somatic mutational landscapes of adherens junctions and their … · 1 1 somatic mutational...

1

Somatic mutational landscapes of adherens junctions and their 1

functional consequences in cutaneous melanoma development 2

3

Praveen Kumar Korla,1 Chih-Chieh Chen,2 Daniel Esguerra Gracilla,1 Ming-Tsung Lai,3 Chih-4

Mei Chen,4 Huan Yuan Chen,5 Tritium Hwang,1 Shih-Yin Chen,4,6,* Jim Jinn-Chyuan Sheu1,4, 6-9,* 5

1Institute of Biomedical Sciences, National Sun Yat-sen University, Kaohsiung 80242, Taiwan; 6

2Institute of Medical Science and Technology, National Sun Yat-sen University, Kaohsiung 80424, 7

Taiwan; 3Department of Pathology, Taichung Hospital, Ministry of Health and Welfare, Taichung 8

40343, Taiwan; 4Genetics Center, China Medical University Hospital, Taichung 40447, Taiwan; 9

5Institute of Biomedical Sciences, Academia Sinica, Taipei 11574, Taiwan; 6School of Chinese 10

Medicine, China Medical University, Taichung 40402, Taiwan; 7Department of Health and 11

Nutrition Biotechnology, Asia University, Taichung 41354, Taiwan; 8Department of 12

Biotechnology, Kaohsiung Medical University, Kaohsiung 80708, Taiwan; 9Institute of 13

Biopharmaceutical Sciences, National Sun Yat-sen University, Kaohsiung 80242, Taiwan 14

15

PKK, CCC and DEG contributed equally to this study. 16

*Correspondence to: Dr. Shih-Yin Chen ([email protected]) at Genetics Center, China 17

Medical University Hospital, Taichung, 40447, TAIWAN; or Dr. Jim Jinn-Chyuan Sheu 18

([email protected]) at Institute of Biomedical Sciences, National Sun Yat-sen 19

University, Kaohsiung City 80424, TAIWAN. 20

21

Running title: mutational landscape of cadherins in melanoma 22

2

Abstract 23

Cell-cell interaction in skin homeostasis is tightly controlled by adherens junctions (AJs). 24

Alterations in such regulation lead to melanoma development. However, mutations in AJs and 25

their functional consequences are still largely unknown. 26

Methods: Cadherin mutations in skin cutaneous melanoma were identified using sequencing data 27

from TCGA dataset, followed by cross-validation with data from non-TCGA cohorts. Mutations 28

with significant occurrence were subjected to structural prediction using MODELLER and 29

functional protein simulation using GROMACS software. Neo-antigen prediction was carried out 30

using NetMHCpan tool. Cell-based fluorescence reporter assay was used to validate β-catenin 31

activity in the presence of cadherin mutations. Clinical significance was analyzed using datasets 32

from TCGA and other non-TCGA cohorts. Targeted gene exon sequencing and 33

immunofluorescence staining on melanoma tissues were performed to confirm the in silico 34

findings. 35

Results: Highly frequent mutations in type-II classical cadherins were found in melanoma with 36

one unique recurrent mutation (S524L) in the fifth domain of CDH6, which potentially destabilizes 37

Ca2+-binding and cell-cell contacts. Mutational co-occurrence and physical dynamics analyses 38

placed CDH6 at the center of the top-four mutated cadherins (core CDHs; all type-II), suggesting 39

altered heterophilic interactions in melanoma development. Mutations in the intracellular domains 40

significantly disturbed CDH6/β-catenin complex formation, resulting in β-catenin translocation 41

into cytosol or nucleus and dysregulation of canonical Wnt/β-catenin signaling. Although 42

mutations in core CDH genes correlated with advanced cancer stages and lymph node invasion, 43

the overall and disease-free survival times in those patients were longer in patients with wild-type. 44

Peptide/MHC-I binding affinity predictions confirmed overall increased neo-antigen potentials of 45

3

mutated cadherins, which associated with T-lymphocyte infiltration and better clinical outcomes 46

after immunotherapy. 47

Conclusion: Changes in cell-cell communications by somatic mutations in AJ cadherins function 48

as one of mechanisms to trigger melanoma development. Certain mutations in AJs may serve as 49

potential neo-antigens which conversely benefit patients for longer survival times. 50

51

Keywords: adherens junctions (AJ), cadherin-6 (CDH6), somatic mutation, melanoma, neo-52

antigen 53

54

Graphical Abstract 55

56

Heterophilic interactions between melanocyte and keratinocyte maintain skin homeostasis. UV-57

induced CDH mutations cause contact control loss from keratinocyte and melanomagenesis. 58

During metastasis, mutation-derived Neo-Ags boost T-lymphocytes and activate anti-cancer 59

immunity. 60

4

Introduction 61

Cutaneous melanoma is the most dangerous form of skin cancer with high invasive and metastatic 62

potentials, resulting in more than 55,000 deaths per year worldwide [1] . Even though the cure rate 63

for patients with localized tumors is high, the overall five-year survival rates could be very low 64

(<20%) for those in whom the spread has occurred, indicating critical roles of cell-cell and cell-65

matrix crosstalk in disease development. In normal epidermis, melanocytes or their stem cells form 66

multiple contacts with keratinocytes, which in turn establish regulatory networks in paracrine 67

manners to fine-tune melanocyte growth and epidermal homeostasis [2, 3]. During 68

melanomagenesis, melanoma cells alter such regulatory contacts via extracellular matrix 69

remodeling and then promote dermal invasion by activating inflammation-dependent cell 70

migration/translocation [4]. These findings suggest that loss of the control by keratinocytes enables 71

melanocyte undergoing uncontrolled cell growth, leading to cell transformation and melanoma 72

development. 73

Adherens junctions (AJs) are cadherin-based dynamic structures which link the physical 74

contact from the extracellular matrix to cytoskeleton networks via complex formation between 75

cadherins and catenins (, and p120-catenin). Dissociation of -catenin by growth factor 76

stimulation leads to loss of cadherin-mediated cell-cell contact and increase of -catenin levels in 77

cytoplasm and nucleus, which subsequently promotes cell proliferation via activating TCF/Wnt 78

signaling [5, 6]. On the other hand, AJs also regulate stress-sensing and force-transmission through 79

signaling cascades, known as mechanotransduction [7, 8]. Clustering of core cadherin-catenin 80

complexes can further strengthen cell-cell adhesion, allowing cadherins to nucleate signaling hubs 81

and establish the regulatory Hippo network [9, 10]. Breakdown of Hippo network by up-regulating 82

expression and nuclear translocation of YAP/TAZ, the nuclear mechanosensor/transducer, was 83

5

found during melanoma progression, resulting in invasion, metastasis and drug resistance [11-14]. 84

Accordingly, AJ-mediated mechanical and cytoskeletal signaling are important for melanoma to 85

create 3D microenvironment for further progression. 86

Recent studies by using lineage-tracing approach or global gene expression analyses 87

suggested that skin cutaneous melanoma cells may originate from adult melanocyte stem cells and 88

show feature properties of multipotent progenitors [2, 4]. Within the specialized anatomical 89

microenvironments, both hemophilic (via type-I cadherins) and heterophilic (via type-II cadherins) 90

cell-cell interactions are required for melanocyte stem cells to develop new cell partnerships with 91

stroma and/or endothelial cells. For examples, cadherin switching from N-cadherin (CDH2) to K-92

cadherin (CDH6) was found critical for neural crest cell formation, emigration and detachment 93

from the neural tube to be melanocyte [15, 16]. Furthermore, down-regulation of key junction 94

proteins for communication between melanocytes and keratinocytes frequently occur during 95

melanoma development, including E-cadherin (CDH1) and P-cadherin (CDH3) [2, 17, 18]. 96

Alterations in such interacting network enable melanocyte stem cells to proliferate and migrate out 97

of the niche after activation or transformation by extrinsic stimuli like UV exposure [2, 4]. 98

With advanced whole genome sequencing and other integrative technologies, a set of major 99

driver mutations in cutaneous melanoma, e.g. BRAF, RAS and NF1, have been discovered which 100

contribute to activation of mitogen-activated protein kinase and phosphoinositol kinase signaling 101

[19, 20]. Notably, immune signatures with higher lymphocyte infiltration have been correlated 102

with improved survival outcomes but are irrespective of the genomic subtypes in patients [19, 21]. 103

These data indicate the need to further explore certain genes that determine microenvironment 104

remodeling and lymphocyte recruitment through cell-cell/cell-matrix interactions. In this study, 105

6

we aimed to study somatic mutational landscapes of key AJ genes in melanoma and characterize 106

functional impacts of those mutations with molecular and clinical meanings. 107

7

Materials and Methods 108

Data collection and functional relevance with mutations in classical cadherin genes 109

Somatic mutation, gene expression and clinical information data of individual cadherins in skin 110

cutaneous melanoma were collected from cBioPortal databank (https://www.cbioportal.org/) Data 111

from TCGA cohort (TCGA, PanCancer Atlas; n = 448) were utilized as the studying cohort, and 112

the resultant data were validated by using data from non-TCGA cohort (n = 396). The functional 113

interactome among classical cadherins were analyzed by using STRING protein-protein 114

interactome database (http://string-db.org/). Oncoprinter and Mutation mapper tools in cBioportal 115

database were performed to analyze mutation types, hotspot mutation sites, and the conserved 116

mutation sites in cadherin and CTNNB1 genes [22]. To map the key pathways associated with 117

cadherin mutations, normalized gene expression data were applied for gene set enrichment analysis 118

(GSEA) by using GSEA Java (version 2.2.3) from Broad Institute [23]. Prediction of peptide-119

binding affinity onto MHC-I complex, with the peptide length of nine amino acids, was performed 120

to estimate the neo-antigen potentials of mutations in cadherin genes by using NetMHCpan 121

algorithm [24]. Mutation-containing peptides with antigen-tolerance score less than 1.0 were 122

considered as neo-antigens. 123

124

Tissue sample collection for targeted exon sequencing and immunofluorescence staining 125

