancer.

The prognosis of patients affected by colorectal cancer, one of the most incident and lethal cancers worldwide [1], is highly variable, mainly dependent on the stage at diagnosis. Through the years, many efforts have been made to identify and validate molecular biomarkers predictive of prognosis and/or drug response in this neoplasm. In recent years, interesting examples of predictors of prognosis concern the colorectal cancer molecular subtypes that have been obtained by unsupervised transcriptomic approaches, i.e., consensus colorectal cancer molecular subtypes (CMSs) [11] and colorectal cancer intrinsic subtypes (CRIS) [12]. However, the clinical utility of such classifications for the single patient has yet to be established. Potential biomarkers predictive of response to adjuvant chemotherapy in the early colorectal cancer stages (i.e., stage II-III) have been suggested, for instance from validated transcriptomic [13–15] or genetic [16] analyses. However, to date, biomarkers predictive of drug response represented by actionable oncogenic drivers (i.e., *RAS* wild-type and MSI-H status for anti-EGFR and anti-PD-1 monoclonal antibodies, *BRAF* V600E mutations, *NTRK* gene fusions and more recently *KRAS* G12C mutations for targeted agents) are used only in the metastatic setting [17].

In this framework, the review of Del Buono et al. [18] contextualized the role of the DNA mismatch repair (MMR) system in colorectal cancer precision medicine. Today, the knowledge of the MSI status provides several advantages by satisfying a number of clinical queries. In fact, MSI, due to an impaired MMR system, plays a role in the inherited predisposition to gastrointestinal cancers, and identifies a subset of colorectal cancer patients who show a substantial better prognosis and who do not obtain an advantage from adjuvant chemotherapy (i.e., low-risk stage II patients). More recently, MSI has become a key biomarker for the treatment of several tumors, including colorectal cancer, with immune checkpoint inhibitors. Thus, the evaluation of MMR/MSI is becoming part of standard care in colorectal cancer, as recommended by major oncological international societies [17].

Overall, therapeutic options in colorectal cancer are related to the cancer stage and, as mentioned above, differ from metastatic and nonmetastatic settings. Stage I and low-risk stage II patients are treated with surgery alone. High-risk stage II, stage III and stage IV (oligometastatic disease) patients are instead treated with pharmacological therapy in addition to surgery, with positive results. However, neoplastic progression due to additional dysregulated molecular events occurs in a substantial percentage of patients, limiting the efficacy of the available drugs administered as adjuvant or neoadjuvant therapies. This occurrence stimulates the search for biomarkers able to predict colorectal cancer prognosis in order to plan preventative pharmacological strategies for patients at high risk of disease progression as well as biomarkers predictive of drug response, in order to avoid the administration of inactive drugs to resistant patients. Immunoscore is a further example of tumor biomarker able to predict disease prognosis in early-stage colorectal cancer [19]. Instead, tumor mutational burden is not yet a recommended biomarker for the prediction of pembrolizumab efficacy in colorectal cancer due to the limited data available in this patient population [17].

The study of Rhyner Agocs et al. [20] evaluated the predictive role of the expression of the lymphocyte-activation gene 3 (LAG-3) in the outcome of 143 stage II colon cancer. LAG-3 is an inhibitory immune-related molecule mainly expressed on T cells, but also on B cells and dendritic cells. LAG-3 may synergize with the PD-1/PD-L1 pathway and is closely related to CD4. The upregulation of LAG-3 on immune cells downregulates T cell expansion and cytokine secretion, and thus contributes to an immunosuppressive microenvironment. In particular, the presence of LAG-3 was evaluated by immunohistochemistry in formalin-fixed paraffin-embedded (FFPE) tissues on tumor-infiltrating lymphocytes (TILs) in the tumor center and tumor front to assess its impact on the survival of stage II colon cancer patients. The authors found no correlations between LAG-3 expression and clinical/pathological characteristics, although they observed a higher percentage of MMR-deficient colon cancers when LAG-3-positive TILS were present. In relation to the primary study end-point, i.e., disease-free survival, the authors found a significant association between the presence of LAG-3 in the tumor front and prolonged disease-free survival. This significant correlation was maintained even when only MMR-proficient colon cancer, (i.e., the majority of the analyzed tumors), were considered. Moreover, in this case, such a correlation was limited to TILs localized at the tumor front. Thus, this manuscript identified LAG-3 as a biomarker potentially useful in predicting patient prognosis in stage II colon cancer, including MMR-proficient tumors.

Similarly, Peyravian et al. [21] analyzed a panel of candidate genes (i.e., 20 genes) whose expression was potentially involved in the development of lymph node metastases in 100 colorectal cancer patients. The selected genes were chosen according to their role in key cancer processes such as carcinogenesis, tumor growth, tumor invasion and metastasis. Overall, about 60% of patients initially diagnosed as stage I-III, were lymph nodes negative. Hierarchical clustering analysis showed that *VANGL1*, *PCSK7*, and *ANXA3* genes were the most expressed among the study genes at mRNA level in the majority of colorectal cancer samples. However, only *VANGL1* was shown to significantly vary between lymph node-negative and -positive patients. The mRNA expression levels of *VANGL1* were also confirmed at protein level. The study also provided associations between two other study genes, *NOTCH1* and *ILR2B,* and overall survival. In particular, the high expression of *NOTCH1* and the low expression of *ILR2B* were associated with prolonged overall survival.

In metastatic colorectal cancer, Taghizadeh et al. [22] provided a molecular profile of a real-world cohort of drug refractory patients for whom no further standard treatment option was available. The molecular profile was performed by a precision medicine platform developed at the author's institution, i.e., the Comprehensive Cancer Centre of the Medical University of Vienna. Based on the biomolecular characteristics of tumors, this study was aimed at providing information on potential further options of targeted therapy. Overall, by exploiting next-generation sequencing panels of mutation hotspots, microsatellite instability testing, and immunohistochemistry, 60 metastatic colorectal cancer

samples were characterized. The analysis revealed 166 mutations in 53 patients, the five most frequent being *TP53*, *KRAS*, *APC*, *PIK3CA*, and *PTEN*. All patients had previously received cytotoxic chemotherapy combined with anti-EGFR or anti-VEGF(R) monoclonal antibodies. The study showed that, in 47% of patients, a molecularly targeted therapy could be recommended whereas the remaining were not suitable for targeted therapy due to the lack of actionable molecular targets. Overall, 20% of the study patients underwent the recommended targeted therapy. In particular, pembrolizumab was offered to four MSI-H patients, consequently obtaining control of disease in all patients and objective response in 75%. Stable disease was observed in two further patients treated with everolimus combined with raltitrexed, and with trastuzumab combined with lapatinib, respectively, according to their specific immunohistochemical and mutational characteristics (i.e., strong m-TOR expression associated with the loss of PTEN and HER2+ overexpression, respectively). Overall, this study highlights how at least a portion of heavily pretreated patients without further standard treatment options may benefit from a molecular-based treatment approach.

Interestingly, by a rationale based on the role that the immune response and inflammation play in tumor growth and in the metastatic process, Fülöp et al. [23] evaluated the prognostic impact of the neutrophil-to-lymphocyte and lymphocyte-to-monocyte ratios (i.e., NLR and LMR) in over 1000 rectal cancer patients. The overall survival was significantly associated with increased NLR and decreased LMR, and no relationship was found between the study ratios and tumor stage, thus potentially suggesting that these markers are independent from cancer stage, even if this occurrence is controversial [24]. Moreover, NLR and LMR were also found to predict response to the neoadjuvant chemoradiotherapy to which patients underwent. In particular, the identification of a cut-off for NLR value (i.e., ≥3.11) allowed the authors to discriminate between chemoradiotherapy responsive and non-responsive rectal cancer patients, although the responsive ones had a low chance of sphincter preservation, or to obtain a complete total mesorectal excision. Although the study ratios may also be affected by factors independent of the neoplastic disease, these data warrant attention due to the high number of patients included in this analysis and deserve further investigation.

Oxaliplatin, widely used in the treatment of gastrointestinal cancers, is a highly neurotoxic agent. Acute or chronic peripheral neuropathy develops in about 90% and 40% of patients, respectively, and the latter form may strongly affect the quality of life of patients and cancer survivors. Unfortunately, to date, no remedy or antidote is available to reverse this side effect. Thus, it would be very important to identify patients susceptible to develop peripheral neuropathy before starting the oxaliplatin treatment, even though, despite the efforts of many researchers, no predictive biomarker has yet been identified.

The review of Velasco et al. [25] discusses the status of the art of strategies that may be implemented pre-emptively to evaluate the risk of developing neurotoxicity. In particular, neurological monitoring through the evaluation of neurophysiological signs of oxaliplatininduced neuropathy may be performed by mechanical strategies (e.g., nerve conduction tests, electromyography). However, this procedure is not part of the common clinical practice. Less invasive blood biomarkers have also been widely investigated. Genetic biomarkers, mainly represented by single-nucleotide polymorphisms in genes encoding detoxification enzymes (e.g., proteins belonging to the glutathione detoxification system), drug transporters (e.g., ATP binding proteins), proteins involved in the mechanism of action of oxaliplatin, as well as proteins implicated in neuronal functions, have drawn attention. In addition, proteins released in blood when nerve damage occurs (e.g., the protein neurofilament light chain (NfL) and nerve growth factor (NGF)) have also been suggested as predictive biomarkers of neurotoxicity. Neuroimaging strategies have also been studied as potential tools for the early detection of neurotoxicity onset.

Overall, the manuscripts included in this Special Issue highlight the need to identify and validate molecular biomarkers predictive of prognosis and drug response in gastrointestinal cancers. To satisfy this goal, biomarkers identified in retrospective studies will need to be validated in large-scale prospective clinical trials. Moreover, the availability of

new and highly predictive biomarkers implies that the discovery of new anticancer drugs, specifically inhibiting these targets, can be accomplished to effectively treat patients who are potentially unresponsive to standard therapies.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Thermal Liquid Biopsy (TLB) Focused on Benign and Premalignant Pancreatic Cyst Diagnosis**

**Sonia Hermoso-Durán 1,2,† , Guillermo García-Rayado 1,3,4,† , Laura Ceballos-Laita 1,2 , Carlos Sostres 1,3,4 , Sonia Vega <sup>2</sup> , Judith Millastre 1,3 , Oscar Sánchez-Gracia 5 , Jorge L. Ojeda 6 , Ángel Lanas 1,3,4,7 , Adrián Velázquez-Campoy 1,2,4,8,9, \* and Olga Abian 1,2,4,8,10, \***


**\*** Correspondence: adrianvc@unizar.es (A.V.-C.); oabifra@unizar.es (O.A.); Tel.: +34-976-762996 (A.V.-C.); +34-876-555417 (O.A.)

† S.H.-D. and G.G.-R. contributed equally to this work and are both first authors of the manuscript.

**Abstract:** Background: Current efforts in the identification of new biomarkers are directed towards an accurate differentiation between benign and premalignant cysts. Thermal Liquid Biopsy (TLB) has been previously applied to inflammatory and tumor diseases and could offer an interesting point of view in this type of pathology. Methods: In this work, twenty patients (12 males and 8 females, average ages 62) diagnosed with a pancreatic cyst benign (10) and premalignant (10) cyst lesions were recruited, and biological samples were obtained during the endoscopic ultrasonography procedure. Results: Proteomic content of cyst liquid samples was studied and several common proteins in the different groups were identified. TLB cyst liquid profiles reflected protein content. Also, TLB serum score was able to discriminate between healthy and cysts patients (71% sensitivity and 98% specificity) and between benign and premalignant cysts (75% sensitivity and 67% specificity). Conclusions: TLB analysis of plasmatic serum sample, a quick, simple and non-invasive technique that can be easily implemented, reports valuable information on the observed pancreatic lesion. These preliminary results set the basis for a larger study to refine TLB serum score and move closer to the clinical application of TLB providing useful information to the gastroenterologist during patient diagnosis.

**Keywords:** pancreatic cysts; thermal liquid biopsy; differential scanning calorimetry; diagnosis; generalized linear models

#### **1. Introduction**

During recent years, the detection of pancreatic cysts has become more frequent due to improvements in abdominal imaging techniques. The incidence of this pathology is approximately 2% in the adult population [1]. Computed tomography (CT) scans are reported detection between 1.2% and 2.6%, and magnetic resonance imaging (MRI) has even a higher detection capability, ranging between 13.5% and 19.9% [1,2]. The management of

**Citation:** Hermoso-Durán, S.; García-Rayado, G.; Ceballos-Laita, L.; Sostres, C.; Vega, S.; Millastre, J.; Sánchez-Gracia, O.; Ojeda, J.L.; Lanas, Á.; Velázquez-Campoy, A.; et al. Thermal Liquid Biopsy (TLB) Focused on Benign and Premalignant Pancreatic Cyst Diagnosis. *J. Pers. Med.* **2021**, *11*, 25. https://doi.org/ 10.3390/jpm11010025

Received: 9 December 2020 Accepted: 29 December 2020 Published: 31 December 2020

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

these incidentally detected pancreatic cysts is still a challenge, because, even though the risk of being malignant is low, the prognosis in case of pancreatic adenocarcinoma and intraductal papillary mucinous neoplasms- (IPMN)-related pancreatic adenocarcinoma is very poor and has not improved recently [3]. Thus, distinguishing between benign and malignant cysts is difficult, and very often requires surgical intervention with considerable morbidity and mortality. Since 2005, some guidelines have been published and updated: the American Society for Gastrointestinal Endoscopy [4], international consensus guidelines by the International Association of Pancreatology (Sendai guidelines) [5], the American College of Gastroenterology [6], and the International Association of Pancreatology (Fukuoka guidelines) [7,8]. More recently, The European Study Group on Cystic Tumors of the Pancreas published an update, replacing the 2013 European Consensus Statement Guidelines [9].

From a clinical point of view, the classification according to the prognosis of the lesion is: (1) cysts with malignant potential (mucinous), (2) cysts without malignant potential, and (3) malignancies. However, most studies provide a description of biomarkers for differentiating the mucinous and non-mucinous type of cysts. The non-mucinous group comprises serous cystadenomas (SCAs) (the most common type), pancreatic pseudocysts (PCs), and a variety of rare cysts (benign epithelial, lymphoepithelial, congenital, and squamoid cysts). Most are found incidentally and none of them represents a risk for becoming malignant [10]. On the contrary, the mucinous group, including mucinous cystic neoplasms (MCNs) and IPMNs, constitutes the majority of neoplastic premalignant cysts identified in the pancreas, and a precise diagnosis technique would be vital for the management of patients [11]. The development of techniques with greater pre-surgical diagnostic precision would make it possible to avoid the morbidity and mortality associated with a high-risk intervention when it is not strictly necessary, as well as reduce the healthcare overload derived from unnecessary outpatient follow-up in cystic lesions without the potential for malignancy.

Cystic fluid markers become especially relevant when transabdominal ultrasonography, CT or MRI are inconclusive. In these cases, it is necessary to employ another risk predictor to indicate surgery as the most appropriate treatment given the estimated risk.

The presence of amylase at high concentration in cyst fluid indicates that there is a communication between the cyst and the ductal system. This occurs in both pseudocysts and IPMN lesions. When amylase levels are lower than 250 U/L, communication with the conduct can be discarded, with a specificity of 98% [12]. However, amylase value alone is not enough to differentiate between mucinous/non-mucinous or MCN/IPMN, which is important from the point of view of patient management, in deciding whether surgical resection or cyst time-monitoring is recommended [13]. A previous episode of pancreatitis can be of help in distinguishing a pseudocyst from an IPMN lesion (occasionally related to pancreatitis) [14].

Carcinoembryonic antigen (CEA) is also employed as a biomarker. This is a set of highly related glycoproteins involved in cell adhesion, mucin being one of them. It can be used to distinguish between cysts with (MCNs and IPMNs) and without (SCAs and PCs) mucinous epithelium [10]. The major inconvenience of CEA is the absence of an appropriate cutoff value [12]. The current accepted value to classify a cyst as mucinous is CEA > 192 ng/mL [15].

In order to improve diagnosis, studies related to the identification of new biomarkers in cystic fluid or serum have been recently reported [16–21].

In 2007 Chaires et al. [22] described the application of differential scanning calorimetry (DSC) in diagnosis using plasma/serum samples from cancer patients. Since then, many studies have confirmed the potential clinical use of this technique, not only applied to plasma/serum [23–30], but also to other biological human samples such as cerebrospinal fluid [31,32]. In 2018 our group coined the name "thermal liquid biopsy" (TLB) for DSC applied to cancer diagnosis and cancer patient's treatment monitoring [33,34]. The TLB thermogram reports the global denaturation profile for all the proteins present in the serum/plasma sample and the influence of potential interactions between blood plasma proteins and metabolites, therefore reflecting any alteration induced by a certain disease. As in the case of plasma or serum, cystic fluid is also composed of a mixture of proteins and TLB may also be applied as a clinical diagnosis tool. In this work, the potential of TLB as a clinical biomarker for cyst classification has been pursued. In this pilot study, 20 cyst fluid samples were analyzed and their TLB cyst profiles were obtained, with the purpose of finding a correlation between the TLB thermogram and the type of cyst. The proteomic analysis also allowed a description of the more abundant proteins in the cyst fluid, as well as the post-translational modifications present in those proteins. TLB was also applied to serum samples in some of the patients, and TLB serum profile differences between groups was studied to try to determine whether or not differences observed in cyst fluid TLB correlated with serum TLB. This would be extremely important because, in case cyst features are reflected in certain serum alterations, a simple and risk-free plasma/serum TLB analysis could be employed for cyst diagnosis/classification. Despite the low number of samples considered in this pilot study, different patterns in cyst fluid and plasma from patients with pathology could be observed.

#### **2. Materials and Methods**

#### *2.1. Subjects and Samples*

Cyst liquid and serum samples from patients with cystic lesions in the pancreas detected by transabdominal ultrasonography or CT or MRI were referred to the Department of Digestive Endoscopy at the Hospital Clínico Universitario Lozano Blesa (HCULB), Zaragoza, Spain, between January 2016 and September 2018. The procedure was in accordance with the recommendations of the local ethics committee and all patients gave their informed consent. Pancreatic cystic fluids were collected by EUS-guided fine needle aspiration (FNA). The EUS-FNA procedure was performed with an Olympus® 140 curvilinear echo-endoscope. Boston ScientificTM Expect® 19 or 22-gauge needles were used depending on the cystic endosonographic features. We noted the characteristics of the aspirated cystic fluid: volume, color, and viscosity. The majority of the fluid was examined by the same cytopathologist for every patient and the rest (at least 1 mL Eppendorf for each patient) was collected for detection of biochemical markers, DSC measurements, and proteomic studies described in this manuscript. The collected cystic fluid samples were then stored at −80 ◦C until they were prepared for analysis.

Serum samples from healthy subjects as control group (HC) consisted of 85 serum samples from Spanish Caucasian subjects, apparently cancer-free, from the FISABIO (Fundación para el Fomento de la Investigacion Sanitaria y Biomedica de la Comunitat Valenciana) biobank with a homogeneous distribution, including gender (53% men and 47% women), with an average age of 45.2 ± 14.2.

#### *2.2. Thermal Liquid Biopsy (TLB) Profile Determination*

DSC thermograms were measured using a high-sensitivity differential scanning VP-DSC microcalorimeter (MicroCal, Malvern-Panalytical, Malvern, UK). Cystic liquid samples, serum samples, and reference solutions were properly degassed and carefully loaded into the cells to avoid bubble formation. The baseline of the instrument was routinely recorded before the experiments. Experiments were performed in cystic liquid samples (diluted 1:10 in phosphate buffered saline, PBS) and serum samples (diluted 1:25 in PBS) at a scanning rate of 1 ◦C/min. Thermograms were baseline-corrected and analyzed using software developed in our laboratory implemented in Origin 7 (OriginLab, Northampton, MA, USA).

#### *2.3. Data Analysis*

We have developed a phenomenological model in which the TLB serum thermogram is deconvoluted into several individual transitions, modeling each individual transition by the logistic peak or Hubbert function [30,33]. This model has been successfully applied in the analysis of serum samples from melanoma and gastric and lung cancer patients [30,34,35].

From this multiparametric analysis, a TLB serum score (between 0 and 1) can be calculated reporting the level of alterations in plasma (TLB serum score < 0.5, absence of alterations; TLB serum score > 0.5, presence of alterations).

The Kolmogorov-Smirnov test was performed to assess the normal distribution of the variables. Medians between two independent groups were compared with the Wilcoxon test, in non-normal distributions. Averages between two independent groups were compared with the *t*-test, in normal distributions.

#### *2.4. Protein Sample Preparation and Protein Identification and Quantification by Mass Spectrometry*

Protein concentration: Measured by Bradford protein assay (Bio-Rad, Madrid, Spain) using purified bovine serum albumin (BSA) (10 mg/mL, New England BioLabs, EVRY cedex, France) in PBS as standard. Absorbance at 595 nm of two dilutions from each serum sample was measured in triplicate in a Synergy HT multimode microplate reader (BioTek Instruments, Winooski, VT, USA).

In solution digestion: Samples were evaporated and resuspended in 10 µL of denaturing buffer (6 M urea, 100 mM Tris buffer pH 7.8). Next, cysteines were reduced with 1.5 µL DTT (200 mM) for 30 min at 37 ◦C and alkylated with 6 µL of iodoacetamide (200 mM) for 30 min in the dark. Unreacted iodoacetamide was consumed adding 6 µL of the reducing agent (200 mM DTT) for 30 min at room temperature. Samples were diluted with 50 mM ammonium bicarbonate to a urea final concentration lower than 1 M. Trypsin digestion (Gold Trypsin, Promega, Madison, WI, USA) was carried out overnight at 37 ◦C at a 1:20 enzyme/protein ratio. Reaction was stopped adding concentrated formic acid (Merck KGaA, Darmstadt, Germany). Samples were evaporated, resuspended in 2% acetonitrile (ACN), 0.1% formic acid, and filtered through 0.45 µm filters.

Protein identification by LC-ESI-MS/MS: Protein identification was performed on a nano-LC 2D system (LC 425, Eksigent Ekspert TM, Dublin, CA, USA) coupled to a hybrid triple quadrupole/linear ion trap mass spectrometer (4000 QTRAP, Sciex, Foster City, CA, USA). On-line pre-concentration and desalting of samples was performed using a C18 trap cartridge (Luna® 0.3 mm id, 20 mm, 5 µm particle size, Phenomenex, CA, USA) at 10 µL/min for 5 min. Peptide separation was performed using a C18 column (Gemini® 0.3 mm id, 150 mm, 3 µm particle size, Phenomenex, CA, USA), at 5 µL/min of flow rate. Column was maintained at 35 ◦C. The elution gradient was from 5 to 35% ACN (0.1% formic acid) in 90 min. The mass spectrometer was interfaced with an ESI source (Turbo V™) using a 25 µm ID hybrid electrode and was operated in the positive ion mode. MS source parameters were as follows: capillary voltage 5000 V, de-clustering potential (DP) 85 V and curtain and ion source gas (Nitrogen) 15 psi. Analyses were performed using an information dependent acquisition (IDA) method with the following steps: single enhanced mass spectra (EMS, 400–1400 *m*/*z*) from which the 5 most intense peaks were subjected to an enhanced product ion [EPI (MS/MS)] scan. Protein identification was carried out using the Mascot search engine (Matrix Science; London, UK) and the non-redundant SwissProt database (553,655 sequences; 198,177,566 residues). Search parameters were monoisotopic mass accuracy, peptide mass tolerance ±0.5 Da, fragment mass tolerance ±0.3 Da; one allowed missed cleavage; allowed fixed modification carbamido-methylation (Cys), and variable modification oxidation (Met). Positive identification was assigned with Mascot scores above the threshold level (*p* < 0.05), with at least two identified peptides with a score above homology

Protein SDS electrophoresis: Samples mixed with NuPAGE LDS Sample buffer (Invitrogen), and heated at 95 ◦C for 4 min, were analysed by sodium dodecyl sulphate– polyacrylamide gel electrophoresis (SDS–PAGE) using 10% acrylamide resolving gels and 4% acrylamide stacking gels (Bio-Rad). The gels were fixed with a mixture of ethanol, acetic acid, and deionized water (40:10:50) for 1 h. After washing in water for 5 min, the gels were stained with Coomassie Brilliant Blue R250 (0.1% in 25% methanol, 10% acetic acid) and de-stained by incubation in 30% acetic acid and 20% methanol. Molecular weights were estimated by comparison with the migration rates of standard proteins (Bio-Rad).

#### **3. Results**

#### *3.1. Clinical Sample Description*

Patients who underwent endoscopic ultrasonography procedure were included in this work. A total of 20 subjects, 60 and 40% men and women, respectively, with an average age of 62 ± 13 years.

Based on imaging and cytopathology, the pancreatic cysts were classified into different categories (Table 1).


**Table 1.** Patient Description.

\* Average ± standard deviation (sd) PC = pseudocyst; WOPN = Walled-off pancreatic necrosis; IPMN = intraductal papillary mucinous neoplasm; SC= Serous Cyst; MCN= Mucinous Cystadenoma; LYM= lymphocele; PDAC= Pancreatic Ductal Adenocarcinoma.

> Clinical information of the samples is detailed in Table 2. All the cysts were between 2 and 15 cm in size and they were located in any region in the pancreas. According to clinical data (amylase and CEA concentrations), samples were divided in two groups: benign cysts (PC, WOPN and SC) and premalignant cysts (IPMN and MCN). There are two samples that turned out not to be cysts, but malignant lesions (PDAC).


**Table 2.** Clinical Cyst Sample Description.

nd = not determined: Amylase < 250 U/L, communication with the conduct can be discarded; CEA > 192 ng/mL to classify a cyst as mucinous.

> Pancreatic pseudocysts (PC) are pockets of fluid, common sequelae of acute pancreatitis or chronic pancreatitis. PCs are important in terms of management and differentiation from other cystic processes or masses in this region. According to the updated Atlanta

classification [36], there are two main groups of mature-well defined fluid collections associated with acute pancreatitis: A/Fluid collections in interstitial edematous pancreatitis (PC), and B/Fluid collections in necrotizing pancreatitis (WOPN). Both PC and WOPN were considered benign cysts. From our PC samples, PC 1 and 4 were in the context of acute pancreatitis, and PC 5 was in the context of chronic pancreatitis. The WOPN cysts had the biggest size, between 3 and 15 cm. Both types of pancreatic collections (PC and WOPN) were amylase positive (above 250 U/L) and CEA negative (below 192 ng/mL).

Serous cysts (SC) are benign neoplasms composed of numerous small cysts that are arrayed in a honeycomb-like formation and most individual cysts are typically <10 mm.

Lymphocele (LYM), also known as cystic lymphangioma, is a rare disease. There are no typical clinical manifestations, and most patients were diagnosed incidentally during imaging or surgery. Therefore, diagnosis is challenging. Surgical resection is still considered as the most effective approach for lymphocele, and prognosis is favorable. In our study, SC and Lym were 5 cm in size, and amylase and CEA negative.

Intraductal papillary mucinous neoplasms (IPMN) are epithelial pancreatic cystic tumors of mucin-producing cells that arise from the pancreatic ducts. They are most commonly seen in elderly patients, with sex distribution roughly balanced, a possible slight male predominance. IPMNs are slow growing tumors that have malignant potential and distinct variants have been described: main duct (IPMN 6 and 7), branch duct (IPMN 1, 3 and 4), and mixed branch and main duct (IPMN 2 and 5). Main duct IPMNs have a very high rate of malignancy (up to 70% in reported surgical series [8]); for this reason, the usual recommendation is surgical removal of the affected portion of the pancreas. Branch duct IPMNs are cystic neoplasms of the pancreas that have malignant potential and their management is challenging; the risk of surgery must be carefully weighed against the risk of malignancy when deciding on surgical removal or surveillance. This is the reason why great efforts are taken to distinguish mucinous cysts from other cyst lesions (specially, main duct IPMNs). All IPMNs are considered as premalignant cysts. They had the smallest size, between 2 and 3.5 cm. Four were amylase positive (above 250 U/L) and all were CEA positive (above 192 ng/mL or very closed in case of IPMN3 with 156 ng/mL).

Mucinous Cystadenoma (MCN) is another type of mucinous cystic neoplasm of the pancreas, traditionally considered typical of middle age females. MCN1 was 3 cm in size, amylase and CEA positive.

#### *3.2. Analysis of TLB from Cystic Liquid Samples*

TLB thermograms of 20 cystic fluid samples were obtained. Protein concentrations and dilutions could be considered, but in this case TLB curves were normalized according to their area under the curve values (AUC); therefore, signals from the different samples can be compared and uncertainties in protein concentration (inherent to colorimetric methods) are avoided. TLB cyst profiles clustered according to their clinical assessment (benign or premalignant nature) are represented in Figure 1.

Regarding the benign cystic group, the WOPN group exhibited a very similar cyst thermogram profile with two peaks at 65 and 82 ◦C. We can easily distinguish this group from the other benign cysts (Figure 2A). PC5 is the only PC lacking the 85 ◦C peak, and it is the only PC in a chronic pancreatitis context.

In the premalignant cyst group, branch duct IPMNs (IPMN1, IPMN3 and IPMN4) exhibited a similar profile (Figure 2C). Main duct IPMNs (IPMN6 and IPMN7) exhibited a single peak.

**Figure 1.** Individual TLB thermograms from cystic liquid samples. Samples were clustered according to their benign or premalignant nature.

#### *3.3. Analysis of Proteomic Signatures from Cystic Liquid Samples*

Proteins identified by LC-ESI-MS/MS (detailed in Table S1) were analyzed and clustered according to the benign or premalignant nature of the cyst (Figure 3). This first classification distinguishes 52 proteins common in the cyst groups, and 12 and 11 proteins present only in benign cysts and premalignant cysts, respectively. A deeper analysis according to different groups of cysts was performed.

#### 3.3.1. Benign Cysts

The WOPN cyst group exhibited a homogeneous proteomic profile (Table S1). 18 out of 41 (44%) proteins were shared by all cysts in this group (Figure 4A). The similarity in these samples was even higher, because 10 more proteins were common in WOPN1 and WOPN3. Low protein concentration in WOPN2 (Figure S1A) could prevent proper identification of more proteins in that sample. The most abundant proteins found in this group were globulins (macroglobulin and immunoglobulins), a type of protein related to immunological response as a consequence of an inflammation process. This is consistent with the nature of this specific type of cyst: walled-off pancreatic necrosis (WOPN) is a well-circumscribed area of necrosis which occurs as a late complication of acute pancreatitis, generally after four weeks from the initial episode. Singular proteins detected in WOPN1 and WOPN3 samples were S-100 proteins. They belong to the S100 protein family, having important roles in inflammation and may also be useful markers for gut inflammation [37]. Once secreted in the extracellular space, S100A9 acts as a chemo-attractant, recruiting

further inflammatory cells and creating an inflammatory microenvironment that promotes tumor development [38].

**Figure 2.** TLB thermograms from cystic liquid samples clustered according to their type. Average curves (colored lines) and standard deviations of curve values (grey) are represented: WOPNs (**A**), PCs (**B**), IPMNs (**C**) and IPMN7/MCNs (**D**).

**Figure 3.** LC-ESI-MS/MS proteomic content of cystic samples. Common proteins were analyzed via Venn diagrams online tool (http://bioinformatics.psb.ugent.be/beg/tools/venn-diagrams). 12 and 11 proteins were detected only in benign cysts (blue) and premalignant cysts (pink), respectively, and 52 proteins appeared in both groups (intersection set). Table comprises the detailed information of the proteins in each set.


**Figure 4.** LC-ESI-MS/MS proteomic content of cystic samples according to cystic types: (**A**) WOPN, (**B**) PC, (**C**) IPMN, and (**D**) MCN+IPMN. Common proteins were analyzed via Venn diagrams online tool. Colors code for different groups and numbers inside each set and shared sub-sets indicate the number of identified proteins.

Common proteins of this group detected by electrophoresis (Figure S1A) are: proteolytic proteins, albumin, glycoside hydrolases, metalloproteases, immunoglobulins, and elastases. This agreed with the proteomic profile previously detailed (Table S1).

The homogeneity found in the proteomic signature of this group was reflected in the TLB cyst profiles of the cystic liquid (Figure 2A). As could be observed, compared to the rest of cystic groups studied in this work, these TLB cyst profiles represent a particular signature for WOPN cysts, easily distinguishing this group from the rest. They showed two peaks at T1max ≈ 65 ◦C and T2max ≈ 81 ◦C, with a peak width at around 10 ◦C. Denatured proteins included in these peaks are detailed in Figure 4A.

In the PC cyst group, proteomic profiles data was also collected (Table S1), except for PC1, for which no protein was detected, except albumin. There are 5 proteins shared by all PCs (Figure 4B) and 10 proteins are shared by at least three PCs: PC2, PC3 and PC5 had 6 common proteins and PC3, PC4 and PC5 4 proteins. These cysts were negative for globulins (macroglobulin and immunoglobulins), indicating there were not any inflammatory processes going on (except for PC4, the rest of PCs in this subgroup were not in an acute pancreatitis context). The PC4 cyst, in acute pancreatitis context, exhibited 17 proteins not found in any other PCs (Table S1). It also contained a small number of carboxypeptidase or pancreatic elastase related proteins, absence of pancreatic triacylglycerol lipase and trypsin-1 and presence of globulins (macroglobulin and immunoglobulins), serotransferrin and lacto-transferrin, protein S100-A9, neutrophil defensin 1, and myeloperoxidase.