Twenty-two skin cutaneous melanoma tissue blocks were collected from Department of Pathology, 126

Taichung Hospital, Ministry of Health and Welfare, Taiwan with an approved IRB study (109028) 127

by the Institutional Review Board of Tsaotun Hospital, Ministry of Health and Welfare, Taiwan. 128

Genomic DNA was extracted from those tissue blocks using QIAamp DNA FFPE Tissue Kit 129

(Qiagen, Venlo, Netherlands), followed by PCR enrichment to establish final DNA libraries for 130

8

targeted exon sequencing (cadherin genes involved in the core interactome) on a HiSeq 2000 131

(Illumina, San Diego, CA). The raw sequencing data were filtered by Cutadapt to obtain clean data 132

for further quality control checks [25]. By using SAMtools, somatic mutations occurred in classic 133

CDHs were identified with an average depth more than 120 counts and an average coverage more 134

than 60% on each chromosome [26]. Among the 22 tissues, 9 of them showed mutations in 135

classical cadherin genes in the Taiwanese cohort (Table S1). 136

For immunofluorescence staining, tissue sections were dewaxed by xylene, followed by 137

alcohol fixation and rehydration. After antigen retrieval in 10 mM citric acid buffer (pH 6.0) and 138

blocking with 5% BSA, tissue sections were incubated with Abs against different core CDHs and 139

co-stained with anti-β-catenin Abs (Cat# MAB14489; Abnova Corp., Taipei, Taiwan) at 4oC for 140

overnight. The Abs against core CDHs were anti-CDH6 (PAB31388; Abnova Corp.), anti-CDH9 141

(PA5-106548; Thermo Fisher Scientific Inc., Waltham, MA), anti-CDH10 (ab196662; Abcam 142

PLC., Cambridge, UK) and anti-CDH18 (A14645; Abclonal Inc., Woburn, MA) Abs. For T-143

lymphocyte infiltration study, anti-CD3 Abs (A19017; Abclonal Inc.) were utilized to co-stain 144

with anti-β-catenin Abs. Secondary Ab staining was carried out by using FITC-conjugated anti-145

rabbit IgG (for core CDHs and CD3; sc-2359; Santa Cruz Biotech. Inc., Dallas, TX) and CFL555-146

conjugated anti-mouse IgG (for β-catenin; sc-362267; Santa Cruz Biotech. Inc.) Abs at 37oC for 1 147

hour. Auto-fluorescence were quenched using True VIEW auto-fluorescence quenching kit 148

(Vector Lab., Burlingame, CA). Cell nuclei in the tissues were counterstained with DAPI and the 149

images were taken under an IX83 fluorescence microscope (Olympus Corp., Shinjuku, Japan). 150

151

Structural prediction of the cadherin-6 (CDH6) protein 152

9

The three-dimensional structures of extracellular cadherin domain 5 (EC5; residues 487-608), 153

juxtramembrane domain (JMD; residues 660-677), and catenin-binding domain (CBD; residues 154

687-784) of CDH6 protein were predicted by using the MODELLER program (version 9v8) [27]. 155

In this study, CDH6 (PDB ID: 5VEB) was selected as the template for predicting the structure of 156

S524L-EC5, p120-catenin in complex with E-cadherin (PDB ID: 3L6X) was selected as template 157

for predicting the structures of WT-, D661N-, and E662K-JMD in complex with p120-catenin, 158

and β-catenin with desmosomal-cadherin (PDB ID: 3IFQ) was selected as the other template for 159

predicting the structures of WT-, R689Q-, D691N-, E722K-, D726N-, S749F-, R773Q-CBD in 160

complex with β-catenin. The qualities of all modeled structures were assessed through 161

Ramachandran plots. The steepest descent algorithm was used for structural optimization, and 162

PyMOL (http://www.pymol.org/) was used for all three-dimensional (3D) structures and charge 163

distribution presentation. 164

165

Molecular dynamics simulations 166

Molecular dynamics (MD) simulations were performed using GROMACS software (version 5.1.4) 167

[28]. The OPLS-AA force field was used for energy calculations. The models were solvated with 168

TIP3P water molecules and simulated in a cubic box with periodic boundary conditions. The 169

energy of the models was first minimized using the steepest descent algorithm. The simulations 170

were performed in the canonical NVT and NPT ensembles through the position-restrained MD 171

simulation for 100 ps, respectively. The linear constraint solver algorithm was used to constrain 172

the bonds. The MD simulations were performed at constant pressure and temperature for 30 ns 173

using an integration time step of 2 fs. The non-bonded interactions were cutoff at 10 Å. The 174

coordinates from the MD simulations were saved every 10 picosecond. 175

10

176

Binding free energy calculation 177

The pmx tool was used to construct the hybrid topologies of the system used in the free energy 178

calculations [29]. This tool automatically generates hybrid structures and topologies for amino 179

acid mutations that represent the two physical states of the system. The double system (combine 180

both branches of a thermodynamic cycle) in a single simulation box was used as the setup to keep 181

the system neutral at all times during a transition (Figure S1) [29]. After obtaining this hybrid 182

structure, free energy simulations were performed using GROMACS. From the 10-ns equilibrated 183

trajectories, 100 snapshots were extracted, and a rapid 100-ps simulation was performed starting 184

from each frame. The lambda (λ) ranged from 0 to 1 and from 1 to 0 for the forward and backward 185

integrations, respectively, thus describing the interconversion of the WT to MT systems and the 186

MT to WT systems, respectively. Finally, the pmx tool was used to integrate over the multiple 187

curves and subsequently estimate free energy differences using the fast-growth thermodynamic 188

integration approach, which relies on the Crooks–Gaussian Intersection (CGI) protocol [30]. In 189

such a double-system/single-box setup, the estimated free energy corresponds to the ΔΔG of 190

binding (Figure S1). 191

192

Clinical association study and statistical analyses 193

Clinical data of melanoma patients, including tumor stage, lymph node invasion, metastasis, 194

lymphocyte score and survival time, were retrieved from TCGA database. GraphPad Prism 195

(GraphPad software, La Jolla, CA) was utilized for statistical analyses. Patients were divided into 196

WT CDHs (no mutation in any cadherin genes) and MT core CDHs (patients with mutation in top-197

four highly-mutated cadherin genes) groups. Statistical differences between two groups were 198

11

estimated by t-test. Kaplan–Meier method was used to estimate survival curves, and a log-rank 199

test was performed for statistical comparison between two groups. To compare neo-antigen 200

potentials, two-way ANOVA was utilized to compare differences between two groups with 201

multiple mutation sites. 202

203

Gene cloning and site-directed mutagenesis 204

The gene segment of β-catenin was amplified by nested PCR using cDNA from human melanoma 205

A375 cells. After purification, the gene segment was further sub-cloned into pmCherry-N1 vector 206

(Clontech Lab., Madison. WI), thus the gene is tagged with an mCherry at C-terminal end. Human 207

CDH6 cDNA (OriGene Technologies, Inc., USA) was amplified and sub-cloned into pAcGFP1-208

N1 vector (Clontech Lab.) to generate GFP-tagged wild type CDH6. CDH6 with mutations in the 209

CBD domain were generated by site directed mutagenesis (F541; Thermo Fisher Scientific Inc.). 210

DNA sequencing was done to confirm the gene sequences and the introduced substitutions that are 211

in-frame with the fusion partners. 212

213

Cell-based complex formation assay between -catenin with CDH6 and its mutant variants 214

To monitor complex formation in live cells, -catenin-mCherry and CDH6 variants gagged with 215

GFP were co-transfected into A375 cells using Lipofectamine 2000 (Thermo Fisher Scientific 216

Inc.) by 1:1 molar ratio. Complex formation and cellular localization were determined by an 217

IX83 fluorescent microscope (Olympus Corp., Shinjuku, Japan) 24 hours after gene transfection. 218

Cell image and fluorescent intensity were analyzed using Image-Pro Premier (Media Cybernetics 219

Inc., Rockville, MD). 220

12

Results 221

Highly-frequent mutations in adherens junction (AJ) cadherin genes in melanoma 222

Dysregulation in interactions between melanocyte and keratinocyte or melanocyte and matrix is a 223

key step during melanoma malignant transition. We therefore wanted to know mutational profiles 224

of major adhesion molecules, mainly adherens junction (AJ) molecules, in melanoma and their 225

biological impacts on cancer progression (Figure 1). A total of 20 classical cadherins that 226

participate in cell-cell interactions are identified in human genome [31]. STRING protein-protein 227

interactome analyses defined a compact network composed by 16 classical cadherins (6 type-I and 228

10 type-II) that determines AJ organization, homophilic cell adhesion via plasma membrane 229

adhesion molecules, and cell junction assembly with significant scores (q-values < 10-5) (Figure 230

2A). TCGA-melanoma data with 448 sequenced tumor samples were utilized to analyze mutation 231

statuses of those 16 genes by using cBioportal databank [32]. Surprisingly, frequent mutations (> 232

10%) were found in 5 classical cadherin genes (CDH6, CDH18, CDH10, CDH9 and CDH4) 233

(Figure 2B). Although no mutually exclusive pattern, genetic alterations in the classic cadherin 234

genes sum up a total mutation rate of 56% in melanoma tissues. Consistent with previous findings 235

indicating rare E-cadherin (CDH1) mutations in cancer [33], only 2.7% of melanomas carry CDH1 236

mutations. In addition, no mutation was found in P-cadherin (CDH3). Mutation rates in other 237

minor AJ genes including nectins and nectin-like molecules were also relatively low (< 10%) 238