Common proteins of this group were detected in electrophoresis (Figure S1B) (except for PC1, where the total amount of protein was very low): glycoside hydrolases and metalloproteases (all PCs), proteolytic proteins (PC3 and PC4), albumin (all but PC2), lipases (all but PC5), immunoglobulins (PC3 and PC4) and elastases (PC2 and PC3). This was consistent with the proteomic profiles.

TLBs of PC cystic liquid (Figure 2B) were similar for PC1, PC2 and PC3, with three peaks at T1max ≈ 55 ◦C, T2max ≈ 65 ◦C (two close peaks) and T3max ≈ 85 ◦C (peak width around 10 ◦C). Common proteins were found in these samples: amylase, IgGA, carboxypeptidase pancreatic elastase, chymo-trypsinogen, trypsin, glycoprotein GP2 and phospholipase A2. In PC5, T3max ≈ 85 ◦C was missing, and this could be the result of the absence of any of these proteins: IgGA, glycoprotein GP2, phospholipase A2 and chymo-trypsinogen. In PC4 there was a shift in the transitions, from 55 to 40 ◦C and from 85 to 75 ◦C (peak width was maintained in both cases).

In this group the clinical explanation for these differences could lie in the pancreatitis context for samples PC4 (acute) or PC5 (chronic).

#### 3.3.2. Premalignant Cysts

The IPMN cyst group was more homogeneous from the clinical point of view. IPMN was the cystic group in this study with a number of proteins identified in the proteomic profile (Table S1). For example, it was not possible to identify proteins for IPMN5. One interesting observation was that none of the IPMNs contained globulins (macroglobulin or immunoglobulins), or they were negligible in other cyst groups, and only IPMN7 clearly contained them. These proteins are related to immunological responses as a consequence of inflammation and, therefore, it seemed that inflammation was not associated with this type of cyst.

Branch duct IPMNs (IPMN1, IPMN3 and IPMN4) shared 28 proteins with main duct IPMNs (IPMN2, IPMN6 and IPMN7) (Figure 4C). Branch duct IPMNs shared seven common proteins, and main duct IPMNs shared only two proteins. Neither of these proteins were unique in any of the two groups (they were included in the 28 proteins in common).

Common proteins of this group detected in the electrophoresis gel are albumin and immunoglobulins (IPMN1, IPMN2 and IPMN5); glycoside hydrolases and elastases (IPMN1, IPMN3, IPMN4 and IPMN5); metalloproteases (IPMN1 and IPMN3); and lipases (IPMN3

and IPMN4). Proteolytic proteins were not present in any of IPMNs samples (Figure S1D). This was consistent with the proteomic profiles.

TLBs of IPMN cystic liquid (Figure 2C) were similar for IPMN1, IPMN3 and IPMN4, all being branch duct IPMNs. They exhibited two close peaks (around T1max ≈ 55 ◦C, T2max ≈ 65 ◦C) and a third peak at T3max ≈ 85 ◦C (peak width around 10 ◦C). These peaks could correspond to the common proteins found in this group (amylases, carboxypeptidase and pancreatic elastases).

TLB of main duct type IPMN cystic liquids had a wider single peak at ≈55 ◦C and ≈75 ◦C for IPMN6 or IPMN7, respectively.

TLB of mixed type IPMN cystic liquids presented characteristics from both branch or duct IPMN groups with a low signal peak at ≈55 ◦C and one peak at ≈80 ◦C for IPMN5, or two peaks at ≈85 ◦C and 90 ◦C for IPMN2.

There was no proteomic profile available for IPMN5, but, according to electrophoresis, there was no protein from the amylase family, and additionally IPMN 2 and IPMN 7 lacked amylases, carboxy-peptidase and pancreatic elastases. These three cysts did not exhibit any peak (or it was neglectable) at ≈55 ◦C.

The MCN 1 cyst was in the mixed cysts group, but it had a premalignant nature.

MCN1 shared 11 proteins with IPMNs, either branch or main duct type, (Figure 4D) and 19 more proteins in common only with main duct IPMNs (Figure 4D) with five of these protein shared with IPMN7. When comparing MCN1 and IPMN7 profiles (Table S1), 15 common proteins were found. In fact, IPMN7 was the only IPMN cyst showing globulins (macroglobulin and immunoglobulins). In addition, protein profile in electrophoresis for MCN1 was similar to IPMNs (Figure S1E). However, 17 proteins were detected in IPMNs, but not in MCN1. There were no proteins shared only with branch duct IPMNs.

When comparing MCN1 and IPMN7 profiles (Table S1), 15 common proteins were found. In fact, IPMN7 was the only IPMN cyst showing globulins (macroglobulin and immunoglobulins).

Proteins detected in electrophoresis for MCN1 were similar to IPMNs (Figure S1E).

TLB for cystic liquid sample exhibited a wide transition (around 30 ◦C width) with two peaks (around T1max ≈ 70 ◦C, T2max ≈ 85 ◦C), very similar to the IPMN7 profile (Figure 2D). MCN1 did not exhibit any peak at around 55 ◦C and, again, no amylases, carboxypeptidase or pancreatic elastases were detected by proteomics.

Mixed Cyst Group was a completely heterogenous group comprising a serous cyst, a lymphocele, and two samples of pancreatic ductal adenocarcinoma (PDCA) (which cannot be considered as a cyst).

SC and LYM: The main difference between serous cyst and lymphocele (cystic benign samples), compared to the rest of samples, was the presence of apolipoprotein A-I (only MCN1 and IPMN7 seemed to contain it) (Table S1). In addition, they were positive for globulins (macroglobulin and immunoglobulins), as with the samples with an active inflammatory process. Some typical pancreatic cystic proteins were missing: amylase, metalloproteases, lipases, elastases. Some of these proteins were already employed as clinical biomarkers for pancreatic cyst diagnosis [9], such as amylase.

Common proteins of this group detected in electrophoresis are: serotransferrin, albumin, lipoproteins and immunoglobulins (Figure S1C). This agreed with the proteomic profiles.

All TLB cyst profiles in this group showed particular features and looked different from the other cystic profiles. They exhibited one single peak: SC at T1max ≈ 70 ◦C (peak width around 20 ◦C) and LYM at T1max ≈ 75 ◦C (peak width around 25 ◦C). Common proteins (Table S1) are albumin, lipoproteins, hemoglobins, globulins, and serotransferrin.

PDACs: Proteomic profiles of PDACs (Figure S2) showed 20 shared proteins from hemoglobin, immune-globulins and transferrin groups. PDAC1 also contained other proteins, and PDAC2 contained proteins we previously identified in cysts, such as amylases, metalloproteases, lipases, elastases.

TLB cyst profile in PDAC1 exhibited one peak (T1max ≈ 70 ◦C) and PDAC2 showed two peaks (T1max ≈ 65 ◦C and T2max ≈ 80 ◦C) (Figure 1). This difference could be related to the different proteomic profile between the two samples mentioned above. Electrophoresis also confirmed this different pattern (Figure S1F).

#### *3.4. Analysis of TLB from Serum Samples*

TLB thermograms from serum samples were obtained. The goal was to search for any potential reflection of the cystic pathology in plasmatic serum. TLB serum profiles were normalized according to their area under the curve values (AUC), again avoiding protein concentration influence. Then, they were clustered according to the clinical assessment of the cyst (Figure 5). We focused our attention on distinguishing between premalignant and benign cysts. Unfortunately, it was not possible to obtain serum samples at the same time as the eco-endoscopy procedure for all the patients included in this study.

PCs are benign lesions found in the context of acute or chronic pancreatitis. It has been previously reported that inflammatory processes can be reflected in TLB serum profile [23]. According to the PC TLB serum thermograms, they seemed to be somewhat different (Figure 5A).

IPMNs, being premalignant lesions, could also exhibit some distinctive features compared to healthy patients in their TLB serum profile on the basis of the previous studies on TLB applied to cancer diagnosis [30,33,34]. Apparently, in the case of IPMNs, TLB serum profiles, except for IPMN7, seem quite similar to healthy controls (Figure 5B).

**Figure 5.** Examples of serum TLB thermograms from different types of cystic patients, clustered according to cyst types: premalignant (**A**), benign (**B**) and non-cyst malignant (**C**). Typical TLB thermogram for a healthy subject is represented with black line in (**A**–**C**).

The multiparametric analysis previously developed in our group [30] was applied to TLB serum profiles. The purpose was to help in the identification of cystic disease-related TLB features and the quantification of the TLB serum score for patients' serum (Table S2). Mono-variant analysis of the individual TLB parameters obtained from the thermograms showed that 6 out of 15 parameters were statistically different (*p*-value < 0.05). Only these six parameters were used to construct the classification model and calculate the TLB serum score for each sample, as previously described in [33].

According to the results in Figure 6, the TLB serum score comparison between healthy subjects and cyst patients (both, benign and premalignant) indicated that the differences were statistically significant using the Wilcoxon test (*p*-value < 0.001). Similarly, TLB serum score could differentiate between healthy subjects and benign cyst patients (*p*-value < 0.001), and between healthy subjects and premalignant cyst patients (*p*-value < 0.001).

TLB serum score values are between 0 and 1: the closer to 0, the smaller the alterations in plasma (healthy status), while the closer to 1, the larger the alterations in plasma (diseased status). TLB serum score values were mainly under 0.5 for healthy (82 out of 84, 98% true negative rate) and over 0.5 for cysts patients (10 out 14, 71% true positive rate). This meant that this score may be useful for detecting the presence of cystic lesion. The area under the ROC curve is 0.94 (Figure S3) with sensitivity of 71%, specificity of 98%, a positive predictive value (PPV) of 83%, and a negative predictive value (NPV) of 98%. Unfortunately, there was no statistical difference between both cysts' groups; that is, TLB serum score did not discriminate between benign and premalignant cysts (*p*-value = 0.501).

**Figure 6.** TLB serum score from serum TLB profiles from healthy controls and both types of cysts (benign and premalignant). Median values were compared using the Wilcoxon test: \* *p*-valuehealthy-benign = 0.00026; \*\* *p*-valuehealthy-premalignant = 0.000057; \*\*\* *p*-valuehealthy-cysts = 0.0000012; ns *p*-valuebenign-premalignant = 0.501.

TLB serum values closest to 1 in PC corresponded to those associated to the acute or chronic pancreatitis context (inflammatory process can be reflected in serum).

TLB serum values of IPMN were around 0.5 (except in IPMN7). CEA value for IPMN7 was over 50,000 ng/mL, the highest of all premalignant cysts.

All types of cyst could be clustered differentially from healthy controls by using this single TLB serum score. These results agreed with our previously published results on lung cancer disease [33] in which TLB serum score was able to discriminate between diseased and healthy subjects.

Cystic pathologies are local lesions for which a systemic reflection in blood might not be expected. However, if cystic pathology is accompanied by inflammation, blood alterations may be important, even for a benign lesion.

As a further development, in this study we also proposed to evaluate whether TLB serum score could provide useful information to gastroenterologists for the diagnosis before and besides endoscopic ultrasound procedures. Despite the small number of samples (14 serum samples from cyst patients), a restricted TLB serum score excluding healthy control subjects and considering benign cysts as control samples was performed. Benign and premalignant cysts were 43% and 57% of the samples, respectively. First, we performed a mono-variant analysis of TLB parameters (Table S3) and none of the individual parameters was statistically different (*p*-values > 0.05). Therefore, only parameters presenting *p*-value below 0.25 were considered in constructing the classification model and calculating this new TLB serum score (benign vs. premalignant cysts). For a TLB score threshold of 0.5, 4 out 6 (67%) benign cysts had a TLB score below 0.5, and 6 out 8 (75%) premalignant cysts had a TLB score above 0.5. The area under the ROC curve is 0.875 (Figure 7B) with

sensitivity of 75%, specificity of 67%, a positive predictive value (PPV) of 75% and a negative predictive value (NPV) of 67%. When using the Youden index as a threshold (0.75), all benign samples were well classified (TLB score below 0.75).

**Figure 7.** (**A**)/TLB serum score parameters from serum TLB serum profiles from benign and premalignant cyst patients. **Figure 7.** (**A**)/TLB serum score parameters from serum TLB serum profiles from benign and premalignant cyst patients. (**B**)/ROC curve illustrating the statistical performance of TLB serum score (benign vs. premalignant cysts). AUC = Area Under Curve (95%CI).

#### **4. Discussion**

Pancreatic neoplasms are generally discovered incidentally and consists of IPMNs, predominantly [7]. The diagnosis of malignant IPMN lesions involves a certain degree of subjectivity and variability, which is undesirable for clinical practice, due to the lack of standardized guidelines. Thus, there is a necessity of new diagnosis tools to differentiate between benign and premalignant pancreatic cysts to help in making decisions about surgical intervention or periodic surveillance.

The analysis of cyst fluid may provide information regarding established biomarkers that helps the physician in the diagnosis [13,39]. This type of biological sample includes a diverse amount and type of proteins, representing an interesting challenge for thermal liquid biopsy. TLB for blood serum samples has been proven useful for different types of cancer, and premalignant pancreatic cysts will eventually evolve to pancreatic cancer. Additionally, TLB has been applied to other body fluids (serum, plasma, urine, synovial and cerebrospinal fluids [25,31,40,41], or even tumor digestion), provided that the TLB thermogram is a reflection of the protein composition, interactions, and modifications in the complex sample [32]. Similarities in protein composition of cysts (which is a known biomarker for cyst classification) will result in similarities in TLB cyst profiles, thus, providing the basis for a diagnostic procedure based on TLB. As has been previously described, TLB cyst profile reflects protein composition, but also protein interactions. The presence or absence of a certain protein could promote changes in the profile for other proteins in the sample, especially if they were interacting (interactome concept) [25]. Similar effects could be envisaged for protein modifications as a result of metabolic or pathologic processes.

In our study, we recruited 20 patients who, after being diagnosed with a cystic lesion by CT scan or MNR, underwent routine endoscopic ultrasonography where a sample of cyst fluid was obtained. The final clinical diagnosis of the lesions found was based

on the different clinical features of the patient (fluid aspect, biochemical biomarkers and pathological anatomy results) according to current clinical guidelines [8,36].

To our knowledge, this was the first time that pancreatic cyst fluid was characterized using a TLB technique and thermal profiles were clustered according to the clinical information on the cysts. We confirmed that cyst liquid thermograms reflect protein content in the samples. There was a high intra-group variability within the TLB cyst thermograms. Only WOPN cysts could be easily differentiated from PCs, because there was a clear profile type that could be associated with them (Figure 2A) and they also shared 18 proteins (Figure 4A). More samples are required to clearly define a common cyst thermogram for benign or premalignant cysts. Having the TLB thermogram available for specialist appraisal, perhaps the diagnosis could be oriented to the MCN cyst type. As we have already discussed, there is a considerable subjective component in the physician diagnosis, and the more complementary tools available for a better discrimination, the better for the gastroenterologist in producing an appropriate evaluation statement.

Another goal of this study was linking cyst thermograms to proteomic signatures of cyst samples, bearing in mind that, as we said before, TLB cyst profile reflects sample composition and also potential interactions and modifications in proteins. Each protein unfolds in a certain manner and the calorimetric cyst profile obtained is characteristic for that unfolding process. When two or more proteins are together in the same sample, the resulting profile will be the sum of the signal of the independent profiles when proteins are not interacting; or, in case the proteins interact, the resulting profile could change [42].

A deeper analysis of the proteomic profile of pancreatic intra-cyst sample can be found in the literature [43]. We used proteomic identification to assess the correlation between the TLB thermograms and the protein content. The TLB methodology is based on apparently simple curves, but they contain relevant information for diagnosis.

This is clear when looking at WOPN samples (as we already described above): TLB thermograms were quite similar and overlapped, and proteomic profiles also confirmed this similarity. Nevertheless, results are not so clear for the rest of the cyst groups and further studies to increase the number of intra-cyst samples are needed.

This pilot study reveals some interesting aspects. In IPMN, less inflammatory process indicators have been detected, specifically less immunoglobulins in comparison to benign cysts. There were also less proteins related to iron metabolism (serotransferrin, lactotransferrin, ferritin, hemopexin and haptoglobin) which has been referenced as an indicator of health alteration [44,45].

The information obtained through the cyst TLB thermogram is not focused on the detection of a specific biomarker but comprises information about a large amount of proteins and their relations (interactions and modifications), which could be interesting when doubts about the diagnosis of a cyst are cast. This study represents a proof of concept to confirm that TLB cyst thermograms contain valuable information. According to these preliminary results, a further, larger study can lead to the establishment of certain specific profiles for different type of cysts and can help during diagnosis.

The analysis of a cyst fluid sample implies that the patient has undergone an endoscopic ultrasonography procedure, a risky, invasive test that nowadays is the only way to achieve a reasonably accurate diagnosis of the benign or premalignant nature of cysts. Here we wondered if alterations due to the cyst could have a systemic reflection in blood serum, and whether or not they could be detected through TLB. If metabolic alterations in cyst fluid are translated into blood serum alterations, a quick, low-risk, minimally invasive tool to obtain information on the cyst would be available to differentiate between healthy/cystic patients and/or benign/premalignant cystic patients. TLB has been previously applied in our lab to study different type of tumor disease such as gastric, lung cancer or melanoma [30,33,34]. In the present study, we also obtained 15 serum samples from the 20 patients and their TLB serum profiles were obtained (Figure 5). As normally occurs with TLB serum thermograms, it is difficult to visually discriminate and distinguish among patients' profiles. In previous studies from our group, we contributed to develop, first, a

multiparametric analysis [30] and, later, a TLB serum score [33] to manage and assess TLB thermograms according to a simple interpretation index (TLB serum score from 0 to 1) that could be implemented in diagnosis, easily allowing the stratification of the patients.

We first compared the TLB serum thermograms from patients to thermograms from healthy subjects that were not suffering from the disease. We applied a general TLB serum score previously reported [33] and the results were statistically different when comparing healthy subjects with benign or malignant cyst patients, individually or together as a pooled group of patients (*p*-values lower than 0.05) (Figure 6). Specificity and sensitivity values (98% and 71%, respectively), as well as PPV and NPV values (83% and 98%, respectively), were quite high.

TLB serum scores (from benign or premalignant cyst groups) were not statistically different when using the TLB general formula when comparing to healthy subjects. Therefore, on that basis it was not possible to distinguish between both types of cyst. Therefore, we focused our attention on specifically comparing both patients' groups and obtaining a cyst-specific TLB serum score that could be applied to evaluate intra-group cyst patient variability. This new TLB serum score would strengthen the discrimination power, based on the specific parameters reflecting differences between cysts. The challenge is considerable, because a mono-variant analysis of cyst TLB parameters (Table S3) showed that none of the individual parameters was statistically different between the benign and premalignant cyst group. As we previously confirmed in our studies, the combination of all the parameters in a multiparametric-based single TLB serum score increased the discrimination ability. Despite the small patient sample cohort, it was possible to distinguish between benign and premalignant cysts. This specific TLB serum score (values from 0 to 1) were over 0.75 (according to Youden) in 6 out of 8 (75%) premalignant samples, while TLB serum scores were below 0.75 in 6 out of 6 (100%) benign samples. The diagnosis accuracy based in the area under the curve (AUC) was 0.875, a promising starting point for extending the study. Therefore, more serum samples from patients with pancreatic cysts should be included in a larger future study, but these preliminary results are promising and allow us to foresee that the TLB serum score could be applied routinely in the clinic as an additional complementary tool helping physicians in making better diagnostic decisions.

#### **5. Conclusions**

TLB analysis can be applied to both plasmatic serum and cyst fluid as a sort of high information content tool. Despite the small number of samples in this pilot study, it represents a proof of concept for developing a useful technique for classifying and evaluating risk in pancreatic cysts based on liquid biopsy in both body fluids. A future, larger study, with a larger number of samples for each cyst category, could confirm whether TLB of plasma or cyst fluid could be a new diagnosis tool to differentiate between benign and premalignant pancreatic cysts to help clinician gastroenterologists in making decisions about disease management. In case of serum, TLB would represent a quick, low-risk, minimally invasive tool easily translated to clinical practice for diagnosis and patient monitoring. In addition, TLB is reasonably cheap for serum tests (an estimated cost of 100–200 €/\$ per test, although with a higher cost for cyst fluid test), and could be performed with predefined frequency for patient surveillance.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2075-442 6/11/1/25/s1, Figure S1: Electrophoresis analysis of proteomic profiles, Figure S2: LC-ESI-MS/MS proteomic content of cyst samples types identification, Figure S3: ROC curve illustrating the statistical performance of TLB serum score (healthy vs. cysts); Table S1: Detailed information of cyst proteomic profiles, Table S2: Mono-variant analysis of TLB parameters of healthy controls and cysts patients, Table S3: Mono-variant analysis of TLB parameters of benign and premalignant cysts patients.

**Author Contributions:** Conceptualization, A.V.-C., and O.A.; methodology, S.H.-D., J.L.O., O.S.-G., A.V.-C., and O.A.; software, S.H.-D., J.L.O., O.S.-G., A.V.-C., and O.A.; validation, J.L.O., S.V., O.S.-G., A.V.-C., and O.A.; formal analysis, J.L.O., S.V., O.S.-G., O.A., and A.V.-C.; investigation, L.C, S.V., A.V.-C., and O.A.; resources, G.G.-R., J.L.O., O.S.-G., Á.L., C.S., A.V.-C., and O.A.; data curation, S.H.-D., G.G.-R., C.S., J.L.O., and O.A.; writing—original draft preparation, S.H.-D., L.C.-L., J.L.O., A.V.-C., and O.A.; writing—review and editing, S.H.-D., G.G.-R., L.C.-L., J.L.O., S.V., O.S.-G., Á.L., J.M., A.V.-C., and O.A.; visualization, J.L.O., L.C.-L., O.S.-G., A.V.-C., and O.A.; supervision, Á.L., A.V.-C., and O.A.; project administration, O.A. and A.V.-C.; funding acquisition, Á.L., O.A., and A.V.-C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Spanish Ministry of Economy and Competitiveness and European ERDF Funds (MCIU/AEI/FEDER, EU) (BFU2016-78232-P to A.V.C.); Projects funded by Instituto de Salud Carlos III and co-funded by European Union (ESF, "Investing in your future"): "PI15/00663 (FIS project to O.A)", "PI18/00349 (FIS project to O.A. and Contract to LC)", "FI19/00146 (PFIS contract for SHD)", "CPII13/00017 (Miguel Servet Program to OA)"; Diputación General de Aragón (Protein Targets and Bioactive Compounds Group E45\_17R to A.V.C. and Digestive Pathology Group B25\_17R to O.A.); and the Centro de Investigación Biomédica en Red en Enfermedades Hepáticas y Digestivas (CIBERehd).

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of CEICA (PI16/0228).

**Informed Consent Statement:** All subjects gave their informed consent for inclusion before they participated in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** Proteomic analyses were performed in the Proteomics Platform of Servicios Científico Técnicos del CIBA (IACS-Universidad de Zaragoza), ProteoRed ISCIII member, Zaragoza, Spain.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Arianna Dal Buono 1 , Federica Gaiani 2 , Laura Poliani 1,3 , Carmen Correale <sup>1</sup> and Luigi Laghi 2,4, \***


**Abstract:** Microsatellite instability (MSI) is the landmark feature of DNA mismatch repair deficiency, which can be found in 15–20% of all colorectal cancers (CRC). This specific set of tumors has been initially perceived as a niche for geneticists or gastroenterologists focused on inherited predispositions. However, over the years, MSI has established itself as a key biomarker for the diagnosis, then extending to forecasting the disease behavior and prognostication, including the prediction of responsiveness to immunotherapy and eventually to kinase inhibitors, and possibly even to specific biological drugs. Thanks to the contribution of the characterization of MSI tumors, researchers have first acknowledged that a strong lymphocytic reaction is associated with a good prognosis. This understanding supported the prognostic implications in terms of the low metastatic potential of MSI-CRC and has led to modifications in the indications for adjuvant treatment. Furthermore, with the emergence of immunotherapy, this strong biomarker of responsiveness has exemplified the capability of re-activating an effective immune control by removing the brakes of immune evasion. Lately, a subset of MSI-CRC emerged as the ideal target for kinase inhibitors. This therapeutic scenario implies a paradox in which appropriate treatments for advanced disease are effective in a set of tumors that seldom evolve towards metastases.

**Keywords:** microsatellite instability; colorectal cancer; immunotherapy; targeted therapy

#### **1. Introduction**

Colorectal cancer (CRC) is the third most common malignancy and cause of cancer mortality in Europe and the United States, accounting for nearly 900.000 deaths every year worldwide [1]. Among the newly diagnosed CRC, approximately 20% of patients still present with a metastatic disease, and a further 25% of those with an initially localized disease will eventually develop distant metastases [2,3]. Despite the fact that staging has traditionally represented the backbone of the prognostic factors in oncology, the growing knowledge of the molecular mechanisms of CRC has revolutionized the traditional or "old school" methods of managing tumor conditions. Indeed, CRC is a highly heterogeneous disease in regard to molecular expression and genetic abnormalities. It is known that a small subset of CRCs, approximately 15% of the cases, demonstrate microsatellite instability (MSI) due to an impaired DNA mismatch repair (MMR) system, though the vast majority of CRCs belong to the microsatellite stable (MSS) biomarker list [4]. MSI-CRCs are mostly sporadic, while approximately 3% of all CRCs harbor a germline mutation of mismatch repair genes (i.e., *MLH1*, *MSH2*, *MSH6*, *PMS2*, and *EpCAM*) identifying the Lynch syndrome [5]. The understanding of the carcinogenesis of MMR deficient tumors and subsequent clinical research has had an enormous therapeutic impact in the field of gastrointestinal oncology. In particular, the MSI status defines the largest group of inherited predispositions to

**Citation:** Dal Buono, A.; Gaiani, F.; Poliani, L.; Correale, C.; Laghi, L. Defects in MMR Genes as a Seminal Example of Personalized Medicine: From Diagnosis to Therapy. *J. Pers. Med.* **2021**, *11*, 1333. https://doi.org/ 10.3390/jpm11121333

Academic Editors: Enrico Mini and Stefania Nobili

Received: 3 October 2021 Accepted: 6 December 2021 Published: 8 December 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

gastrointestinal cancers and impacts the prognosis of CRC, giving better stage-adjusted survival rates compared to MSS tumors [6,7]. Moreover, MSI colorectal tumors are more frequently seen at early stages (i.e., stage II–III), and only 3.5% of the cases present with a metastatic disease [8], in accordance with a reduced distant metastasis, which is intrinsic to MSI status. MMR/MSI testing is increasingly being incorporated as a standard of care for all CRC patients and is collectively recommended by the most important scientific societies involved in the field, such as AGA, ASGE, ASCRS, ASCO, and ESMO [9]. This review summarizes the evidence demonstrating the value of MSI as a diagnostic and prognostic tool and eventually also a predictive biomarker in the personalized approach to CRC.

#### **2. Discovery of MSI, Its Relevance in Lynch Syndrome and Understanding the Different Molecular Pathogenesis of CRC**

#### *2.1. Parallel Discovery*

The discovery of DNA mismatch repair (MMR) defects is an interesting outcome, which testifies how the contemporary efforts of different teams have helped to elucidate the molecular basis of Lynch syndrome (LS) in a relatively short period of time. However, in addition to contributing to the development of a new era in molecular medicine, it has also raised other lessons in LS management that are worth recalling. The reason for this is chiefly that different methodological approaches were used by the groups involved in the research. To be precise, finding the mechanism behind LS was not the shared aim of these teams. The study led by Perucho was involved in identifying a particular mechanism of carcinogenesis through an unbiased molecular approach, defined as an "arbitrarily primed polymerase chain reaction" (PCR) [10,11]. In doing so, his group found that a fraction of CRCs harbored un-corrected frame-shifted DNA tracts, and they referred to such changes as ubiquitous somatic mutations. The team led by Thibodeau [12] was looking for allelic losses (and gains) by PCR and noted that there was "instability" at the amplified microsatellite sequences (hence microsatellite instability or MSI), in some proportion of the CRCs. Neither study was familiar with or looking for familial cancer or Lynch syndrome genes. Meanwhile, an international consortium with a strong membership from Finland, including Aaltonen, was trying to identify the loci associated with Lynch syndrome by employing an allelotyping approach to search for loss of heterozygosity [13,14]. With the exploration of dinucleotide repeats in tumor DNA compared to normal subjects, CRC patients were found to have frame-shifted sequences, which they described as replication errors (RER). Subsequently, the term MSI was used to describe the same phenomenon that these groups identified and described, although the degree of competition was very high. Perucho's reference to a probable inherited syndrome was incorporated within the manuscript after Aaltonen and Vogelstein's group had mapped and reported a Lynch syndrome locus on 2p, a finding already detected by Perucho. In a timely editorial, it was noted that "the cancers whose cells carry shortened repeats are differently distributed in the colon from others and metastasize less frequently. If Perucho is right in believing that the underlying fault may be a mutation of a DNA repair gene, the ramifications of that may be exceedingly important" [15]. These words summarized the relevant biological and clinical implications of the discovery of DNA MMR defects.

These inherent differences led to a dual development of research efforts in the field. On one side, the genes involved in DNA mismatch repair in humans were targeted, being first identified by Kolodner [16] and subsequently largely addressed in their relevance by various teams, including that led by Bert Vogelstein [17], as part of his landmark work unravelling the molecular bases of CRC, before and after the discovery of MMR defects.

On the other side, the research focused on the molecular pathogenesis of MMR deficient CRC and addressed the role of these types of mutations in the peculiar behavior of MSI tumors. It soon became evident that these cancers remain in a class of their own among tumors [18], as compared to other known genetic pathways to CRC, mainly driven by *APC* gene damage both in inherited (i.e., Familial Adenomatous Polyposis) and sporadic carcinogenesis. In this respect, MMR-deficient tumors appear mainly a disease marked by accelerated tumor progression rather than by an accelerated tumor initiation. It was

appreciated that the burden of unrepaired mutations in these tumors contributes to their indolent behavior [19,20] and to the amount of immune response that they elicit [21,22]. Surprisingly, these areas of investigation took years to generate translational research aimed at systematically identifying prognostic markers for CRC and then influencing clinical practice. It should be mentioned that for the first time since the discovery of MMR defects and MSI, a molecular phenotype has recently been proposed for the molecular screening of a specific disease subtype [23]. This long journey led to the exclusion from adjuvant therapy of patients with stage IIA MSI CRC, even though they displayed high-risk hallmarks and contributed to defining the role of tumor-infiltrating lymphocytes (TILs) as a prognostic marker in CRC staging (see below).

#### *2.2. Unraveling the Pool of Genes Involved in DNA MMR and Deranged in Lynch Syndrome*

MMR is a mechanism whereby proteins identify and repair mismatched bases occurring mostly by statistical chance during DNA replication or genetic recombination, a mechanism that is present among many species. DNA mismatching, however, is also enhanced by chemical or physical damage. The high conservation rate among species accounts for its importance, as does the discovery of its involvement in human disease by a basic scientist [16]. He was able to cross its defects with the by-then emerging phenotype of MSI in human CRC, thus developing a strategy to identify one of its components (namely, *MSH2*) as the culprit for a fraction of the cases of Lynch syndrome, moving from the similarities of molecular signatures in yeasts. That is why the human genes were initially labelled as homologues of their counterpart in yeasts.

As the result of a plurality of efforts, we now know that this system is constituted by multiple proteins, including MLH1 (MutL homologue), PMS2 (post-meiotic segregation protein), MSH2 (MutS homologue), MSH6, MLH3, MSH3, and PMS1, which form heterodimers with different roles: MSH2/MSH6 and MSH2/MSH3 heterodimers recognize and bind base–base mismatches and insertion/deletion loops, and subsequently, they recruit MLH1/PMS2 heterodimers to excise and allow the resynthesis of corrected strands [4,24,25]. Later, deletions of the 3′ distal portion of the *EPCAM* gene, containing the termination codon, have been demonstrated to influence the MMR system by leading to the methylation of the promoter of the downstream neighbor *MSH2* and therein to its silencing [26]. Genetic or epigenetic events leading to the silencing of one of the genes of the MMR system ensues in the appearance of the mutator phenotype. Irrespective of the underlying molecular mechanisms, the inactivation of any of the members of the MMR genes leads to the disappearance of the encoded protein. However, the loss of MSH2 or MLH1 also leads to the loss of expression of that protein itself and its heterodimer partner, whereas the loss of MSH6 or PMS2 results in the loss of expression only of the specific protein. Accordingly, germline inactivating mutations of the genes encoding for one among the MMR proteins stay at the basis of MSI as the first pathogenetic damage of the Lynch syndrome and should be followed by a second somatic inactivation hit according to the Knudson hypothesis turning off the second allele [24,27].

In the seminal phase of the late 1990s, addressing MSI in clinical practice was mostly based on clinical criteria, namely the Bethesda ones [28,29]. In other words, the clinical criteria used to define Lynch syndrome (by then referred to as Hereditary Non-Polyposis CRC, HNPCC) or Amsterdam criteria [30,31] were loosened and expanded to identify those patients suitable for the analysis of the MS-status of their CRC and then to germline sequencing if the results of the somatic analysis revealed MSI. Initially, the characterizations of tumor samples based on MS-status comprised the classification into microsatellite instability high (MSI-H) if two or more of the microsatellite markers show instability (or >30% of unstable markers if a larger panel is used) and microsatellite instability low (MSI-L) if only one marker shows instability, as opposed to MSS cancers [24,32]. However, such a classification has been variably criticized, and the distinction in MSI-H and MSI-L progressively lost relevance, and the latter group is cumulated with MSS tumors [33,34].