(Figure S2), suggesting that mutations in classical cadherins play more important roles in 239

melanoma development. Such conclusions can be further validated by using sequencing data from 240

non-TCGA cohort, especially for CDH6, CDH9, and CDH18 genes which also show mutation 241

rates larger than 10% (Figure S3). Significantly, the top-four mutated genes (core CDHs) are all 242

type-II classical cadherins, suggesting alterations in heterophilic interactions may function as one 243

13

possible mechanism to drive melanoma development. At nucleotide levels, most of the detected 244

substitutions (around 90%) were G to A or C to T, a signature for DNA damage repair after UV-245

induced DNA adducts [19] (Figure 2C), indicating the etiological relevance of mutations in those 246

adhesion genes during melanoma development [4]. 247

248

Cadherin mutations in melanoma development 249

Mutational profiles revealed that somatic mutations in cadherins could be detected at a variety of 250

positions throughout coding regions with the existence of some truncations and out-of-frame 251

insertions/deletions (Figure 3A, left panel). The feature of most detected somatic mutations 252

suggests a tumor suppressor role for those adhesion proteins in cancer development [34]. 253

Nevertheless, the S524L substitution on the fifth extracellular (EC5) domain of CDH6 (K-cadherin) 254

stood out to be a unique mutation with a relatively higher frequency, indicating its potent role in 255

melanoma development. Especially, this recurrent missense mutation can be confirmed by the non-256

TCGA cohort (Figure 3A, right panel). On the other hand, mutations in the intracellular domains 257

of top-five cadherins show mutual exclusivity with β-catenin (CTNNB1) mutations (Figure 3B), 258

suggesting potent effects of cadherin mutations on canonical Wnt/β-catenin signaling. 259

Co-occurrence analysis among the core CDHs (CDH6, CDH9, CDH10 and CDH10) 260

indicated the association of CDH6 mutations with mutations in CDH9, CDH10 and CDH18 261

(Figure 3C). Gene expression profiling using the GEO dataset (GDS1965) indicated relatively high 262

levels of core CDHs during melanoma malignant transition (Figure 3D). Such expression patterns 263

can be also confirmed at protein levels by immunofluorescence staining (Figures S4-S7). Through 264

biophysical dynamics approach, previous study confirmed strong heterophilic interactions among 265

these core CDHs, and CDH6 was the only one that can form stable hemophilic interactions with 266

14

other three [35]. Interestingly, CDH6 was also the only one to be expressed predominantly in 267

normal melanocytes but not in normal squamous epithelial cells (Figure S4, upper). The above 268

data suggested CDH6 as the key player among the core CDHs during melanoma development and 269

progression (Figure 3E). Thus, mutations in type-II cadherins, especially in CDH6, may gain 270

functional advantages during melanoma development. 271

272

Structural dynamics of the S524L substitution on the fifth extracellular (EC5) domain of 273

CDH6 (K-cadherin) and the potential impacts on Ca2+-binding/stabilization 274

To know functional impacts of mutations in CDH6, the structure of the EC5 domain (CDH6-EC5) 275

was constructed by MODELLER program using the published human CDH6 crystal structure 276

(Protein Data Bank ID: 5VEB) as the template. To compare the similarity in three-dimensional 277

structure of the EC5, Root-Mean-Square Deviation (RMSD) analyses indicated longer average 278

distances between the atoms in the S524L mutant as compared to wild type (WT) CDH6, 279

suggesting its impacts on the EC5 structure (Figure 4A, upper). Analyses of the residue-specific 280

Root-Mean-Square Fluctuations (RMSF) (Figure 4A, lower) further revealed higher atomic 281

fluctuations of the S524L mutant at the β2/β3 loop, a region for Ca2+-binding, when compared 282

with the WT. The 4/β5 loop, which was predicted to stabilize the Ca2+-binding, was also found 283

less stable in this mutant due to its higher RMSF and B-factors (Figure 4A-B). To confirm this, 284

free energy differences of Ca2+-binding between WT and S524L (∆∆𝐺𝐵𝑖𝑛𝑑𝑖𝑛𝑔𝑆524𝐿 ) was calculated in a 285

series of alchemical free energy simulations by using CGI protocol [30]. The double-free energy 286

was calculated according to ΔG1 (Ca2+-bound) – ΔG2 (Ca2+-free), as shown in the thermodynamic 287

cycle (Figure 4C). The data indicated that the Ca2+-binding affinity toward CDH6-EC5 was 288

reduced by S524L substitution (ΔΔG = 4.4 kJ/mol). These analyses conclude that S524L 289

15

substitution may induce additional fluctuations in the β2/β3 and β4/β5 loop regions, which can 290

promote instability in this region and thus hamper Ca2+-binding. 291

292

Mutations in the intracellular regions of cadherins (CDH-C) and their effects on catenin-293

binding 294

The process of AJ assembly to establish cell-cell contacts involves complex formation between 295

cadherins and catenins. To study functional roles of mutations in the intracellular domain (CDH-296

C), we performed sequence alignment across cadherins involved in the cadherin interactome core. 297

Based on structural and functional analyses, CDH-C can be divided into two major domains: the 298

juxtramembrane domain (JMD) and the catenin-binding domain (CBD) that can bind to p120-299

catenin (-catenin) and -catenin, respectively (Figure 5A). Most of mutations (>85%) were found 300

involved in the binding regions or conversed residues, indicating potential impacts on cytoplasmic 301

complex formation. We next predicted 3-D structures of CDH6-JMD (residues 660-677) and 302

CDH6-CBD (residues 687-784) in complexes with p120-catenin and -catenin, respectively, in 303

which orange spheres indicate the mutations (Figure 5B). Significantly, most of the mutations can 304

change the charge (D661N, E662K, R689Q, D691N, E722K, D762N, and R773Q); therefore, the 305

salt bridge interactions between cadherins and catenins can be interrupted, which may 306

consequently reduce cadherin/catenin complex formation. 307

The cadherin-catenin binding free energy differences between WT and MT structures were 308

calculated by using CGI protocol [30]. The catenin binding affinity differences (∆∆𝐺 ) 309

were calculated according to ΔG1 (catenin-bound) – ΔG2 (catenin-free), as shown in the 310

thermodynamic cycle when placed in one simulation box (Figure S1A). The net change of the 311

simulation box was considered as zero and negative ΔΔG values indicate stabilizing mutations 312

16

(Figure S1B). The resulting binding free energy differences (∆∆𝐺 ) in mutants as compared 313

to WT CDH6 were 4.60 (D661N) and 7.50 (E662K) kJ/mol for p120-catenin (Figure 5C and 314

Figure S8), whereas 8.17 (R689Q), 5.96 (D691N), -0.07 (E722K), 5.98 (D762N), and 8.70 (R773Q) 315

kJ/mol for -catenin (Figure 5C and Figure S9), suggesting that the binding affinity for both p120-316

catenin and -catenin were disrupted in most mutants (∆∆𝐺 0). The residue S749 lies 317

within the invariant cadherin sequence 749Ser-Leu-Ser-Ser-Leu753, which has been shown as an 318

important phosphorylation site to form hydrogen bonds with -catenin [36]. The S749F 319

substitution therefore can block the phosphorylation in this region, and consequently reduce 320

cadherin/-catenin complex formation. Similar strategies were also applied to calculate cadherin-321

catenin binding free energy differences caused by somatic mutations in other top-four mutated 322

genes. Except CDH10, most substitutions in JMD and CBD domains of those type-II CDHs 323

resulted in less stable cadherin-catenin complexes with increased free energy differences (Table 324

S2). 325

326

Mutations in the CDH-C of CDH6 and their functional consequences 327

To study the functional relevance, GFP-tagged CDH6 mutants with different known mutations in 328

CDH-C were prepared individually and co-transfected with wild type -catenin tagged with 329

mCherry into A375 cells, a melanoma cell line proven to express the lowest CDH6 level in cell 330

line screening. As shown in Figure 6A-B, wild type CDH6 was predominantly expressed on the 331

cell surface in complex with -catenin. Except D691N, CDH6 mutants with mutations in -332

catenin-binding domain significantly reduced complex formation on cell membrane by promoting 333

-catenin translocation into cytosol and/or nucleus, indicating destabilization of CDH6/-catenin 334

17

complex. For D661N and E662K, the portions of nuclear -catenin were also increased, which 335

could be due to indirect effects of unstable CDH6/p120-catenin complex formation (Figure 6A-336

B). 337

To know clinical and biofunctional impacts, we used dual-color immunofluorescence 338

staining to confirm protein expression patterns of core CDHs and -catenin in clinical samples. 339

For melanoma cells with wild type CDHs, the-catenin staining usually co-localized with CDH 340

staining on the cell surface. However, nuclear and cytosol -catenin staining as found in the cell-341

based study can be frequently detected in melanoma samples with mutated core CDHs (Figure 6C 342

and Figures S4-S7). We also analyzed mRNA levels of genes involved in Hippo and Wnt/-catenin 343

signaling in patients with mutations in top-four cadherins (MT core CDHs). Patients without any 344

mutation in classical cadherin genes (WT CDHs) were used as the controls. The Hippo-345

transcription factor TEAD4 was found significantly up-regulated (p = 0.0034) in MT core CDHs 346

group which associated with activation of target genes, including BIRC5 (p = 0.0004) and WNT1 347

(p = 0.0291) (Figure 6D). The Wnt/-catenin signaling responsive genes WNT1 (p = 0.0291) and 348

FOSL1 (p = 0.0012) were also found up-regulated. Coordination among TEAD4, FOSL1, BIRC5 349

and WNT1 was previously implicated in YAP/Tead-mediated cancer migration/invasion and ion 350

metabolism [37, 38]. C-MYC, WISP1 and PPARD were slightly increased but did not reach 351

statistical significance. 352

353

Clinical significance of cadherin mutations during melanoma development 354

To address the translational relevance, we performed clinical association study by subgrouping 355

patients into MT core CDHs and WT CDHs groups. Our data revealed that patients in the MT 356

group are at higher risks to get more aggressive tumors at advanced stages (stage III/IV) (p = 0.010; 357