The systematization of the characterization of the MS status in CRC has confirmed the initial findings by Perucho et al. that most MSI tumors are not the epiphenomenon of LS but are instead sporadic. In fact, considering that MSI cancers account for 15% of all CRCs, only 3% of the total (or 20% among MSI cases) are attributable to Lynch syndrome [35]. It is also now clear that hereditary MSI cancers differ from sporadic ones by means of the type of underlying alteration causing the impairment of the MMR system (as well as in their clinical behavior).

#### *2.3. Sporadic MSI Cancers and Hypermethylation*

Patients with sporadic MSI CRC are significantly older than those affected by Lynch syndrome, and most of them lack any significant familial clustering, nevertheless maintaining a better prognosis than those with MSS tumors [36]. The molecular features of sporadic MSI tumors, instead of germline pathogenic variants of MMR genes plus second hit on the other allele, are the methylation of *MLH1* promoter frequently coupled with the mutation *BRAF*(V600E) [4,37].

Understanding the molecular pathogenesis of sporadic MSI CRC was the sequel of the discovery of germline MMR defects, which has helped clarify the mechanism for a portion of otherwise unexplained cases, as well as introducing one additional cancer phenotype [38–40]. In fact, the main mechanism for a sporadic MSI CRC going through the inactivation of the promoter region of the DNA mismatch repair gene *MLH1* by hypermethylation [41] mostly occurs in the context of the CpG island methylator phenotype (CIMP) [42]. CpG islands are genomic regions rich in cytosine and guanine repeats present in about 40–50% of human genes, usually located at the promoter region and crucial for the epigenetic inactivation of gene transcription by hypermethylation [42].

Although the methylator phenotype can be intended as the main molecular biomarker of sporadic MSI tumors, CIMP can also be found in a group of patients who present no anomalies of the MMR system. Further studies by Ogino et al. [43] and Samowitz et al. [44] demonstrated that not all sporadic MSI tumors with *MLH1* hypermethylation have a methylator phenotype. The scenario of CRC molecular characterization has become more and more complex over the years, adding the CIMP status as a separate parameter of classification [41,45]. CIMP+ (or CIMP-high) CRCs are reported to be more frequent in the elderly and in women, are often located in the proximal location, show poor differentiation, and have a high frequency of MSI and *BRAF* mutation [41,46,47], largely overlapping with sporadic MSI cases. CIMP was originally described as the *de novo* methylation of the 5′ CpG island of p16 (now *CDNK2A*) detectable in approximately 1/5 of different tumor types and acting as an alternative mechanism for the silencing of tumor suppressor genes [48].

Although the value of CIMP is not well known, CIMP+ CRC seems to have a better outcome than CIMP-low (particularly if showing wild-type *BRAF*) and appears to respond more efficiently to adjuvant treatments [41].

#### *2.4. Lynch Syndrome versus Lynch-Like Syndrome*

The seminal report on what will be later referred to as HNPCC and Lynch syndrome dates to the end of the XIX century by Aldred S. Warthin, who reported the pedigree of "family G" with a cluster of uterine, gastric, and abdominal cancer, which led him to suspect the existence of a form of predisposition [49]. Years later, Henry Lynch reported similar familial clusters of cancer and reviewed the history of family G, with a predominance of cancers of the colon, uterus and stomach [50]. Notably, Lynch concluded the culprit was an autosomal dominant inheritance of this otherwise unrecognized syndromic cluster, referred to as "Cancer Family Syndrome" (for an exhaustive perspective on the historical development of the medical perspective on the topic, see Boland, 2013) [50]. Later, the term HNPCC was used to refer to the lack of a phenotypic hallmark compared to polyposis syndromes [51]. However, once a molecular phenotype had been identified and its basis clarified, the term Lynch syndrome was encouraged and adopted for those cases with a defined MMR defect and a germline mutation in the MMR genes. Alternately, the lack of a

pathogenic germline mutation in a patient with an MSI CRC and features suggestive of an underlying predisposition is called "Lynch-like" syndrome [52,53]. The two syndromes have the development of MSI CRCs at a young age and the presence of extracolonic cancers in common. However, although in patients affected by Lynch-like syndrome, the onset of cancer is in the fifth decade (mean age, 54.9 years) [53], the standardized incidence ratios of CRC and extracolonic cancers is lower (2.12 vs. 6.04 and 1.69 vs. 2.81, respectively) [54].

Although Lynch-like syndrome patients lack germline mutations of the MMR system, they exhibit in almost half of all cases the biallelic somatic inactivation of DNA MMR genes within the tumor [54,55]; moreover, they might harbor germline mutations of unknown genes other than MMR ones. Nevertheless, due to the increased cancer risk for the proband and his or her relatives, a careful follow-up remains advisable from a clinical perspective [54,55].

#### **3. Prognostic Value of MSI in CRC**

#### *3.1. Lower Metastatic Potential and Better Survival of MSI CRC*

MSI is undoubtedly a positive prognostic factor in CRC patients, which is promptly explained by the low prevalence of MSI tumors among metastatic CRCs, corresponding to 2–4% of stage IV cases [4,25], as compared to their prevalence in earlier stages [56,57]. MSI CRCs typically present a dense immune cell infiltration, particularly rich in TILs, which has been associated with a better prognosis and a reduced tendency to metastasize [8]. Substantial evidence supports that MSI is a strong prognostic marker in early-stage CRCs with a favorable impact on survival, beyond the TNM staging system also from pooled retrospective analyses [58]. With respect to stage II CRC patients, in the ACCENT database analysis, the MSI profile significantly improved the disease-free survival and the overall survival [59].

Compared to stage II, the prognostic value of MSI in stage III CRC is less defined, and contradictory data have emerged from randomized clinical trials (RCTs) and metaanalysis [60–62] (see below).

Summarizing the available data, MSI confers a favorable prognosis in stage II CRC, and this effect seems to be progressively reduced with advancing stage (i.e., stage III) [60–62]. A speculative explanation of this phenomenon lies in the evasion of immune surveillance that is possibly acquired in more advanced stages of the disease. In accordance with the above statement, in stage IV CRCs, MSI no longer provides an advantage in terms of prognosis [63,64], though interactions with chemotherapy, as the standard adjuvant treatment for stage III CRC, should not be disregarded despite being difficult to disentangle.

#### *3.2. Adaptive Immune Response and the Relevance of Immune Parameters*

MSI-CRCs attract a dense lymphocytic infiltrate [21,22], parallelly driving the infiltration of specific subsets of immune cells (i.e., cytotoxic and helper T-lymphocytes) that are associated with an improved prognosis and reduced recurrence rates after surgery [63], especially in patients with early, node-negative CRC, largely contributing to the prognostic advantage of high densities of infiltrating lymphocytes [65].

Among the immune subpopulations recruited by MSI-CRC, dendritic cells and T cells activate the immune antitumoral response, which is downstream accomplished by activated memory CD4 + T cells, NK cells, M1 macrophages, and neutrophils [66]. The attempt to measure the immune infiltrate in the primary tumor and to assess its prognostic value has been pursued by trying to build a reliable "immuno-score" that quantifies the amount of infiltrating T-lymphocytes and allows inferences on CRC outcomes [67]. The immuno-score has been suggested to be superior to the conventional TNM classification in CRC, given its ability to differentiate patients with a better or worse prognosis in MSS and MSI disease, as across the various stages according to AJCC/UICC [68]. The measure of CD8+ cells and CD45RO+ memory cells in specific tumor regions (i.e., at the invasive front) has been, in fact, linked to longer overall survival in MSI-CRC patients [69]. This parameter

is likely to be included in the TNM staging, similarly to its use for MSI, although some refinement is necessary in order to better define its reliability in stage III disease [36,44].

#### **4. Predictive Value of MSI**

#### *4.1. Implication for the Adjuvant Treatment: Stage 2 vs. Stage 3*

In stage II CRC, MSI has been endorsed as a reliable predictive indicator associated with a lack of benefit from adjuvant chemotherapy (5-fluorouracil-based (5FU)). This clinical endorsement first moved from the better prognosis and lower metastatic potential of MSI CRCs [56,57].

The initial report on non-responsiveness came from a study by Ribic et al. in which patients with MSI CRC were found to have a better overall 5-year survival, especially when not receiving adjuvant chemotherapy [70]. Subsequently, Sargent, in a collaborative study, confirmed this finding by showing that MSI interacted significantly with chemotherapy and that there was no improvement in patients with stage II MSI CRC who had received 5-FU [71]. Sinicrope et al. shortly after confirmed that patients with MSI CRC have lower rates of tumor recurrence, delayed time to relapse, and improved survival rates, with respect to MSS CRC patients [72]. Adjuvant treatment also reduced the rate of distant recurrences in patients with stage III CRC, which could be significant in patients with germline pathogenic variants compared to those with sporadic tumors [73].

A milestone in modern oncology was placed in the phase III Quick and Simple and Reliable (QUASAR) trial that randomized more than 2000 patients affected by stage II CRC to either receive adjuvant chemotherapy with 5-FU or for observation [74]. The study showed a significantly reduced risk of recurrence for MMR-deficient CRC (risk ratio, 0.53, 95% C.I., 0.40–0.70; *p* < 0.001) as compared to proficient ones, and the subanalysis for MMR status demonstrated no benefit from adjuvant chemotherapy [74]. This evidence has been confirmed by several meta-analyses that established MSI status as a predictive factor for both therapy response and relapse rates as concerns in stage II CRC [75–77]. Overall, data support MSI as the leading molecular marker with clinical value in early-stage CRC; no further molecular stigma has been incorporated in the management algorithms of CRC yet.

The situation in stage III appears more complex. In a study on patients included in a randomized trial on adjuvant 5-FU plus Oxaliplatin and folinic acid (FOLFOX) after resection of stage III CRC, Sinicrope et al. found that *KRAS* and *BRAF* mutations had a negative prognostic effect on disease-free survival, while MSI was not prognostic in all patients but significantly interacted with the tumor site and nodal status [78]. Accordingly, only patients with right-sided MSI CRC had a better outcome, and such an advantage was lost in those with N2 tumors [78].

In an interesting study assessing the value of lymphocyte infiltration in patients included in the PETACC8 phase III study [79], the authors found that MSI was not a predictive factor for overall survival in treated patients [79]. However, a larger study adding patients from the NCCTG N0147 trial [80] found that patients with MMR-deficient CRCs had significantly longer disease-free survival than those with proficient tumors at multivariate analyses (HR, 0.73; 95% CI, 0.54–0.97; *p* = 0.03), although such advantage may become evident only after 18 months at Kaplan–Meier survival curves. One issue involves the benefit of oxaliplatin added to 5-flurouracil [80]. Interestingly, it had been shown earlier that in an MSH2-deficient mouse model developing CRC, FOLFOX treatment led to a reduction in tumor volume, and MMR status was found not to modify responsiveness to oxaliplatin in previous studies [81,82].

Other studies further clarified that *KRAS* and *BRAF* mutations act as negative prognostic factors in MSS CRC patients treated with adjuvant FOLFOX, but not in MSI patients [83].

#### *4.2. Removing the Breaks from the Immune Response: Immunotherapy*

In the last decade, translational research in oncology has been focusing on the molecular mechanisms driving the interaction between MSI CRC and the immune system. The MSI status influences the tumoral microenvironment and the interactions with the immune

system through multiple aspects, therefore impacting the efficacy of immunotherapy. A defective MMR leads to a high tumor mutational burden (TMB) [11,19,20], which means that tumoral cells profusely generate highly immunogenic soluble and surface neoantigens able to attract cytotoxic and helper T-lymphocytes [22,84]. The higher somatic mutational load that increases the presentation of neoepitopes has been epitomized as one of the mediators of the observed augmented response to immunotherapy as well in MSI tumors [85,86]. The immunogenicity of these neoantigens, structurally frame-shifted peptides, lies in their ability to bind with major histocompatibility complex class I (MHC-I) alleles [87]. Moreover, the neo-antigen load was directly associated with the T-cell memory tumoral infiltration [87].

Secondly, as demonstrated by Llosa et al., neoplastic cells with MMR defect overexpress several immune checkpoint proteins (e.g., PD-1, PD-L1, CTLA-4, LAG-3, and IDO), compared to MSS cancers [88].

These findings, together with evidence stemming from clinical trials, initially led immune checkpoint inhibitors (i.e., anti-PD1) to be approved by the regulatory authorities exclusively according to the MSI status, regardless of cancer type [8].

Recent studies investigating anti-programmed death-1 (PD-1) checkpoint inhibitors have identified and demonstrated MS status as a biomarker predictive of therapy response [89,90]. MMR-deficient cancers are now acknowledged to be sensitive to anti-PD1 (nivolumab, pembrolizumab) with or without anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) antibodies [89,90].

#### *4.3. Silencing Map Kinases in Sporadic MSI*

In the current landscape, it has become clear that *BRAF*-mutant CRC represents a distinct biologic entity, typically refractory to the traditional chemotherapy regimens [91]. *BRAF* is a serine/threonine kinase that acts downstream of *KRAS* in the mitogen-activated protein kinase (MAPK) cellular signaling pathway. *BRAF*-mutant CRC commonly exhibits a valine to glutamic-acid variation, specifically at codon 600 (V600E; or 1799T>A). The effect of this change is a constitutively activated protein. The *BRAF* V600E mutation overlaps with sporadic MSI-CRC in up to 33% of the cases [92].

Historically, *BRAF*-mutated CRCs have been associated with a significantly worse prognosis [93]. The therapeutic implications of targeting this mutation came as a lesson from the management of *BRAF*-mutated melanomas, and currently, several ongoing clinical trials are investigating the efficacy of BRAF-inhibitors (i.e., dabrafenib, vemurafenib, or encorafenib) alone or in combination in patients with metastatic *BRAF* (V600E)-mutated CRC [94].

In terms of personalized medicine, the inhibition of MAPK signaling in sporadic MSI-CRCs has been explored with promising results. In a pivotal, single-arm study that included 43 patients with *BRAF*-V600E metastatic CRC treated with the adjunct of a MEK inhibitor (Trametinib), the results showed improved response rates compared with BRAF inhibition alone [95]. A further phase II study, comparing dabrafenib, trametinib, and panitumumab triple therapy with double therapies (either dabrafenib plus panitumumab or trametinib plus panitumumab), assessed a disease control rate (response and stable disease together) in 86% of patients [96]. The median progression-free survival (PFS) and the duration of response were 4.2 and 7.6 months, respectively [96].

Based on these preliminary data, the combination therapies of BRAF/MEK have not been approved by the Food and Drug Administration (FDA) for the treatment of metastatic *BRAF* V600E CRC yet. Lastly, clinical trials examining immunotherapy in combination with inhibitors of the MAPK pathway are expected.

#### **5. Discussion and Concluding Remarks**

This review illustrates the current evidence on the prognostic and predictive value of MSI as a trail maker of the personalized medicine approach to CRC. Compared to MSS CRC, MSI status is associated with a more favorable prognosis in early-stage CRCs [58].

Furthermore, based on the evidence that adjuvant chemotherapy does not add any advantage for the prognosis in stage II, knowledge of MSI status drives clinical decisions for these patients [59]. Conversely, the prognostic value of MSI with respect to stage III disease appears attenuated, and these patients are, so far, recommended to receive standard adjuvant chemotherapy.

Regarding the predictive value of MSI status, it has been extensively demonstrated to be a robust biomarker for a good response to immune checkpoint inhibitors in patients with metastatic disease [89,90]. However, the precise role of immunotherapy in earlier-stage CRCs needs to be clarified by ongoing randomized studies. The studies on the molecular heterogeneity and tumoral microenvironment surrounding MSI tumors have led to an increased understanding of possible innovative therapeutic targets.

Figure 1 summarizes the timeline of the gradual achievement of a progressively wider clinical usefulness of MSI status in the field of CRC.

**Figure 1.** MMR story: lessons from a long-lasting biomarker. Timeline of its gradual achievement of wider clinical usefulness.

Finally, research has recently been focusing on the relationship between gut microbiota and CRC tumorigenesis, with a particular interest in the induced molecular profile, such as MSI. What is emerging is that, among the different microbiological species, *Fusobacterium nucleatum* is linked to the development of MSI tumors [97,98]. Indeed, tumors with high levels of *Fusobacterium nucleatum* tend to occur in the proximal colon and have a higher incidence of MSI with rather poor survival, as reported in a prospective cohort study [98]. This seems somehow counterintuitive, and it has been associated with the capability of *Fusobacterium nucleatum* to suppress the adaptive immune response in MSI-CRCs [99].

In the foreseeable future, gut bacterial modulation or a fecal microbiota transplant could stimulate the immune response in patients with MSI-CRCs that have developed a secondary resistance to immunotherapy. Thus, the modulation of the microbiota and increased antigen presentation appear to be two possible therapeutic targets for new and personalized strategies aimed, for example, at restoring a competent immune response and immunotherapy efficacy in MSI tumors. As we have gleaned much more than we would have expected from the MSI tumor subtype, we should be confident there is yet more to learn.

T3N0M0 CRCs, or stage IIA, invade through the muscolaris propria into the subserosa but have not reached nearby organs and lymph nodes and have not spread to distant organs [100]. FOLFOX, comprising of 5-FU, Oxaliplatin, and Folinic acid, is administered after surgery as adjuvant treatment.

**Author Contributions:** Conceptualization, L.L.; methodology, A.D.B., F.G.; writing—original draft preparation, A.D.B., F.G.; writing—review and editing, L.P. and C.C.; supervision, L.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Abbreviations**


#### **References**


## *Article* **LAG-3 Expression Predicts Outcome in Stage II Colon Cancer**

**Gaëlle Rhyner Agocs 1,2 , Naziheh Assarzadegan 3,4 , Richard Kirsch 3 , Heather Dawson 5 , José A. Galván 5 , Alessandro Lugli 5 , Inti Zlobec <sup>5</sup> and Martin D. Berger 1, \***

	- heather.dawson@pathology.unibe.ch (H.D.); jose.galvan@pathology.unibe.ch (J.A.G.);

**Abstract:** Introduction: LAG-3 is an inhibitory immune checkpoint molecule that suppresses T cell activation and inflammatory cytokine secretion. T cell density in the tumor microenvironment of colon cancer plays an important role in the host's immunosurveillance. We therefore hypothesized that LAG-3 expression on tumor-infiltrating lymphocytes (TILs) predicts outcome in patients with stage II colon cancer. Patients and Methods: Immunohistochemical staining for LAG-3 was performed on tissue microarrays (TMAs) of formalin-fixed paraffin-embedded tissue from 142 stage II colon cancer patients. LAG-3 expression was assessed in TILs within both the tumor front and tumor center and scored as either positive or negative. The primary endpoint was disease-free survival (DFS). Results: In patients diagnosed with stage II colon cancer, the presence of LAG-3 expression on TILs was significantly associated with better 5-year DFS (HR 0.34, 95% CI 0.14–0.80, *p* = 0.009). The effect on DFS was mainly due to LAG-3-positive TILs in the tumor front (HR 0.33, 95% CI 0.13–0.82, *p* = 0.012). Conclusion: Assessment of LAG-3 might help to predict outcomes in patients with stage II colon cancer and potentially identify those patients who might benefit from adjuvant chemotherapy. Therefore, LAG-3 may serve as a prognostic biomarker in stage II colon cancer.

**Keywords:** biomarker; LAG-3; immune checkpoint; colon cancer; survival

#### **1. Introduction**

Colon cancer is a major cause of cancer-associated death worldwide and its global incidence is continuously increasing [1,2]. The prognosis of patients with colon cancer is mainly dependent on the stage according to the TNM classification system [3]. While patients diagnosed with stage I colon cancer have an excellent outcome, those with stage IV disease have a low chance of cure and a significantly worse survival [4,5]. Interestingly, some patients with stage II colon cancer, especially the ones with a pT4b tumor (stage IIC), have a worse outcome compared with those with a pT1 or pT2 N+ tumor (stage IIIA and IIIB). In fact, one might assume that a node-positive disease indicates a more aggressive tumor biology translating into a poorer clinical outcome [4]. According to current guidelines, adjuvant chemotherapy is indicated for stage III colon cancer and for stage II disease with additional clinicopathological risk factors [6]. Despite curative surgery and adjuvant chemotherapy, relapses occur and pose significant challenges for our health care system. However, some patients will not benefit from adjuvant chemotherapy because they have already been cured by surgery alone [7]. Therefore, new prognostic and predictive biomarkers have to be developed to define subgroups of patients, especially

**Citation:** Rhyner Agocs, G.; Assarzadegan, N.; Kirsch, R.; Dawson, H.; Galván, J.A.; Lugli, A.; Zlobec, I.; Berger, M.D. LAG-3 Expression Predicts Outcome in Stage II Colon Cancer. *J. Pers. Med.* **2021**, *11*, 749. https://doi.org/10.3390/ jpm11080749

Academic Editors: Enrico Mini and Stefania Nobili

Received: 16 July 2021 Accepted: 20 July 2021 Published: 30 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

those with low- or standard risk stage II cancer who have a high chance of relapse and may derive the most benefit from chemotherapy [8].

Recently, substantial progress has been made in understanding the role of immune cell infiltration in colon cancer, which has led to the development of the Immunoscore in stage I–III colon cancer, based on the quantification of CD3+ and CD8+ lymphocytes in the tumor and its invasive margin [9]. Remarkably, this scoring system predicts disease-free and overall survival even more precisely than the TNM classification and might therefore guide our treatment decisions in the future [9]. Moreover, the identification of immune-related biomarkers on tumor-infiltrating lymphocytes (TILs) might help to predict prognosis and direct clinical decisions. Recent trials demonstrated that the presence of immune-related biomarkers such as PD-1/PD-L1 on tumor-infiltrating cells might serve as prognostic biomarkers in colorectal cancer [8,10]. The lymphocyte-activation gene 3 (LAG-3 or CD223) is another inhibitory immune-related molecule that is expressed on T cells, especially on activated CD8+ and CD4+ T cells, but also on B cells and dendritic cells, which may act in synergy with the PD-1/PD-L1 pathway [11,12]. LAG-3 mainly binds to the major histocompatibility complex II (MHC II), and thus prevents the interaction of the MHC II with the T cell receptor (TCR) on CD4+ T cells, resulting in decreased CD4+ activity. Another ligand of LAG-3 is Galectin-3, which is mainly expressed on epithelial and immune cells and preferentially binds to LAG-3 on CD8 cells [13]. The protein liver sinusoidal endothelial cell lectin (LSECtin) is a further potential ligand that binds to LAG-3 [14,15]. Upregulation of LAG-3 on immune cells downregulates T cell expansion and cytokine secretion, and thus contributes to an immunosuppressive microenvironment [16,17].

Given the growing interest in the role of LAG-3 in cancer, we sought to evaluate the presence of LAG-3 on tumor-infiltrating lymphocytes (TILs) in the tumor center and tumor front and to assess its impact on outcomes in stage II colon cancer.

#### **2. Patients and Methods**

#### *2.1. Patient Cohort*

Between 1992 and 2010, patients at the Mount Sinai University Hospital in Toronto, Canada with curatively resected stage II colon cancer in which archival material was available were consecutively included in this retrospective study. Patients with rectal cancers were excluded from our analysis.

A histopathological review was performed according to the 6th edition of the TNM classification system. Clinical data were obtained from patient records. The baseline characteristics comprised age at diagnosis, gender, tumor location, pT stage, tumor grade and lymphatic and venous vessel invasion. In addition, tumor budding, considered as a supplementary prognostic factor, was scored according to the International Tumor Budding Consensus Conference 2016 [18]. Moreover, the mismatch repair (MMR) status was determined by immunohistochemistry. The study was approved by the research ethics board of the Mount Sinai Hospital (nr 13-0136).

#### *2.2. Next-Generation Tissue Microarray (ngTMA®) Construction*

H&E-stained (hematoxylin and eosin-stained) whole slides of each case were digitized using a slide scanner (3DHistech, P250, Hungary). Each scan was annotated twice using a 0.6 mm tool in four different regions of interest: tumor center (encompassing mostly tumor epithelium), tumor front (targeting 50% tumor and 50% stromal areas at the invasion front), and tumor stroma (including largely stromal areas at the invasion front with only little, if any tumor). This produced 11 ngTMA® blocks with six cores per tumor.

#### *2.3. Immunohistochemistry*

2.5 µm ngTMA® sections were mounted onto glass slides, dried and baked at 60 ◦C for 30 min prior to use. All immunostainings were performed by automated staining using Bond RX (Leica Biosystems, Muttenz, Switzerland). All slides were dewaxed in Bond dewax solution (product code AR9222, Leica Biosystems). Heat-induced epitope retrieval at

pH 9 in Tris buffer base (code AR9640, Leica Biosystems) followed for 30 min at 95 ◦C. LAG-3 rabbit monoclonal antibody (Cell Signaling, clone D2G4O Ref 15372) was diluted 1:200 and incubated for 30 min at room temperature. Then, the samples underwent incubation with HRP (Horseradish Peroxidase)-polymer for 15 min and were subsequently visualized using 3,3-Diaminobenzidine (DAB) as brown chromogen (Bond polymer refine detection, Leica Biosystems, Ref DS9800) for 10 min. Finally, the samples were counterstained with hematoxylin, dehydrated and mounted with Tissue-Tek® Glas™ Mounting Media (Sakura). Slides were scanned and photographed using Pannoramic 250 (3DHistech). The immunostainings for the mismatch repair proteins were performed using the VENTANA MMR IHC Panel and the VENTANA BenchMark automated staining system (Roche Diagnostics, Mannheim, Germany) according to the manufacturer's instructions. The following antibodies were used: anti-MLH1 (mouse, clone M1, Roche Diagnostics, Ref 8504946001), anti-MSH2 (mouse, clone G219, Roche Diagnostics, Ref 8504946001), anti-MSH6 (rabbit, clone SP93, Roche Diagnostics, Ref 8504946001) and anti-PMS2 (mouse A16-4, Roche Diagnostics, Ref 8504946001).

#### *2.4. Evaluation of Immunohistochemistry*

All TMA cores for each individual case were evaluated for the presence or absence of LAG-3 immunohistochemical staining (G.R). Consistent with Sobottka et al., LAG-3 expression on TILs within both the tumor front and tumor center was dichotomously scored as either positive or negative [19]. Representative images are outlined in Figure 1A–D. LAG-3 positivity on TILs was defined as membranous staining of any intensity regardless of the number of LAG-3 positive immune cells (≥1), whereas the absence of any staining was determined as LAG-3 negative. We reported the scores for LAG-3 positive TILs as absolute numbers and used the maximum score of all analyzed tissue cores from each patient. The tumor front was defined as the area where the most advancing cancer cells reached the edge of the tumor. In a control set of normal, non-neoplastic colon tissues, no LAG-3 positivity could be observed. Mismatch repair deficiency (dMMR) was defined as the loss of nuclear expression of at least one of the four MMR proteins (MLH1, MSH2, MSH6 and PMS2) in the tumor cells in the presence of an internal positive control such as lymphocytes, normal epithelium or fibroblasts in the close vicinity of the tumor. Focal weak and dotted nuclear staining were considered negative. Retained nuclear expression of all MMR proteins was determined as mismatch repair proficiency (pMMR) [20].

#### *2.5. Statistical Analysis*

Correlations between LAG-3 expression in TILs within the tumor center/front and categorical variables were tested using the chi-square test. Continuous or ordinal variables were analyzed with the Kruskal-Wallis or Wilcoxon rank sum test. Disease-free survival (DFS) analysis was performed with Kaplan-Meier survival curves and log-rank test. Hazard ratios and 95% confidence intervals (CIs) were used to determine the effect of each variable on outcome, using Cox regression analysis. DFS was calculated from the time of surgery to local or distant recurrence or death. A *p*-value < 0.05 was considered statistically significant. Statistical analysis was carried out by SPSS version 26 (United States).

μ μ **Figure 1.** (**A**,**B**): Absence of immunohistochemical staining for LAG-3 on TILs (magnification: A 5× and B 40×). (**C**,**D**): TILs expressing LAG-3 (magnification C 5× and D 40×). Scale bar (**A**,**C**): 100 µm. Scale bar (**B**,**D**): 50 µm. All images (**A**–**D**) represent the tumor center.

#### **3. Results**

#### *3.1. Patients Characteristics*

Our study population comprised 142 patients with curatively resected stage II colon cancer. The median age of the patients was 70 years (range, 24–98 years). 42.2% (*n* = 60) of the patients were female and 57.8% (*n* = 82) male. 85.8% (*n* = 121) of the study cohort had a pT3 tumor, whereas 14.2% (*n* = 20) of the patients presented with a pT4 tumor. 91% of the tumors (*n* = 122) were well (G1) or moderately (G2) differentiated and 9% (*n* = 12) presented with a G3 grading. 87.4% (*n* = 118) of the tumor specimens displayed no extramural venous invasion (V0), while vascular invasion could be observed in 12.6% (*n* = 17). There was a slight predominance of left- as compared to right-sided tumors (52.5%, *n* = 73 versus 47.5%, *n* = 66). In total, 124 patients (87.3%) had more than 12 lymph nodes examined, whereas the lymph node yield was less than 12 in 12.7% of the patients (*n* = 18). From 134 evaluable tumor tissue samples, 75.4% (*n* = 101) were MMR-proficient, whereas 24.6% (*n* = 33) were MMR-deficient (Supplementary Table S1). The median tumor budding count was 10 (0–74 buds). The 5-year DFS rate of the cohort was 85%.

#### *3.2. LAG-3 Expression on TILs and Its Correlation with Clinicopathological Characteristics*

69% (*n* = 98) of all patients exhibited LAG-3 expression on tumor-infiltrating lymphocytes. No significant correlation could be observed between LAG-3 expression on TILs and age, gender, pT stage, grade, vascular invasion, tumor location or tumor budding.

Interestingly, the percentage of MMR-deficient colon cancers was higher if LAG-3 positive TILs were present in the tumor front or center. Conversely, a lower ratio of MMR-deficient colon cancers was observed in the absence of LAG-3 positive TILs (*p* = 0.034). Additionally, LAG-3 expression on TILs in the tumor center was associated with better differentiation (grade 1, *p* = 0.021) (Tables 1 and 2).

#### *3.3. LAG-3 Expression on TILs and Its Association with DFS*

The presence of LAG-3 expression on TILs either in the tumor front or tumor center was associated with better DFS (5-year DFS 89.9% (LAG-3 positive TILs) versus 74.7% (LAG-3 negative TILs), HR 0.34, 95% CI 0.14–0.80, *p* = 0.009; Figure 2A). Further analysis demonstrated that the favorable association of LAG-3 positive TILs with DFS was restricted to those that were localized at the tumor front (5-year DFS 91.2% versus 75.2%, HR 0.33, 95% CI 0.13–0.82, *p* = 0.012; Figure 2B). Although statistically not significant, there was a trend towards a longer DFS among LAG-3 positive versus LAG-3 negative TILs in the tumor center (5-year DFS 91.3% versus 81.1%, HR 0.42, 95% CI 0.41–1.24, *p* = 0.106; Figure 2C).

**Table 1.** Association of LAG-3 (combined tumor front and center) with clinicopathological features on the stage II colon cancer cohort.


Data are presented as *n* (%), unless otherwise stated. Abbreviations: SD = standard deviation; pT = pathological T stage (TNM classification system); EMVI = extramural vascular invasion; ITBCC = International Tumor Budding Consensus Conference. Bold indicates statistical significant *p*-values

> The favorable association of LAG-3 expression on TILs either in the tumor front or tumor center with the outcome remained significant, even when we considered only MMR-proficient colon cancers (5-year DFS 90.2% versus 67.9%, HR 0.30, 95% CI 0.11–0.83, *p* = 0.014; Figure 3A). Again, the favorable correlation between LAG-3 positive TILs and outcome among MMR-proficient tumors was limited to those localized at the tumor front (Figure 3B), whereas no association with outcome could be observed among LAG-3 positive versus LAG-3 negative TILs in the tumor center (5-year DFS 90.6% versus 75.7%, HR 0.35, 95% CI 0.12–0.99, *p* = 0.039 and 90.4% versus 80.6%, HR 0.44, 95% CI 0.13–1.56, *p* = 0.192, respectively; Figure 3C).

> Due to the low number of events in our cohort of stage II colon cancer, we were not able to conduct a multivariate analysis. However, after adjustment for the pT stage in a bivariate analysis, the favorable effect of LAG-3 expression on DFS remained significant (HR 0.35, 95% CI 0.15–0.83, *p* = 0.017).


**Table 2.** Association of LAG-3 (front/tumor center) with clinicopathological features on the stage II colon cancer cohort.

Data are presented as *n* (%), unless otherwise stated. SD = standard deviation; pT = pathological T stage (TNM classification system); EMVI = extramural vascular invasion; ITBCC = International Tumor Budding Consensus Conference. Bold indicates statistical significant *p*-values.

**Figure 2.** *Cont*.

**Figure 2.** (**A**): The impact of LAG-3 expression on TILs on DFS in patients with stage II colon cancer (tumor center and tumor front). (**B**): LAG-3 expression on TILs at tumor front and its effect on DFS in stage II colon cancer. (**C**): LAG-3 expression on TILs in the tumor center and its impact on DFS in stage II colon cancer.

**Figure 3.** *Cont*.

**Figure 3.** (**A**): The impact of LAG-3 expression on TILs on DFS in patients with stage II MMRproficient colon cancer (tumor center and tumor front). (**B**): LAG-3 expression on TILs at tumor front and its effect on DFS in stage II MMR-proficient colon cancer. (**C**): LAG-3 expression on TILs in the tumor center and its impact on DFS in stage II MMR-proficient colon cancer.