18

Figure 7A, left) and lymph node invasion (p = 0.021; Figure 7A, middle). Although we found 358

higher tendency of metastasis for the MT group (6.2 % vs. 5.1 %; Figure 7A, right), the data did 359

not reach statistical significance. Surprisingly, such tumor-promoting activity did not determine 360

worse survival outcomes. In general, patients with mutations in core CDH genes showed better 361

overall (p = 0.008; Figure 7B, upper) and disease-free (p = 0.002; Figure 7B, lower) survival times. 362

To know possible signaling signatures in melanomas with mutations in core CDH genes, gene set 363

enrichment analysis (GSEA) was performed using WT group as the control. Several up-regulated 364

pathways were found to contribute to cancer progression, DNA replication/cell proliferation, and 365

cell cycle regulation (Table S3 and Figure S10), indicating the pro-oncogenic activity of cadherin 366

mutations. These data may explain why melanomas with cadherin mutations correlate with 367

advanced tumor stages and metastatic potentials. 368

Interestingly, T-lymphocyte-up and antigen response pathways were also found to be 369

activated in MT group (Figure 7C), suggesting that anti-tumor immune responses may be triggered 370

once tumor cells enter the circulation system. Consistent with those findings, patients in MT group 371

showed higher lymphocyte scores (including lymphocyte density and distribution) in their tumor 372

lesions as compared to patients in the control group (Figure 7D). To confirm such finding in real 373

clinical samples, melanoma tissue sections with matched genetic information (Table S1) were 374

utilized to detect T-lymphocytes (CD3+; red) in melanomas. Our data revealed higher T-375

lymphocyte (CD3+) infiltrations in melanoma lesions with mutations in core CDH genes as 376

compared to tissues with wild type CDHs (Figure 7E). Our data suggest that cadherin mutations 377

can potentially trigger anti-cancer immunity in patients via promoting lymphocyte infiltration, 378

resulting in prolonged survival times. 379

380

19

Cadherin mutations generate neo-antigens that favor better immunotherapy outcomes 381

To understand how cadherin mutations activate T-cell immunity, we used machine-learning-based 382

neo-antigen prediction program, NetMHCpan [24], to score neo-antigen potentials of defined 383

mutations. Notably, the recurrent S524L substitution on CDH6 was found to be a potent neo-384

antigen due to increased MHC-I binding affinity and less antigen-tolerance (Figure 8A). 385

Predictions on other three core cadherins confirmed an overall decrease of antigen-tolerance scores 386

associated with known somatic mutations (Figure S11). This finding indicated higher 387

immunogenic potentials generated by mutations in core CDH genes. Accumulating evidence have 388

revealed that tumors with higher mutational burdens correlate with the likelihood of higher neo-389

antigen loads, leading to better therapeutic outcome for patients after immunotherapy [39, 40]. By 390

using immunogenetic datasets from MSKCC and UCLA groups [39, 41], we found that mutations 391

in core CDH genes highly correlated with total mutational and neo-antigen loads in patients with 392

metastatic melanoma (Figure 8B-C). Those findings suggest an anti-tumor microenvironment 393

immune type associated with mutations in core CDH genes that favors a better clinical response 394

to immunotherapeutic agents (Figure S12) [42]. To confirm this, we also compared clinical 395

outcomes of anti-CTLA4 (Ipilimumab) [39] and anti-PD1 (Pembrolizumab) [41] therapies 396

between patients with and without mutations in core CDH genes. Consistent with higher 397

mutational and neo-antigen loads, patients with CDH mutations usually showed longer therapeutic 398

duration times for Ipilimumab treatment (Figure 8D) and better drug responses for Pembrolizumab 399

therapy (Figure 8E). 400

20

Discussion 401

Alterations in cell-cell contacts have long been considered as an important determinant of cancer 402

initiation and progression. Although molecular bases of many well-known cadherins e.g. CDH1 403

(E-cadherin) and CDH2 (N-cadherin) in cancer development have been established, we know very 404

little about genetic/functional impacts of other AJ genes on cancer especially cutaneous melanoma, 405

a unique cancer type with a high mutation load. Through data mining using TCGA dataset, we 406

discovered high mutation frequencies in cadherin genes which can be validated by using data from 407

non-TCGA cohort (Figure 2 and Figure S3A). Interestingly, we found the top-four highly mutated 408

genes (core CDHs, including CDH6, CDH18, CDH10 and CDH9) are all type-II cadherins which 409

harbor the capability to form heterophilic interactions. By contrast, mutation frequencies in type-I 410

cadherins, which form homophilic interactions, were relatively low. Even though mutation rate of 411

CDH4 (R-cadherin) was also high (the fifth highly mutated gene) and defined as a type-I cadherin, 412

the heterophilic interaction between CDH4 and CDH2 has been previously reported [43]. Dislike 413

CDH1 being genetically deleted or epigenetically silenced during cancer development, those top-414

four cadherins were consistently expressed to form cell-cell contacts during the transition from 415

melanocyte to melanoma. Our study therefore provides genetic evidence to support the view that 416

the heterophilic interactions between different cell types, such as melanocytes with keratinocytes 417

or stroma cells, play critical roles in skin tissue homeostasis. Breaking such balance by mutations 418

in key cadherins may serve as a newly-defined driving force to promote melanomagenesis, which 419

could be a possible common mechanism shared by other cancer types. Pan-cancer study of AJ 420

genetic alterations is under investigation. 421

Among AJ genes in cutaneous melanoma, CDH6 stood out to be a very interesting one 422

which showed the highest mutation rate with several novel functional mutations. Through cadherin 423

21

binding dynamics assays [35], CDH6 was also found at the center of the interactome formed by 424

the top-four core CDHs. Previous study has shown CDH6 a node protein for the interplay of 425

multiple cellular signaling such as TGF-, PI3K/AKT and FAK/ERK1/2, thus participates in the 426

regulation of stem cell formation and function [44, 45]. Both tumor-suppressing [46, 47] and pro-427

oncogenic activities [45, 48, 49] were associated with CDH6 expression in cancer development. 428

Emerging evidence in recent years suggest a pro-EMT role of CDH6 to promote cancer invasion 429

and metastasis [48, 49]. Those controversial results signify complex regulation in inter-/intra-tissue 430

crosstalk during cancer development and progression. Based on our study, recurrent mutation 431

S524L and functional mutations in the intracellular domain of CDH6 can result in loss of cell-cell 432

contacts and -catenin translocation into nucleus, which associate with tumors at more advanced 433

stages and lymph node invasion (Figures 6 and 7), suggesting those mutations as active drivers but 434

not passengers during melanogenesis. To our knowledge, this is the first report to address the 435

functional impacts of CDH6 and its mutations on melanoma development. Interestingly, we also 436

found nuclear staining for mutated CDH6 in tumor samples (Figure 6C), which could be novel 437

phenomena in certain types of melanoma awaiting for further investigation. 438

Type II cadherins have been shown to exhibit either homophilic or heterophilic binding 439

capability in cell aggregation studies [50-52] and their interactions define the development of 440

nervous system in vivo [51, 53, 54]. Expression of specific complements of type II cadherins in 441

neuron cells can determine their cell sorting during embryonic development into distinct sub-442

populations with different phenotypes/functions [55]. It has been known that melanocytes as well 443

as their stem cells in the skin originate from neural crest cells migrating out along the dorsal neural 444

tube. In this process, changing cell-cell contacts by cadherin switching plays important roles to 445

trigger cell de-polarization and epithelial–mesenchymal transition, leading to cell detachment and 446

22

migration [56, 57]. The well-known type-II cadherins involved in the underlying mechanisms 447

include CDH6 [15], CDH7 [58], and CDH11 [59]. Recent studies on biophysics of collective cell 448

migration revealed that cancer cell metastasis is a process very similar to neural crest cell spreading 449

[60-62]. This viewpoint highlights the requirement of changes in heterotypic mechanical 450

interactions for cells to pass through the mechanical barriers generated by primary cell-cell 451

contacts in the original tissue. For examples, fibroblast-like synoviocytes can migrate and invade 452

into pigmented villonodular synovitis by expressing CDH11 [63], Silencing of CDH10 was also 453

found as another mechanism to enhance breast tumor cell mobility in hypoxic conditions [64]. 454

These studies support our findings that mutations in core type-II cadherins may provide advantages 455

for melanoma cells to alter the original connecting networks with keratinocytes, enabling them to 456

further migrate out of melanocyte stem cell niche and form tumor lesions on the skin. 457

Based on genetic analyses, molecular modeling and experimental investigations, our data 458

highly suggest the causal consequence of mutations in CDH-C domains with activation of Wnt/-459

catenin signaling. Firstly, mutations in CDH-C domains of the core CDHs showed mutual 460

exclusivity with β-catenin (CTNNB1) mutations (Figure 3B), suggesting that they share redundant 461

biological functions; secondly, CDH-C mutations correlated with increased thermodynamics of 462

the CDH-C/-catenin or CDH-C/p120-catenin complexes (Figure 5C and Figures S8-S9), 463

indicating higher levels of free -catenin in cells; finally, live imaging study confirmed an overall 464

increased nuclear -catenin in cells expressing CDH6 mutants with mutations in the CDH-C 465

domain (Figure 6A-B). Consistent with this, the recurrent S524L mutation in CDH6-EC5 also 466

affect the stability and organization of CDH6-mediated AJ complex (Figure 4A-C), leading to -467

catenin release from the membrane and translocation into cell nucleus as well as loss of contact 468