#### **4. Discussion**

To the best of our knowledge, this is the first study evaluating the impact of LAG-3 expression on disease-free survival in stage II colon cancer patients. We demonstrated that LAG-3 expression on TILs was associated with a favorable DFS, especially when LAG-3 positive TILs were identified at the tumor front.

Our results are consistent with previous findings from other studies. Lee et al. found that patients with stage I–III MMR-deficient colon cancer exhibiting LAG-3 positive TILs had a longer DFS compared to those whose MSI tumors did not contain LAG-3 positive TILs [21]. However, in contrast to our study, Lee et al. included only patients with MSI-high tumors ranging from stage I to stage III, whereas our cohort comprised a homogenous series of patients with stage II colon cancer. Additionally, we did not restrict our analysis to patients with MSI stage II colon cancers alone, but also included patients with MSS tumors, representing the majority of stage II cancer patients.

This is particularly important, as we know from previous studies that not only MSI but also MSS tumors may be enriched by immune infiltrates, representing an immunogenic tumor microenvironment (TME) [22]. In a recently published landmark study comprising tumor tissue samples from 2681 patients with stage I–III colon cancer, Pagès et al. could demonstrate that the numbers of CD3+ and cytotoxic CD8+ T cells in the tumor directly correlated with time to recurrence in both the training and validation cohorts, independent of MSI, T and N stages and other clinicopathological factors. A high Immunoscore was not only associated with a longer time to recurrence (TTR) but also translated into better disease-free survival (DFS) and overall survival (OS). Remarkably, patients with a MSI tumor and a high Immunoscore had a similar outcome compared with those who presented with a MSS tumor and a high Immunoscore. Conversely, patients with MSI tumors and a low Immunoscore exhibited a shorter DFS than those with MSS tumors and a high Immunoscore. In the subgroup of stage II colon cancer patients, these associations remained significant [9].

Thus, further characterization of the TME in both MSI and MSS colon cancers is of utmost importance to gain insight into the complex interplay between immune stimulatory and inhibitory effects within the TME to improve our treatment strategies and to better identify patients who benefit most from systemic treatment.

Similarly, Zhang et al. could demonstrate that LAG-3 expression in a mixed cohort of patients with esophageal squamous cell carcinoma encompassing all stages (I–IV) was

associated with improved survival, whereas the favorable effect of high versus low LAG-3 expression on outcome was restricted to stage I–II cancers [23]. In an unselected cohort of patients with stage I –IIIB non-small cell lung cancer, including mainly squamous cell carcinoma and adenocarcinoma but also other histological types such as adenosquamous and large cell carcinoma, LAG-3 expression on TILs was correlated with improved survival [24].

Likewise, another study demonstrated that the presence of LAG-3 positive intraepithelial TILs was associated with a longer disease-specific survival among estrogen receptornegative breast cancer patients [25].

At first glance, it may not seem obvious that the expression of an inhibitory immunecheckpoint molecule such as LAG-3 correlates with a favorable outcome in various solid tumors. Rather, one may assume that the presence of LAG-3 results in increased inhibition of both T cell activation and proliferation and thus contributes to an immune-suppressive tumor microenvironment facilitating tumor growth and metastasis. However, the increased expression of LAG-3 on TILs may not be seen as an independent and 'isolated' immuneinhibiting effect but rather be interpreted as an indicator of an enhanced inflammatory immune response, where TILs are stimulated to exert their antitumor response.

Contrary to our results, Chen et al. reported that patients with stage I–IV colorectal cancer exhibiting a high percentage of LAG-3+ cells in the tumor tissue had a shorter survival compared with those with a low percentage of LAG-3+ cells [26]. However, there are several reasons for these opposite findings. First, the study cohort of Chen et al. comprised 108 patients with a mixture of stage I to stage IV colon and rectal cancers, whereas our cohort consisted of stage II patients with colon cancer alone. Interestingly, Chen et al. found that the percentage of LAG-3+ cells in tumor tissues was significantly higher in stage III and IV colorectal cancers compared to that observed in stage I and II cancers. Given that two-thirds of the patients included in the cohort of Chen et al. had stage III and IV colorectal cancers, it might not be surprising that high versus low LAG-3+ expression in the mixed stage I–IV cohort was associated with shorter survival. Additionally, they could demonstrate that a higher percentage of LAG-3+ cells was associated with poor differentiation, lymph node metastasis and invasion [26], whereas no correlation of LAG-3 expression with any of the clinicopathological characteristics such as tumor grade, vascular invasion or tumor budding could be observed in our study cohort of stage II colon cancer. However, we could demonstrate an association between LAG-3 counts and MMR status. Whereas high LAG-3 expression at the tumor front correlated with microsatellite instability (MSI), this association could not be observed in the tumor center. Second, there is no established scoring method for LAG-3. While Chen et al. divided the cohort into tumors with high and low LAG-3-expressing cells [26], we classified our study cohort according to Sobottka et al. [19], into those tumors exhibiting either LAG+ or LAG- TILs at the tumor front or center. Third, there is a significant diversity of antibodies used for immunohistochemistry across several studies. While we used a rabbit monoclonal antibody (Cell Signaling, clone D2G4O Ref 15372), Chen et al. utilized a different LAG-3 antibody from Abcam (Cambridge, MA, USA) without providing any further information [26].

All these different points mentioned above may partly explain the contradictory findings among our studies. Therefore, it is crucial to develop a standard protocol regarding LAG-3 scoring enabling us to better interpret findings from various studies. Saleh et al. could demonstrate that LAG-3 mRNA expression levels in tumor tissues versus paired normal tissues of colorectal cancer patients were approximately similar [27]. Additionally, Toor et al. could demonstrate in a small and mixed cohort of stage I–IV colorectal cancer patients using a flow cytometry assay that LAG-3 expression on peripheral mononuclear leukocytes was significantly lower compared to the levels observed on both tumorinfiltrating lymphocytes (TILs) and lymphocytes from adjacent normal colon tissue (NILs). Again, no significant difference in LAG-3 expression could be detected between TILs and NILs [28].

With the introduction of immune checkpoint therapies such as PD1 and PD-L1 inhibitors, the prognosis of cancer patients with various malignancies such as melanoma [29],

non-small cell lung cancer [30], renal carcinoma [31], urothelial carcinoma [32], head and neck carcinoma [33] and Hodgkin lymphoma has significantly improved [34]. Currently, there are several clinical trials evaluating the effect of LAG-3 inhibitors in different tumor types [35,36]. Therefore, both the diagnostic and therapeutic value of LAG-3 might be evolving in the near future. Whereas immune checkpoint therapy is associated with improved tumor control and longer survival in patients with MSI metastatic colorectal cancer (mCRC), and thus represents the standard of care treatment [37,38], its effect on outcome in patients with microsatellite stable (MSS) mCRC, who represent 95% of all patients with mCRC, is so far disappointing [39]. Additionally, the role of immune checkpoint therapy as part of the adjuvant treatment strategy in both MSS and MSI stage II and stage III still remains elusive, with several studies ongoing [40,41].

The limitations of this study are its monocentric retrospective design, the small sample size and the lack of an independent clinical validation cohort. Moreover, we restricted our analysis to the immunohistochemical assessment of LAG-3 without using further assays such as in situ hybridization or assessment of ELISA serum LAG-3 concentrations. The latter could not be done due to the lack of blood samples. Additionally, the lack of a consensus guideline regarding immunohistochemical LAG-3 scoring makes it challenging to draw any cross-comparisons between studies. However, our strengths are that we restricted our analysis to a well-defined homogenous cohort of patients with stage II colon cancer. In accordance with Sobottka et al., we performed a binary scoring algorithm for LAG-3 expression. By using a dichotomous scoring method rather than a quantitative scoring method with different thresholds, we may minimize the interobserver variability, increase the reproducibility and facilitate further clinical validation studies in other patient cohorts. Given the lack of a general scoring guideline, this simple binary assessment of LAG-3 allows for efficient, cost-effective and easily reproducible scoring that might be implemented in the diagnostic algorithm of stage II colon cancer, enabling clinicians to decide whether a patient should undergo adjuvant chemotherapy.

#### **5. Conclusions**

In conclusion, we were able to demonstrate that LAG-3 expression on TILs at the tumor front of stage II colon cancers was associated with better outcomes in both the overall stage II cohort and within the subgroup of stage II MSS tumors. Therefore, LAG-3 might serve as a potential prognostic biomarker. However, further studies are needed to explore whether the assessment of LAG-3 in stage II colon cancer may help us to identify those patients who derive the most benefit from adjuvant chemotherapy.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/jpm11080749/s1, Table S1: Baseline characteristics.

**Author Contributions:** Conceptualization, M.D.B.; methodology, G.R.A., J.A.G., I.Z. and M.D.B.; software, I.Z., J.A.G.; formal analysis, I.Z.; investigation, G.R.A., N.A., R.K., H.D., J.A.G., A.L., I.Z. and M.D.B.; resources, N.A., R.K., H.D., A.L., I.Z. and M.D.B.; data curation, I.Z.; writing–original draft preparation, G.R.A., J.A.G., I.Z. and M.D.B.; writing–review and editing, G.R.A., N.A., R.K., H.D., J.A.G., A.L., I.Z. and M.D.B.; visualization, G.R.A., J.A.G., I.Z.; supervision, M.D.B.; project administration, G.R.A., N.A., R.K., A.L., I.Z., M.D.B.; funding acquisition, M.D.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** MDB received a grant from the "Bernese Foundation of Clinical Cancer Research".

**Institutional Review Board Statement:** The study was approved by the research ethics board of the Mount Sinai Hospital (nr 13-0136).

**Informed Consent Statement:** Since this was a retrospective tissue study, no informed consent was required from the research ethics board.

**Data Availability Statement:** The data presented in this study are available from the corresponding author upon reasonable request.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Increased Expression of** *VANGL1* **Is Predictive of Lymph Node Metastasis in Colorectal Cancer: Results from a 20-Gene Expression Signature**

**Noshad Peyravian 1 , Stefania Nobili 2 , Zahra Pezeshkian 1 , Meysam Olfatifar 1 , Afshin Moradi 3 , Kaveh Baghaei 1 , Fakhrosadat Anaraki 4 , Kimia Nazari 1 , Hamid Asadzadeh Aghdaei 1 , Mohammad Reza Zali 5 , Enrico Mini 6,\* and Ehsan Nazemalhosseini Mojarad 5, \***

	- or E.nazemalhosseini@sbmu.ac.ir (E.N.M.)

**Abstract:** This study aimed at building a prognostic signature based on a candidate gene panel whose expression may be associated with lymph node metastasis (LNM), thus potentially able to predict colorectal cancer (CRC) progression and patient survival. The mRNA expression levels of 20 candidate genes were evaluated by RT-qPCR in cancer and normal mucosa formalin-fixed paraffin-embedded (FFPE) tissues of CRC patients. Receiver operating characteristic curves were used to evaluate the prognosis performance of our model by calculating the area under the curve (AUC) values corresponding to stage and metastasis. A total of 100 FFPE primary tumor tissues from stage I–IV CRC patients were collected and analyzed. Among the 20 candidate genes we studied, only the expression levels of *VANGL1* significantly varied between patients with and without LNMs (*p* = 0.02). Additionally, the AUC value of the 20-gene panel was found to have the highest predictive performance (i.e., AUC = 79.84%) for LNMs compared with that of two subpanels including 5 and 10 genes. According to our results, *VANGL1* gene expression levels are able to estimate LNMs in different stages of CRC. After a proper validation in a wider case series, the evaluation of *VANGL1* gene expression and that of the 20-gene panel signature could help in the future in the prediction of CRC progression.

**Keywords:** colorectal cancer; gene signature; mRNA expression; *VANGL1*; FFPE

#### **1. Introduction**

Radical surgery and adjuvant chemotherapy improve the clinical outcome of stage III and high-risk stage II colorectal cancer (CRC) patients. However, it is known that 5-year overall survival (OS) highly varies according to important prognostic factors, such the pTs stage (stages II and III) and the involvement of lymph nodes (pN1 and pN2 stage III) [1–3]. Overall, lymph node metastasis (LNM) is a key prognostic factor for the determination of

**Citation:** Peyravian, N.; Nobili, S.; Pezeshkian, Z.; Olfatifar, M.; Moradi, A.; Baghaei, K.; Anaraki, F.; Nazari, K.; Aghdaei, H.A.; Zali, M.R.; et al. Increased Expression of *VANGL1* Is Predictive of Lymph Node Metastasis in Colorectal Cancer: Results from a 20-Gene Expression Signature. *J. Pers. Med.* **2021**, *11*, 126. https://doi.org/ 10.3390/jpm11020126

Academic Editor: Lisa Salvatore

Received: 26 December 2020 Accepted: 7 February 2021 Published: 14 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

CRC outcomes and significantly relates to poorer prognosis, disease-free survival (DFS), and OS [4,5].

Available clinical data suggest that the accurate diagnosis of LNM is not only important for the prediction of the prognosis of patients but also useful for further therapeutic management, such as for the selection of patients who would benefit from adjuvant/neoadjuvant chemotherapy or chemo-radiotherapy [2,6,7]. In fact, the number and sites of lymph nodes involved have a direct impact on the stage of disease as established by the AJCC tumor-node-metastasis (TNM) staging classification for colon cancer [8].

However, it should be noted that pathological methods are not able to diagnose occult LNMs (micrometastases). Thus, the use of advanced methodologies (e.g., gene expression profiling (GEP)), the investigation of specific biomarkers (e.g., microsatellite instability (MSI)), CpG island methylator phenotype (CIMP), the application of the immune score recommended for increasing the detection power of disease recurrence, and in this regard the use of genes associated with lymph node involvement are very useful to enhance this evaluation [9–12]. In this regard, GEP has been used to discover biomarkers associated with lymph nodes in epithelial neoplasms, such as pancreatic cancer [13], oral squamous cell carcinoma [14], invasive breast cancer [15], and CRC [16]. However, the power of diagnosis may vary based on gene selection.

High-throughput studies in which biomarkers of tumor suppressor genes and oncogenes are potentially able to predict the prognosis of CRC patients at different stages of disease and according to the lymph node involvement are available [4,8,17].

In order to find a suitable biomarker to predict LNM involvement, we evaluated gene expression profiling studies and selected 20 genes (*VANGL1*, *SMAD2*, *BUB1*, *EGFR*, *HES1*, *MAP2K1*, *NOTCH1*, *ANXA3*, *SMAD4*, *MTA1*, *LEF1*, *RHOA*, *TGF-ß*, *CD44*, *CD133*, *IL2RA*, *IL2RB*, *PITX2*, *PCSK7*, and *FOLH1*) that play a key role in carcinogenesis, tumor growth, LNM development tumor invasion, and metastasis by regulating a variety of cellular processes (Table 1) [4,16–20].

**Table 1.** Information \* on the biological functions of 20 candidate genes and on the primer sequences used for RT-qPCR



#### **Table 1.** *Cont.*


**Table 1.** *Cont.*

\* Information available at https://www.ncbi.nlm.nih.gov/ (accessed on 10 February 2021).

Thus, the aim of this study was to identify at the mRNA level and to validate at the protein level the potential prognostic role of these candidate genes in relation to the LNM of CRC patients.

#### **2. Materials and Methods**

#### *2.1. Patients and Sample Collection*

This retrospective study was performed in 100 formalin-fixed paraffin-embedded (FFPE) tumor tissues of CRC patients at stage I or II (*n* = 52) and at stage III or IV (*n* = 48) and 10 FFPE samples of normal tissues (colonic mucosa) as calibrators in RT-qPCR, as well as paired samples of normal colonic tissues from the same CRC patients. All samples were anonymized. Patients underwent surgical resection for CRC from February 1998 to December 2018 at Taleghani Hospital, Shahid Beheshti University of Medical Sciences, Tehran, Iran. They were chemo- and radiotherapy naïve, and none of them experienced previous neoplastic disease. Clinical information, such as colonoscopy/pathology report, follow-up data, and cause of death, was collected from medical records. All patients were carefully followed up to confirm their clinical outcomes.

This study was approved by the ethical committee (IR.SBMU.RIGLD.REC.1396.947) of the Research Institute for Gastroenterology and Liver Disease, Shahid Beheshti University of Medical Sciences, Tehran, Iran. Written informed consent was obtained from all patients. The inclusion criteria for the patients were the following: (1) signed informed consent; (2) availability of the pathology report to confirm the tumor histology; (3) nonresident in an institution, such as a prison, nursing home, or shelter; (4) no severe illness in the intensive care unit; and (5) no preoperative chemotherapy and radiotherapy. The exclusion criteria were the following: patients affected by familial adenomatous polyposis (FAP), hereditary nonpolyposis CRC (HNPCC), cancer at any site at the time of selection, and patients who received neoadjuvant chemotherapy or radiotherapy. The FFPE tissue blocks were cut 10–15 µm and 4–7 µm in thickness for mRNA extraction and immunohistochemistry (IHC), respectively.