23

inhibition (Figure 6B-D). Especially, there is a 510-His-Ala-Val-512 (HAV) motif at the EC5 469

domain which is considered responsible for stabilization and clustering of adjacent cadherin 470

monomers in a calcium-dependent fashion [65, 66]. Also, the EC5 is the only extracellular domain 471

that contains critical cysteine residues (at the positions of 497, 589, 591, and 600) in CDH6 for 472

possible cis- and trans-dimerization of neighboring cadherins. Therefore, S524L mutation may 473

restrain calcium-induced CDH6 rigidification and dimerization, leading to reduced cell contacts 474

with neighboring cells. 475

Although mutations in the core CDH genes showed pro-oncogenic activity, patients with 476

such mutations showed better overall and disease-free survival times (Figure 7B). These results 477

indicate possible complications associated with mutations in those type-II cadherins in controlling 478

immunosurveillance. Several cadherins including CDH6 were identified as RGD motif-containing 479

proteins, which can function as ligands to activate metastasis-associated integrins, including 1-480

containing integrins such as α2β1, and promote cancer cell proliferation, migration and invasion 481

[67-69]. Signaling activated by those metastasis-associated integrins can also enhance tumor 482

angiogenesis [68-70]. By contrast, activation of 1-containing integrins in T cells, a specialized 483

skin-resident/infiltrating T cells, can also promote cell adhesion to endothelial cells, extravasation 484

and lymphocyte infiltration into tumors [71-73]. T cells trigger immune response rapidly just 485

like innate-like lymphocytes do, yet they can direct the development of adaptive immunity with 486

TCR-dependent reactivity/specificity [74, 75]. By using the immunogenetic dataset from MSKCC 487

group [39], we also noticed patients with mutations in CDH6 or other core CDH genes showed 488

more γδT cell recruitments in melanoma lesions with more activated CD4+ memory T cells and 489

higher cytolytic scores (Figure S12). Accordingly, there could be a possibility that mutations in 490

24

cadherins, when the melanoma cells spread through the blood or lymph, can stimulate T cell 491

activation and enhance lymphocyte recruitment into core tumor lesions, leading to up-regulation 492

of T-lymphocyte and antigen responses which were evidenced in our study (Figure 7C-E). 493

A recent study indicated increased hydrophobicity of UV-derived mutations that can 494

enhance antigen presentation on the MHC complexes with stronger neo-antigen potentials [76]. 495

Since most of detected CDH mutations in melanoma are highly relevant to UV exposure, CDH 496

mutations should play certain key roles in triggering anti-tumor immunity in patients. Especially, 497

the S524L substitution in CDH6 has been predicted to be a strong neo-antigen (Figure 8A), 498

peptides with such substitution could be utilized as potent adjuvants to trigger stronger anti-499

melanoma immunity when co-treated with other immuno-therapeutics such as anti-PD-1/L1 or 500

anti-CTLA4 Abs. Whether mutations in cadherins can change cytokine profiles in the local 501

microenvironment to direct lymphocyte homing and activate anti-melanoma immunity will be 502

investigated in our future study. 503

Some limitations in this study need to be addressed here. Firstly, the mutational profiles 504

detected in the Taiwanese cohort were found a little bit different from the TCGA cohort which is 505

Caucasian dominant (96%) [19]. Those differences suggest the involvement of genetic factors 506

between different populations during melanomagenesis. However, CDH6, CDH10, CDH18 are 507

still within the top-five most frequently mutated cadherin genes in the Taiwanese cohort (Table 508

S1). Also, most of detected mutations are related to UV-exposure (G to A or C to T substitutions). 509

Therefore, protein staining data by using Taiwanese samples might be relevant but mutation sites 510

are different. Tumor samples from Caucasian population with matched genetic information may 511

provide the opportunity to confirm biological effects of those mutations shown in Figure 6A-B. 512

25

Secondly, CDH mutations alone may not be the only contributor to make melanoma becoming 513

more immunogenic. Functional impacts from other genetic events during melanoma development 514

should be also considered and investigated. 515

26

Abbreviations 516

AJs: adherens junctions; CDH: cadherin; core CDHs: the top-four mutated cadherins including 517

CDH6, CDH9, CDH10 and CDH18; Neo-Ag: neo-antigen; GSEA: gene set enrichment analysis; 518

EC5: extracellular cadherin domain 5; JMD: juxtramembrane domain; CBD: catenin-binding 519

domain; MD: molecular dynamics; WT: wild type; MT: mutant type; ΔΔG: free energy 520

differences. 521

522

Acknolwedgments 523

The authors thank technical assistance from Mr. Louice Chun-Chi Lai at National Sun Yat-sen 524

University/Taiwan. The authors also thank critical comments from Prof. Daniel Riveline at 525

IGBMC/France, Prof. Kuang-Hung Cheng at National Sun Yatsen University/Taiwan, and Prof. 526

Anna C Jan at National Cheng Kung University/Taiwan. This study was supported by grants from 527

Ministry of Science and Technology (MOST)/Taiwan (106-2314-B-110-001-MY3; 109-2314-B-528

110-003-MY3; 108-2811-B-110-506; 108-2221-E-110-061-MY2), Ministry of Health and 529

Welfare/Taiwan (MHW 10819), China Medical University Hospital (DMR-108-121), and the 530

NSYSU-KMU joint research projects (108-P013). 531

532

Author contributions 533

CCC, SYC and JJCS designed the experiments and validated the data. PKK, DEG, MTL, HYC 534

and TH performed bioinformatics and clinical association study. CCC and PKK performed 535

molecular modeling and free-energy calculation. DEG, CMC, and TH made the gene constructs, 536

27

conducted cell-based studies, and analyzed the data. CCC, PKK, DEG, MTL, SYC and JJC 537

confirmed the data and wrote the manuscript. 538

539

Competing Interests 540

The authors declare no competing interests in this study. 541

542

Data availability 543

Data from the experiments presented in this study are included in this published article and its 544

Supplementary Information file. Other data generated or analyzed during this study are also 545

available from the corresponding authors on request. 546

28

Supplementary Materials 547

Table S1. Genetic mutation rates of core CDH genes based on targeted exon sequencing in 548

melanoma samples from Taiwanese patients 549

Table S2. Calculation of catenin-binding free energy differences between WT and known 550

mutations in core CDH genes in melanoma 551

Table S3. Gene set enrichment analyses of reactomes/pathways concurrently up-regulated with 552

mutations in top-four core CDH genes in melanoma 553

Figure S1. Double-system/single-box setup 554

Figure S2. Mutations in nectin and nectin-like genes in skin cutaneous melanoma 555

Figure S3. Validation of mutation frequencies in key adherens junction (AJ) genes in non-TCGA 556

cohort 557

Figure S4. Dual-color immunofluorescence staining of CDH6 and -catenin on melanoma tissues 558

from Taiwanese patients 559

Figure S5. Dual-color immunofluorescence staining of CDH9 and -catenin on melanoma tissues 560

from Taiwanese patients 561

Figure S6. Dual-color immunofluorescence staining of CDH10 and -catenin on melanoma 562

tissues from Taiwanese patients 563

Figure S7. Dual-color immunofluorescence staining of CDH18 and -catenin on melanoma 564

tissues from Taiwanese patients 565

Figure S8. The predicted CDH6/p120-catenin (-catenin) complex structures and their binding 566

free energy differences (ΔΔG) between WT and MT CDH6 567

29

Figure S9. The predicted CDH6/-catenin complex structures and their binding free energy 568

differences (ΔΔG) between WT and MT CDH6 569

Figure S10. Reactomes/pathways associated with mutations in top-four highly mutated cadherin 570

(core CDHs) genes in skin cutaneous melanoma 571

Figure S11. Neo-antigen potentials of mutations in core CDH genes in skin cutaneous melanoma. 572

Figure S12. Changes of tumor microenvironments in skin cutaneous melanoma by mutations in 573

core CDH genes 574

30

References 575

1. Little EG, Eide MJ. Update on the current state of melanoma incidence. Dermatol Clin. 576

2012; 30: 355-61. 577

2. Wang JX, Fukunaga-Kalabis M, Herlyn M. Crosstalk in skin: melanocytes, keratinocytes, 578

stem cells, and melanoma. J Cell Commun Signal. 2016; 10: 191-6. 579

3. Dvořánková B, Szabo P, Kodet O, Strnad H, Kolář M, Lacina L, et al. Intercellular 580

crosstalk in human malignant melanoma. Protoplasma. 2017; 254: 1143-50. 581

4. Moon H, Donahue LR, Choi E, Scumpia PO, Lowry WE, Grenier JK, et al. Melanocyte 582

stem cell activation and translocation initiate cutaneous melanoma in response to UV 583

exposure. Cell Stem Cell. 2017; 21: 665-78 e6. 584

5. Nelson WJ, Nusse R. Convergence of Wnt, beta-catenin, and cadherin pathways. Science. 585

2004; 303: 1483-7. 586

6. Heuberger J, Birchmeier W. Interplay of cadherin-mediated cell adhesion and canonical 587

Wnt signaling. Cold Spring Harb Perspect Biol. 2010; 2: a002915. 588

7. Buckley CD, Tan J, Anderson KL, Hanein D, Volkmann N, Weis WI, et al. Cell adhesion. 589

The minimal cadherin-catenin complex binds to actin filaments under force. Science. 2014; 590

346: 1254211. 591

8. Malinova TS, Huveneers S. Sensing of cytoskeletal forces by asymmetric adherens 592

junctions. Trends Cell Biol. 2018; 28: 328-41. 593

9. Harris TJ, Tepass U. Adherens junctions: from molecules to morphogenesis. Nat Rev Mol 594

Cell Biol. 2010; 11: 502-14. 595

10. Yap AS, Gomez GA, Parton RG. Adherens junctions revisualized: Organizing cadherins 596

as nanoassemblies. Dev Cell. 2015; 35: 12-20. 597

11. Kim JE, Finlay GJ, Baguley BC. The role of the hippo pathway in melanocytes and 598

melanoma. Front Oncol. 2013; 3: 123. 599

12. Nallet-Staub F, Marsaud V, Li L, Gilbert C, Dodier S, Bataille V, et al. Pro-invasive 600

activity of the Hippo pathway effectors YAP and TAZ in cutaneous melanoma. J Invest 601

Dermatol. 2014; 134: 123-32. 602

13. Sanchez IM, Aplin AE. Hippo: hungry, hungry for melanoma invasion. J Invest Dermatol. 603

2014; 134: 14-6. 604

14. Fisher ML, Grun D, Adhikary G, Xu W, Eckert RL. Inhibition of YAP function overcomes 605

BRAF inhibitor resistance in melanoma cancer stem cells. Oncotarget. 2017; 8: 110257-606

72. 607

15. Clay MR, Halloran MC. Cadherin 6 promotes neural crest cell detachment via F-actin 608

regulation and influences active Rho distribution during epithelial-to-mesenchymal 609

transition. Development. 2014; 141: 2506-15. 610

16. Inoue T, Chisaka O, Matsunami H, Takeichi M. Cadherin-6 expression transiently 611

delineates specific rhombomeres, other neural tube subdivisions, and neural crest 612

subpopulations in mouse embryos. Dev Biol. 1997; 183: 183-94. 613

17. Hsu MY, Wheelock MJ, Johnson KR, Herlyn M. Shifts in cadherin profiles between human 614

normal melanocytes and melanomas. J Investig Dermatol Symp Proc. 1996; 1: 188-94. 615

18. Kuphal S, Bosserhoff AK. E-cadherin cell-cell communication in melanogenesis and 616

during development of malignant melanoma. Arch Biochem Biophys. 2012; 524: 43-7. 617

31

19. Cancer Genome Atlas Network. Genomic classification of cutaneous melanoma. Cell. 618

2015; 161: 1681-96. 619

20. Hayward NK, Wilmott JS, Waddell N, Johansson PA, Field MA, Nones K, et al. Whole-620

genome landscapes of major melanoma subtypes. Nature. 2017; 545: 175-80. 621

21. Ascierto PA, Simeone E, Sileni VC, Pigozzo J, Maio M, Altomonte M, et al. Clinical 622

experience with ipilimumab 3 mg/kg: real-world efficacy and safety data from an expanded 623

access programme cohort. J Transl Med. 2014; 12: 116. 624

22. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis 625

of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. 2013; 626

6: pl1. 627

23. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene 628

set enrichment analysis: a knowledge-based approach for interpreting genome-wide 629

expression profiles. Proc Natl Acad Sci U S A. 2005; 102: 15545-50. 630

24. Jurtz V, Paul S, Andreatta M, Marcatili P, Peters B, Nielsen M. NetMHCpan-4.0: Improved 631

peptide-MHC class I interaction predictions integrating eluted ligand and peptide binding 632

affinity data. J Immunol. 2017; 199: 3360-8. 633

25. Kechin A, Boyarskikh U, Kel A, Filipenko M. cutPrimers: A new tool for accurate cutting 634

of primers from reads of targeted next generation sequencing. J Comput Biol. 2017; 24: 635

1138-43. 636

26. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, et al. Share the sequence 637

alignment/Map format and SAMtools. Bioinformatics. 2009; 25: 2078-9. 638

27. Webb B, Sali A. Comparative Protein Structure Modeling Using MODELLER. Curr 639

Protoc Protein Sci. 2016; 86: 291-37. 640

28. Van Der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, Berendsen HJ. GROMACS: 641

fast, flexible, and free. J Comput Chem. 2005; 26: 1701-18. 642

29. Gapsys V, Michielssens S, Seeliger D, de Groot BL. pmx: Automated protein structure and 643

topology generation for alchemical perturbations. J Comput Chem. 2015; 36: 348-54. 644

30. Goette M, Grubmuller H. Accuracy and convergence of free energy differences calculated 645

from nonequilibrium switching processes. J Comput Chem. 2009; 30: 447-56. 646

31. Gul IS, Hulpiau P, Saeys Y, van Roy F. Evolution and diversity of cadherins and catenins. 647

Exp Cell Res. 2017; 358: 3-9. 648

32. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer 649

genomics portal: an open platform for exploring multidimensional cancer genomics data. 650

Cancer Discov. 2012; 2: 401-4. 651

33. Strathdee G. Epigenetic versus genetic alterations in the inactivation of E-cadherin. Semin 652

Cancer Biol. 2002; 12: 373-9. 653

34. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med. 2004; 654

10: 789-99. 655

35. Brasch J, Katsamba PS, Harrison OJ, Ahlsén G, Troyanovsky RB, Indra I, et al. 656

Homophilic and heterophilic interactions of yype II cadherins identify specificity groups 657

underlying cell-adhesive behavior. Cell Rep. 2018; 23: 1840-52. 658

36. Huber AH, Weis WI, The structure of the beta-catenin/E-cadherin complex and the 659

molecular basis of diverse ligand recognition by beta-catenin. Cell. 2001; 105: 391-402. 660

37. Liu X, Li H, Rajurkar M, Li Q, Cotton JL, Ou J, et al. Tead and AP1 Coordinate 661

Transcription and Motility. Cell Rep. 2016; 14: 1169-80. 662

32

38. Torti SV, Torti FM. Iron and cancer: more ore to be mined. Nat Rev Cancer. 2013; 13: 663

342-55. 664

39. Snyder A, Makarov V, Merghoub T, Yuan J, Zaretsky JM, Desrichard A, et al. Genetic 665

basis for clinical response to CTLA-4 blockade in melanoma. N Engl J Med. 2014; 371: 666

2189-99. 667

40. Samstein RM, Lee CH, Shoushtari AN, Hellmann MD, Shen R, Janjigian YY, et al. Tumor 668

mutational load predicts survival after immunotherapy across multiple cancer types. Nat 669

Genet. 2019; 51: 202-6. 670

41. Hugo W, Zaretsky JM, Sun L, Song C, Moreno BH, Hu-Lieskovan S, et al. Genomic and 671

transcriptomic features of response to anti-PD-1 therapy in metastatic melanoma. Cell. 672

2016; 165: 35-44. 673

42. Chen YP, Zhang Y, Lv JW, Li YQ, Wang YQ, He QM, et al. Genomic analysis of tumor 674

microenvironment immune types across 14 solid cancer types: immunotherapeutic 675

implications. Theranostics. 2017; 7: 3585-94. 676

43. Shan WS, Tanaka H, Phillips GR, Arndt K, Yoshida M, Colman DR. Functional cis-677

heterodimers of N- and R-cadherins. J Cell Biol. 2000; 148: 579-90. 678

44. Park SR, Kim HJ, Yang SR, Park CH, Lee HY, Hong IS. A novel endogenous damage 679

signal, glycyl tRNA synthetase, activates multiple beneficial functions of mesenchymal 680

stem cells. Cell Death Differ. 2018; 25: 2023-36. 681

45. Sancisi V, Gandolfi G, Ragazzi M, Nicoli D, Tamagnini I, Piana S, et al. Cadherin 6 is a 682

new RUNX2 target in TGF-beta signalling pathway. PLoS One. 2013; 8: e75489. 683

46. Goeppert B, Ernst C, Baer C, Roessler S, Renner M, Mehrabi A, et al. Cadherin-6 is a 684

putative tumor suppressor and target of epigenetically dysregulated miR-429 in 685

cholangiocarcinoma. Epigenetics. 2016; 11: 780-90. 686

47. Sellar GC, Li L, Watt KP, Nelkin BD, Rabiasz GJ, Stronach EA, et al. BARX2 induces 687

cadherin 6 expression and is a functional suppressor of ovarian cancer progression. Cancer 688

Res. 2001; 61: 6977-81. 689

48. Gugnoni M, Sancisi V, Gandolfi G, Manzotti G, Ragazzi M, Giordano D, et al. Cadherin-690

6 promotes EMT and cancer metastasis by restraining autophagy. Oncogene. 2017; 36: 691

667-77. 692

49. Zhao L, Jiang R, Xu M, Zhu P, Mo XM, Wang N, et al. Concomitant high expression of 693

BRAFV600E, P-cadherin and cadherin 6 is associated with High TNM stage and lymph 694

node metastasis in conventional papillary thyroid carcinoma. Clin Endocrinol (Oxf). 2016; 695

84: 748-55. 696

50. Nakagawa S, Takeichi M. Neural crest cell-cell adhesion controlled by sequential and 697

subpopulation-specific expression of novel cadherins. Development. 1995; 121: 1321-32. 698

51. Patel SD, Ciatto C, Chen CP, Bahna F, Rajebhosale M, Arkus N, et al. Type II cadherin 699

ectodomain structures: implications for classical cadherin specificity. Cell. 2006; 124: 700

1255-68. 701

52. Shimoyama Y, Tsujimoto G, Kitajima M, Natori M. Identification of three human type-II 702

classic cadherins and frequent heterophilic interactions between different subclasses of 703

type-II classic cadherins. Biochem J. 2000; 349: 159-67. 704

53. Basu R, Duan X, Taylor MR, Martin EA, Muralidhar S, Wang Y, et al. Heterophilic Type 705

II Cadherins Are Required for High-Magnitude Synaptic Potentiation in the Hippocampus. 706

Neuron. 2017; 96: 160-76 e8. 707

33

54. Osterhout JA, Josten N, Yamada J, Pan F, Wu SW, Nguyen PL, et al. Cadherin-6 mediates 708

axon-target matching in a non-image-forming visual circuit. Neuron. 2011; 71: 632-9. 709

55. Dewitz C, Duan X, Zampieri N. Organization of motor pools depends on the combined 710

function of N-cadherin and type II cadherins. Development. 2019; 146: dev180422. 711