To ensure the quality of the presence of tumor and normal cells in FFPE tissue blocks, before performing the laboratory process, each section was evaluated for tumor and normal cells (>80% representative) by the pathologist using hematoxylin and eosin (H&E) staining.

#### *2.2. RNA Isolation*

Ten–fifteen micrometer thick sections were cut from the FFPE blocks, and each section was transferred into a microcentrifuge tube. Deparaffinization was performed with 1 mL xylene, incubating twice for 10 min, and 1 mL absolute ethanol, also incubating twice for 10 min.

#### *2.3. Quantitative and Qualitative Analysis of the Isolated RNA Samples*

Total RNA was extracted from the target tissues using the Rneasy Kit (Qiagen, Chatsworth, CA) according to the company's protocol. To avoid genomic DNA contamination, RNA samples were treated with Dnase I according to the manufacturer's protocol (Invitrogen, Carlsbad, CA, USA). RNA concentration was measured by a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies Inc., Rockland, DE, USA). An A260/A280 ratio was used to evaluate the RNA purity, and values in the range of 1.8–2.0 were accepted.

#### *2.4. Real-Time PCR Analysis*

cDNAs were generated with a PrimeScript RT Reagent kit (Takara, Shiga, Japan) according to the manufacturer's protocol.

The mRNA levels of 20 candidate genes (i.e., *VANGL1*, *IL2RA*, *IL2RB*, *TGF-ß*, *SMAD2*, *SMAD4*, *CD44*, *CD133*, *HES1*, *NOTCH1*, *LEF1*, *MTA1*, *EGFR*, *MAP2K1*, *FOLH1*, *BUB1*, *RHOA1*, *PCSK7*, *PITX2*, and *ANXA3*) (Table 1) and of the housekeeping gene β2-microglobulin (*B2M*) were analyzed by RT-qPCR using the SYBR Fast qPCR Mix Kit (Takara). The cDNA samples were amplified by the 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) with an initial denaturation at 95 ◦C for 30 s, followed by 40 cycles each at 95 ◦C for 5 s and 60 ◦C for 34 s. Relative expression abundances of the target genes were determined by normalizing to *B2M* and *β-actin* using the 2−∆∆Ct method. Each measurement was performed in triplicate. *B2M* was utilized to calculate the relative quantitation (RQ) of mRNA transcripts using the 2−∆∆Ct method.

#### *2.5. Unsupervised Hierarchical Clustering*

An unsupervised hierarchical clustering was used to graphically display the expression levels of 20 candidate genes in CRC samples. Dendrograms and clustering were generated by using the Gene Cluster version 3.0 software and visualized with the Java TreeView version 3.0, available at http://rana.lbl.gov/EisenSoftware.htm (accessed on 1 February 2021) and http://jtreeview.sourceforge.net (accessed on 1 February 2021), respectively. The color of each square box represents the ratio of gene expression. Green boxes indicate upregulated genes, while red boxes represent downregulated genes.

#### *2.6. Immunohistochemistry and Evaluation of Staining*

To investigate the expression levels of candidate proteins, an IHC analysis was performed on slices of FFPE tissues ranging from 4 to 7 µm in thickness. For deparaffinization,

the slides were incubated at 37 ◦C for 24 h and then washed with xylene (100%), ethanol (100%, 85%, and 75%), and distilled water, respectively. After deparaffinization, slides were incubated in a solution of 10% H2O<sup>2</sup> and methanol at a ratio of 1:9 for 15 min and subsequently washed with the distilled water. Next, the slides were treated in the 10 mM citrate buffer solution (pH = 6) and microwaved with 800 W for 24 min and washed with the Tris-buffered saline (TBS). After treating with the blocking serum for 15 min, the slides were immunostained with mouse anti-human MoAbs for 45 min and later washed with TBS. Later, by treating with the EnVision + visualization system (Dako) for 30 min, followed by DAB (Master Diagnosis, LOT. No 090517C1-01) as the chromogen substrate for 10 min, the bound primary antibody was visualized. Finally, the slides were washed with distilled water, dehydrated with ethanol, and stained in hematoxylin. All the slides were independently checked by investigators who had no knowledge of the patients' characteristics and clinical outcome using a microscope (Nikon, Tokyo, Japan).

The analysis of immunostaining intensity was performed using a qualitative scale and ocular observation. Sections were first scanned at low-power magnification (10x) and were quantitatively assessed as follows: under a light microscope at 400x magnifications, five high-power fields (HPFs) were randomly selected, and the immunostaining intensity was determined.

Mean values were estimated through the scanning of the entire tissue sections of all samples using two graded scales: negative, <10%, and positive, >10%. The positive controls were the following: (a) a normal colonic tissue was taken as an internal control for ß-catenin IHC, and (b) a histologically diagnosed section of colon carcinoma tissues for nuclear positivity by ß-catenin IHC. Negative control was achieved by omitting the primary antibody. The MoAbs used in this study were VANGL1 (Abcam, Anti-VANGL1 antibody ab69227), SMAD4 (Abcam, Phospho-SMAD4 antibody T277), EGFR (Master Diagnosis, Anti-EGFR antibody Lot No. 0664000), and LEF1 (Master Diagnosis, Anti-EGFR antibody Lot No. 07430003).

#### *2.7. Statistical Analysis*

Tumors were divided into lymph node metastases (LNMs) and non-lymph node metastases (non-LNMs) based on the histopathological results. The mRNA expression levels of tumor tissues were represented as the mean ± standard deviation (SD). The Mann– Whitney U and Kruskal–Wallis tests were used to assess the differences of the mRNA expressions of 20 genes between the established groups (i.e., presence/absence of LNMs; stage (I–II vs. III–IV), tumor differentiation grade (well vs. poor differentiated), sex (male vs. female), and age (<50 vs. ≥50)). All statistical analyses were performed by the IBM SPSS Statistics software version 22 (IBM, SPSS, Chicago, IL, USA) and Stata analyzer.

For receiver operating characteristic (ROC) curve analysis, the R 3.6.1 software was used to evaluate the sensitivity and specificity of the prognosis prediction (evaluated by OS) according to the mRNA gene expression by analyzing the area under the curve (AUC). Stratification of patients in high and low tumor gene expression was established according to the cutoff obtained for each gene by ROC analysis. OS analysis was performed by plotting Kaplan–Meier (log-rank test) curves. *p*-Values < 0.05 were considered statistically significant.

#### **3. Results**

#### *3.1. Clinical and Pathological Characteristics of Patients*

The population study consisted of 100 FFPE tissues from CRC patients (59 men and 41 women with an average age of 52.17 years, 20–78 range). The clinical features of the study population are shown in Table 2. Information on age, sex, stage, tumor differentiation, and tumor location is available for all patients. Among patients, 52% had stage I or II CRC, while 48% of the cases had stage III or IV. Of 100 patients, 37 were positive and 63 were negative for LNM.


**Table 2.** Clinical and pathological characteristics of patients.

#### *3.2. Gene Expression Analysis*

To identify molecular determinants of LNMs, gene expression profiles from patients with or without LNMs and at different stages of disease were compared. Based on literature data, we selected 20 genes that relate to the lymphatic metastatic process and evaluated their expression levels in 100 FFPE blocks.

Relationships of tumor gene expression with demographic (sex, age), clinical (tumor location), and pathological (stage, LNM, grade) features are reported in Tables 3 and 4.

**Table 3.** Relationships between the expression of 20 CRC study genes and age and sex.



**Table 3.** *Cont.*

**Table 4.** Relationships between the expression of 20 CRC study genes and clinical and pathological characteristics of patients.


In particular, the gene expression levels of *VANGL1* varied significantly between patients with and without LNMs. The tumors of patients with LNMs displayed twofold

higher levels of *VANGL1* mRNA expression compared with those of patients without LNMs (*p* = 0.02) (Table 4 and Figure 1). Additionally, the expression levels of this gene varied between patients with stages I-II and III-IV, showing the highest mean level (i.e., 8.831) for stages III and IV. This difference reached a good level of significance, although not fully significant (*p* = 0.05) (Table 4 and Figure 1).

**Figure 1.** *VANGL1* mRNA relative quantification (RQ) established by RT-qPCR analysis according to lymph node metastasis (LNMs) involvement or stage. Gene expression levels of *VANGL1* differed significantly between patients with and without LNMs.

The mRNA expression levels of three genes (i.e., *IL2RB*, *SMAD*, and *ANXA3*) were significantly (*p* < 0.05) different between well-differentiated (i.e., G1 and G2) and poorly differentiated (i.e., G3 and G4) tumors (Table 4 and Figure 2).

**Figure 2.** *ANXA3* and *SMAD2* mRNA relative quantification (RQ) established by RT-qPCR analysis according to histological grade. Gene expression levels of these genes differed significantly between well- and poorly differentiated cancers.

We found significant associations between tumor mRNA expression of *IL2RB* and *NOTCH1* genes and gender and between tumor mRNA expression of *IL2RA* and *MAP2K1* genes and age. In particular, the *IL2RA* gene was significantly downregulated in patients younger than 50 years old compared with patients older than 50 years (Figure 3). Additionally, an increased expression of the *MAP2K1* gene was observed in patients older than

50 years in comparison with patients younger than 50 years old (*p* = 0.02). *IL2RB* exhibited lower expression in females compared with males (*p* = 0.02) (Table 3 and Figure 4). Conversely, the *NOTCH1* gene was significantly upregulated in female patients as compared with male CRC cases (*p* = 0.02).

**Figure 3.** *IL2RA* and *MAP2K1* mRNA relative quantification established by RT-qPCR analysis according to age. Gene expression levels of these genes differed significantly between patients younger than 50 years and patients older than or equal to 50 years.

**Figure 4.** *IL2RB* and *NOTCH1* mRNA relative quantification (RQ) established by RT-qPCR analysis according to gender. Gene expression levels of these genes differed significantly between males and females.

#### *3.3. Heat Maps of Real-Time PCR Data*

Hierarchical clustering of 100 CRC samples is reported in Figure 5. According to the diagram, the *VANGL1*, *PCSK7*, and *ANXA3* genes showed the highest expression levels in most CRC samples.

**Figure 5.** Heat maps of real-time RT-qPCR data representing gene expression variations of the 20 analyzed transcripts in FFPE CRC samples. Green indicates upregulation, and red indicates downregulation.

#### *3.4. ROC Analysis*

The predictive performance of the 20-gene signature was assessed by computing the AUC value of the ROC curve. A logistic regression model was built based on the comparison of tumor samples (*n* = 100) in relation to the following study patient characteristics: stage I–II vs. III–IV and presence vs. absence of LNM. We selected two panels including 5 and 10 genes based on the genes that had the highest AUC and showed a more effective role in CRC progression. One panel included 5 genes (*VANGL1*, *IL2RA*, *SMAD2*, *RHOA1*, and *HES*), and the other 10 genes (previous 5 plus *MTA1*, *CD133*, *FOLH1*, *NOTCH1*, and *TGF-ß*). The total number of genes (i.e., 20-gene panel) was also analyzed.

Figure 6 summarizes the performances of the study gene panels for the prediction of stage in the patient cohort, with the 20-gene panel achieving the highest performance. The AUC value for the 5-gene panel was 68.39%, along with 95% CI, 57.81%–78.97%; 67.30% sensitivity; and 66.66% specificity (Figure 6A); for the 10-gene panel, the AUC was 71.67% (95% CI, 61.51%–81.84%; sensitivity, 61.53%; and specificity, 72.91%) (Figure 6B). The analysis of the total 20 genes resulted in AUC = 78.85% (95% CI, 69.94%–87.75%; sensitivity, 75%; and specificity, 77.08%) (Figure 6C). In Figure 6D, the AUCs of 20-, 10-, and 5-gene panels in relation to stage are reported. When all the three AUCs (5-/10-/total-gene panels) were compared together, these results showed a trend towards significance (*p* = 0.055). When the AUC of the total number of genes was compared with that of the 5 genes, the difference was highly significant (*p* = 0.02). A statistical trend was observed between the AUC of the total panel and that of the 10-gene panel (*p* = 0.08), while no significant

difference between the AUC of the 10-gene panel and that of the 5-gene panel was noted (*p* = 0.34).

**Figure 6.** Comparison of the predictive performance by receiver operating characteristic (ROC) curve analysis for stage. (**A**) The AUC assessment of the logit(*p*) value for the panel of 5 genes. (**B**) The AUC assessment of the logit(*p*) value for the panel of 10 genes. (**C**) The AUC assessment of the logit(*p*) value for the panel of total genes. (**D**) Comparison of the predictive performance for the panel of 5, 10, and 20 genes.

Figure 7 summarizes the performances of the study gene panels for the prediction of LNMs in CRC patients, and also in this case, the 20-gene panel achieved the highest performance. The AUC value was 70.19% (95% CI, 59.18%–81.02%; sensitivity, 84.12%; specificity, 57.86%) when the gene expression of the 5-gene panel was compared in relation to LNM and non-LNM CRC patients (Figure 7A). Comparison of the gene expression of the 10-gene panel in relation to LNM and non-LNM CRC patients resulted in AUC = 71.47% (95% CI, 60.62%–82.32%; sensitivity, 80.95%; specificity, 59.45%) (Figure 7B). The AUC of the total genes in LNM vs. non-LNM CRC patients was the highest of the three AUCs obtained (i.e., 79.84% (95% CI, 70.38%–89.30%; sensitivity, 74.60%; specificity, 75.67%) (Figure 7C). As far as the association with LNMs was concerned, the comparison between all the AUCs together (5-/10-/total-gene panels) pointed out a nearly significant difference (*p* = 0.05). In particular, the AUC of the total-gene panel was significantly higher compared with that of the 5-gene panel (*p* = 0.03) in relation to LNM and non-LNM CRC patients (Figure 7D). A high statistical trend was observed between the AUC of the total gene panel and that of the 10-gene panel (*p* = 0.06), although no statistical difference was observed between the AUC of the 10-gene panel and that of the 5-gene panel (*p* = 0.38). We also analyzed the predictive performance of single genes according to stage and LNM (Tables S1 and S2). Data showed that the *VANGL1* gene was a significant predictor for LNMs with an AUC of 63.99 (95% CI, 52.41%–75.56%; sensitivity, 80.95%; specificity, 45.94).

**Figure 7.** Comparison of the predictive performance by receiver operating characteristic (ROC) curve analysis for lymph node metastasis (LNMs). (**A**) The AUC assessment of the logit(*p*) value for the panel of 5 genes. (**B**) The AUC assessment of the logit(*p*) value for the panel of 10 genes. (**C**) The AUC assessment of the logit(*p*) value for the panel of total genes. (**D**) Comparison of the predictive performance for the panel of 5, 10, and 20 genes.

#### *3.5. Correlation of Gene Expression with Overall Survival*

All patients completed their follow-up by 20 December 2018 (median, 10.8 years, and range, 0.019–21 years). Patients whose tumors expressed higher levels of *NOTCH1* mRNA or lower levels of *IL2RB* mRNA showed a statistically significant prolonged OS compared with their respective counterparts (*p* = 0.042 and *p* = 0.043, respectively) (Figure 8). No other statistically significant correlation was found between OS and expression levels of the other study genes.

**Figure 8.** Association between *NOTCH1* (**A**) and *IL2RB* (**B**) expression and overall survival. The median follow-up was 10.8 (0.95LCL–0.95UCL; 10–11) years. HR, hazard ratio.

#### *3.6. Immunohistochemistry Analysis*

The protein expression levels of four genes that play a critical role in cancer development and progression was evaluated by using IHC. Twenty-five percent of CRC FFPE and normal matched tissues were used in this regard. In particular, we were interested in the evaluation of the expression levels of the products of *VANGL1*, *EGFR*, and *SMAD4* based on literature data and on the relationships we observed between the expression of their respective encoding genes and the clinical/pathological characteristics of the study CRC patients. The fourth protein we selected was LEF1. Although we did not find relationships between *LEF1* gene expression and the clinical/pathological parameters of CRC patients, we were interested in evaluating the potential role of LEF1 as an early biomarker of colorectal carcinogenesis since its activation by MYC has been associated with the activation of the WNT pathway signaling. The protein expression of VANGL1, EGFR, LEF1, and SMAD4 via their antibodies were examined. Results showed that VANGL1 and EGFR proteins were overexpressed (more than 50% of the stained cells in colon adenocarcinoma tissues compared with the normal tissues) (Figures 9 and 10). Additionally, immunohistochemical staining revealed a predominantly nuclear localization of SMAD4 and LEF1, and they showed higher expression in CRC tissues compared with normal colonic mucosa with more than 50% and 20% of the stained cells, respectively (Figures 10 and 11).