56. Scarpa E, Szabó A, Bibonne A, Theveneau E, Parsons M, Mayor R. Cadherin switch during 712

EMT in neural crest cells leads to contact inhibition of locomotion via repolarization of 713

forces. Dev Cell. 2015; 34: 421-34. 714

57. Taneyhill LA, Schiffmacher AT. Cadherin dynamics during neural crest cell ontogeny. 715

Prog Mol Biol Transl Sci. 2013; 116: 291-315. 716

58. Nakagawa S, Takeichi M. Neural crest emigration from the neural tube depends on 717

regulated cadherin expression. Development. 1998; 125: 2963-71. 718

59. Blaue C, Kashef J, Franz CM. Cadherin-11 promotes neural crest cell spreading by 719

reducing intracellular tension-Mapping adhesion and mechanics in neural crest explants by 720

atomic force microscopy. Semin Cell Dev Biol. 2018; 73: 95-106. 721

60. Labernadie A, Trepat X. Sticking, steering, squeezing and shearing: cell movements driven 722

by heterotypic mechanical forces. Curr Opin Cell Biol. 2018; 54: 57-65. 723

61. Barriga EH, Franze K, Charras G, Mayor R. Tissue stiffening coordinates morphogenesis 724

by triggering collective cell migration in vivo. Nature. 2018; 554: 523-7. 725

62. Kerosuo L, Bronner-Fraser M. What is bad in cancer is good in the embryo: importance of 726

EMT in neural crest development. Semin Cell Dev Biol. 2012; 23: 320-32. 727

63. Cao C, Wu F, Niu X, Hu X, Cheng J, Zhang Y, et al. Cadherin-11 cooperates with 728

inflammatory factors to promote the migration and invasion of fibroblast-like synoviocytes 729

in pigmented villonodular synovitis. Theranostics. 2020; 10: 10573-88. 730

64. Casciello F, Al-Ejeh F, Miranda M, Kelly G, Baxter E, Windloch K, et al. G9a-mediated 731

repression of CDH10 in hypoxia enhances breast tumour cell motility and associates with 732

poor survival outcome. Theranostics. 2020; 10: 4515-29. 733

65. Boggon TJ, Murray J, Chappuis-Flament S, Wong E, Gumbiner BM, Shapiro L. C-734

cadherin ectodomain structure and implications for cell adhesion mechanisms. Science. 735

2002; 296: 1308-13. 736

66. Blaschuk OW, Sullivan R, David S, Pouliot Y. Identification of a cadherin cell adhesion 737

recognition sequence. Dev Biol. 1990; 139: 227-9. 738

67. Bartolomé RA, Peláez-García A, Gomez I, Torres S, Fernandez-Aceñero MJ, Escudero-739

Paniagua B, et al. An RGD motif present in cadherin 17 induces integrin activation and 740

tumor growth. J Biol Chem. 2014; 289: 34801-14. 741

68. Desgrosellier JS, Cheresh DA. Integrins in cancer: biological implications and therapeutic 742

opportunities. Nat Rev Cancer. 2010; 10: 9-22. 743

69. Seguin L, Desgrosellier JS, Weis SM, Cheresh DA. Integrins and cancer: regulators of 744

cancer stemness, metastasis, and drug resistance. Trends Cell Biol. 2015; 25: 234-40. 745

70. Naci D, Vuori K, Aoudjit F. Alpha2beta1 integrin in cancer development and 746

chemoresistance. Semin Cancer Biol. 2015; 35: 145-53. 747

71. Siegers GM. Integral Roles for Integrins in gammadelta T Cell Function. Front Immunol. 748

2018; 9: 521. 749

72. Cruz MS, Diamond A, Russell A, Jameson JM. Human alphabeta and gammadelta T cells 750

in skin immunity and disease. Front Immunol. 2018; 9: 1304. 751

73. Quaresma JAS. Organization of the skin immune system and compartmentalized immune 752

responses in infectious diseases. Clin Microbiol Rev. 2019; 32: e00034-18. 753

34

74. Davey MS, Willcox CR, Joyce SP, Ladell K, Kasatskaya SA, McLaren JE, et al. Clonal 754

selection in the human Vdelta1 T cell repertoire indicates gammadelta TCR-dependent 755

adaptive immune surveillance. Nat Commun. 2017; 8: 14760. 756

75. Fay NS, Larson EC, Jameson JM. Chronic Inflammation and gammadelta T Cells. Front 757

Immunol. 2016; 7: 210. 758

76. Pham TV, Boichard A, Goodman A, Riviere P, Yeerna H, Tamayo P, et al. Role of 759

ultraviolet mutational signature versus tumor mutation burden in predicting response to 760

immunotherapy. Mol Oncol. 2020; 14: 1680-94. 761

35

762

763

Figure 1. The workflow scheme presents the whole process of data mining and functional 764

characterizations in this study. 765

• Collect NGS data of melanoma samples (including gDNA and mRNA) from TCGA (n=448) and non-TCGA (n=396) cohortsthrough cBioportal (www.cbioportal.org)

• Mutation mapping and identification of recurrent or conserved mutations using Oncoprint and Mutation mapper

• Mutation validation by using data from non-TCGA cohorts• Key pathways of mRNA expression defined by GSEA (Gene Set

Enrichment Analysis) method (www.genepattern.org)

• Gene cloning of CDH6 (fused with GFP) and -catenin (mCherry)• Site-directed mutagenesis to introduce known mutations in CDH6• Gene transfection and protein co-localization analysis

• 3-D structural predictions by MODELLER program• Molecular dynamics simulations (RMSD and RMSF) by GROMACS

software• Binding free energy calculation according to CGI protocol by

GROMACS software • Neo-Ag predictions of somatic mutations by NetMHCpan tool

(www.cbs.dtu.dk/services/NetMHCpan)

• Clinical association study, including tumor stage, lymph node invasion, metastasis, lymphocyte scores, overall/disease-free survivals, and immunotherapy responses

• Mutation validation by targeted gene sequencing on melanomas• Fluorescent staining on tissue sections/microarrays

1. NGS Data Collection

2. NGS Data Analysis

3. In Silico Functional Prediction

4. Cell-Based Functional Validation

5. Clinical Relevance Analysis

36

766

Figure 2. Mutations in classical cadherin genes in skin cutaneous melanoma. (A) STRING 767

protein-protein interaction analysis (https://string-db.org) on 20 annotated classical cadherins 768

reveals an interactome composed of 16 classical cadherins (6 type-I and 10 type-II) involved in AJ 769

organization, homophilic cell adhesion via plasma membrane adhesion molecules, and cell 770

junction assembly pathways. (B) Mutation frequencies of the cadherin genes in skin cutaneous 771

melanoma were analyzed by using NGS data from the TCGA cohort (TCGA, PanCancer Atlas; n 772

= 448). Our data showed no mutation in CDH3. (C) The heat map indicates nucleotide substitution 773

profiles of mutated cadherin genes in skin cutaneous melanoma. 774

37

775

776

Figure 3. Biological relevance of mutations in classical cadherin genes in skin cutaneous 777 melanoma. (A) Lollipop plots of mutations in classical cadherin genes in skin cutaneous 778

melanoma were presented by using NGS data from TCGA cohort (n = 448) (left) and the non-779

TCGA cohort (n = 396) (right). The non-TCGA cohort includes patients from Broad/Dana Farber 780

(n = 26, Nature 2012), MSKCC (n = 64, NEJM 2014), Broad (n = 121, Cell 2012), Yale (n = 147, 781

Nat Genet 2012), and UCLA (n = 38, Cell 2016). (B) Mutations in the intracellular domains of 782

top-five cadherin genes were compared with mutations in CTNNB1. (C) Co-occurrence analysis 783

among top-four type-II classical cadherin genes (core CDHs) including CDH6, CDH9, CDH10, 784

and CDH18. (D) Gene expression levels of core CDHs during melanoma malignant transition from 785

38

normal neural crest cells and melanocytes to melanoma. Gene expression data were extracted from 786

the GEO dataset (GDS1965). (E) Physical interactions (heterophilic/homophilic) among the top-787

four core CDHs were previously characterized and confirmed [35]. 788

39

789

790

Figure 4. Impact of S524L mutation on the fifth extracellular (EC5) domain of the CDH6 791 protein. (A) Protein stability of wild type (WT) and mutant (S524L) CDH6 were compared by 792

residue-specific Root-Mean-Square Deviation (RMSD) and Root-Mean-Square Fluctuations 793

(RMSF) methods in simulations. The S524L mutation lead to higher instability in the 2/3 loop 794

and 4/5 loop regions. (B) The EC5 structures of the WT and S524L CDH6 are drawn in cartoon 795

putty representation, where the color is ramped by residue from blue as the lowest B-factor value 796

to red as the highest B-factor value. In addition, the size of the tube also reflects the value of the 797

B-factor, where the larger the B-factor the thicker the tube. The calcium atoms and mutation sites 798

are indicated with gray sphere and yellow sphere representations, respectively. (C) The 799

thermodynamic cycle for Ca2+ binding affinity was calculated in one simulation box. The free 800

energy difference in Ca2+ binding corresponds to a double free energy difference: ΔΔG = ΔG1 – 801

ΔG2. 802

40

803

804 Figure 5. Thermo-stability of CDH6/catenin complex with mutations in the intracellular 805 domain (CDH-C) of CDH6. (A) Alignment of CDH-C protein sequences among classical 806

cadherins. Letters with underlines and yellow shadows indicate known mutations in the CDH-C 807

sequences in skin cutaneous melanoma based on the NGS data from TCGA cohort. Letters with 808

orange color represent the conserved residues whereas letters in red color represent identical 809

residues among classical cadherins. Green box and red box indicate the regions of JMD for p120-810

41

catenin (-catenin) and CBD for -catenin binding, respectively. (B) The complex structures of 811

CDH6-JMD (blue)/p120-catenin (green) (left) and CDH6-CBD (blue)/-catenin (green) (right) 812 were predicted by using the MODELLER program. The mutation sites are highlighted in orange 813

spheres. (C) The binding free energy differences (ΔΔG) caused by mutations in the domains of 814

JMD for p120-catenin binding (upper) and CBD for -catenin binding (lower) were calculated by 815

using the fast-growth thermodynamic integration approach. 816

42

817 818

Figure 6. Functional impacts of mutations in the intracellular domain (CDH-C) of CDH6 on 819

-catenin binding. (A) Wild type (WT) CDH6 and its mutants (tagged with GFP) were co-820

transfected with -catenin (tagged with mCherry) into A375 cells. Cells expressing both green and 821

red constructs were monitored and representative images were taken under a fluorescent 822

microscope 24 hrs after transfection. (B) Distributions of -catenin in cells with different CDH6 823

constructs were counted and categorized into nucleus, cytoplasm and membrane (n = 50 cells for 824

each CDH6 construct). (C) Dual-color immunofluorescence staining, CDH6 in red and -catenin 825

in green, was performed on melanoma tissue slides from patients with wild type cadherins and 826

patients with D724N mutation in CDH6. (D) Gene expression levels of genes involved in Hippo 827

(blue box) and Wnt/-catenin (pink box) pathways were compared between patients with 828

mutations in top-four core cadherins (MT core CDHs) and patients without mutation in any 829

classical cadherins (WT CDHs). The expression levels of indicated genes were extracted from the 830

mRNA data of TCGA cohort. Statistical differences between two groups were compared by using 831

43

t-test. The p values were presented as *: p value < 0.05, **: p value < 0.01, and ***: p value < 832

0.001. 833

44

834

835 836 Figure 7. Clinical significance of cadherin mutations during melanoma development. (A) 837

Clinical associations were compared between patients with mutations in top-four core cadherins 838

(MT core CDHs) and patients without any mutation in classical cadherin genes (WT CDHs), 839

including tumor stage (left), lymph node invasion (middle), and metastasis (right). (B) Kaplan-840

Meier analyses were performed to compare overall survival (upper) and disease-free survival 841

(lower) between MT core CDHs and WT CDHs groups. (C) Gene set enrichment analysis (GESA) 842

was performed using mRNA expression data from TCGA group. Two pathways, T-lymphocyte-843

45

up (upper) and antigen response (lower), were found positively associated with mutations in top-844

four core CDH genes. (D) Lymphocyte scores in melanoma tissues of TCGA cohort were 845

compared between MT core CDHs and WT CDHs groups. (E) Dual-color immunofluorescence 846

staining, CD3 in red and-catenin in green, was performed on melanoma tissue slides from 847 patients with wild type cadherins and patients with K467R mutation in CDH6 (left). CD3+ T-848

lymphocytes in the stained tissue sections were quantified and compared between MT core CDHs 849

and WT CDHs groups (right). For data in (A), (D) and (E), Fisher’s exact test was performed to 850

analyze the clinical association. For (B), the long-rank test was used to compare the survival times. 851

The p values were presented as *: p value < 0.05, **: p value < 0.01, and ***: p value < 0.001. 852

46

853 854

Figure 8. Immunogenicity of cadherin mutations and therapeutic advantages for treating 855 melanoma. (A) The mutation counts (gray bars) for known CDH6 mutations in TCGA cohort 856

were shown alone with the amino acid positions in CDH6. Antigen tolerance scores of the 857

indicated mutation (red square) and its corresponding wild type (green circle) were analyzed by 858

using NetMHCpan algorithm and plotted on this bar chart. Immunogenetic datasets of MSKCC 859

[39] and UCLA [41] were utilized to compare (B) overall mutational loads; (C) Neo-antigen loads; 860

(D) therapeutic duration times for Ipilimumab (anti-CTLA-4) treatment; and (E) drug responses 861

for Pembrolizumab (anti-PD-1) therapy between MT core CDHs and WT CDHs groups. For data 862

in (B) and (C), Fisher’s exact test was performed to analyze the statistical significances. For (D) 863

and (E), two-proportions Z-Test was utilized to compare drug effectiveness. The p values were 864

presented as *: p value < 0.05, **: p value < 0.01, and ***: p value < 0.001. 865

0% 20% 40% 60% 80% 100%

1

2

圖表標題

數列1 數列2

0% 20% 40% 60% 80% 100%

1

2

圖表標題

數列1 數列2

mutation count wild type score mutant type score

0 100 200 300 400 500 600 7000

2

4

6

8

0

2

4

6

8

0 100 200 300 400 500 600 7000.001

0.01

0.1

1

10

100

1000

Ant

igen

tole

ranc

e sc

ore

Mutation counts

Amino acid position on CDH6

Po

ten

tial

neo

-ant

ige

nN

orm

al

self-

antig

enS524L

MSKCC, NEJM 2014

0

1000

2000

3000

4000

5000

MSKCC, NEJM 2014

0

1000

2000

3000

4000

MSKCC, NEJM 2014

p = 0.0001***

UCLA, Cell 2016

p = 0.0049**

Mut

atio

nal l

oad

WT MT Core WT MT Core

MSKCC, NEJM 2014

0

2000

4000

6000

MSKCC, NEJM 2014

0

1000

2000

3000

MSKCC, NEJM 2014

p = 0.0034**

UCLA, Cell 2016

p = 0.0001***

Neo

Ant

igen

load

WT MT Core WT MT Core

Pembrolizumab (anti-PD-1) responseIpilimumab (anti-CTLA-4) effects

16 6 11 5

WT

MT core

WT

MT core

Responder

Non-responder

Effective more than 2 years

Effective less than 2 yearsMSKCC, NEJM 2014; p = 0.0401* UCLA, Cell 2016; p = 0.0485*

0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%

21 21 9 12

A

B C

D E

Table S1. Genetic mutation rates of core CDH genes based on targeted exon sequencing in melanoma samples from Taiwanese patients

Gene Extracellular domain

and transmembrane

mutation rate (%)

Substitution eventsa Intracellular domain

mutation rate (%)

Substitution eventsa Total mutation rate

(%) (cases/ total

samples)

CDH6 9.09% exon9:c.A1400G:p.K467R 18.18% exon12:c.G2170A:p.D724N

exon12:c.G2236A:p.V746M

22.73%

(5/22)

CDH10 18.18% exon3:c.C346G:p.R116G

exon3:c.C377T:p.T126I

exon5:c.G607A:p.E203K

exon7:c.C1163T:p.S388F

9.09% exon12:c.G2044A:p.E682K 22.73%

(5/22)

CDH18 13.63% exon4:c.G310A:p.D104N

exon8:c.G1124A:p.G375E

exon9:c.G455A:p.G152E

13.63% exon16:c.G2345A:p.G782E 22.73%

(5/22)

CDH12 18.18% exon6:c.C28T:p.L10F

exon6:c.C65T:p.P22L

exon12:c.T1301C:p.I434T

exon15:c.G1802A:p.C601Y

exon16:c.C2186T:p.T729I

0% None detected 18.18%

(4/22)

CDH9 9.09% exon2:c.C62T:p.T21I

exon4:c.G541A:p.V181I

0%

None Detected 13.63%

(3/22)

CDH24 9.09% exon3:c.C395T:p.P132L

exon3:c.C470T:p.A157V

4.54% exon13:c.G2036A:p.R679Q 13.63%

(3/22)


CDH1 9.09% exon8:c.C1082T:p.A361V


4.54 exon14:c.G2287A:p.E763K 9.09%

(2/22)

CDH2 4.54% exon9:c.G1276A:p.D426N

exon11:c.C1725A:p.F575L

9.09% exon15:c.G2476A:p.A826T

exon16:c.T2593G:p.S865A

9.09%

(2/22)

CDH4 9.09% exon1:c.C44T:p.P15L

exon14:c.A2338C:p.I780L


(2/22)

CDH5 9.09% exon4:c.C556T:p.H186Y

exon11:c.G1798A:p.A600T


(2/22)

CDH7 9.09% exon3:c.G433A:p.D145N



4.54% exon12:c.C2264T:p.S755F 9.09%

(2/22)

CDH8 9.09% exon3:c.G340A:p.V114I

exon3:c.C445G:p.P149A

exon10:c.A1589C:p.Y530S


(2/22)

CDH11 9.09% exon3:c.C65T:p.A22V



(2/22)

CDH13 4.54% exon5:c.C557T:p.S186F


9.09% exon14:c.C2143T:p.P715S


9.09%

(2/22)

CDH15 4.54% exon9:c.G1306A:p.V436M 0% Not detected 4.54%

(1/22) ac- chromosomal base substitution; p- amino acid substitution

Table S2. Calculation of catenin-binding free energy differences between WT and known

mutations in core CDH genes in melanoma

Mutation Protein Domain Binding protein

∆∆𝑮𝑩𝒊𝒏𝒅𝒊𝒏𝒈𝑴𝒖𝒕𝒂𝒕𝒊𝒐𝒏

(kJ/mol)

D661N CDH6, CDH9, CDH10 JMD p120-catenin 4.60 ± 0.69

E662K CDH6, CDH9 JMD p120-catenin 7.50 ± 2.11

G663E CDH18 JMD p120-catenin 42.29 ± 7.57

G665R CDH10 JMD p120-catenin -15.34 ± 13.85

E666K CDH9 JMD p120-catenin 21.95 ± 4.39

E667K CDH18 JMD p120-catenin 1.56 ± 2.88

R689Q CDH6 CBD -catenin 8.17 ± 4.91

D691N CDH6 CBD -catenin 5.96 ± 0.96

E722K CDH6 CBD -catenin -0.07 ± 5.04



E741K CDH9 CBD -catenin 16.38 ± 2.35

G742E CDH10 CBD -catenin -19.63 ± 6.25

L767F CDH10 CBD -catenin -1.60 ± 5.56

R773Q CDH6, CDH9 CBD -catenin 8.70 ± 5.05

Table S3. Gene set enrichment analyses of reactomes/pathways concurrently up-regulated

with mutations in top-four core CDH genes in melanomaa

Reactome /Pathwayb ES NES NOM

p-value

1. Liver_Cancer_Subclass_G3_Up 0.6493 1.7587 0.0159

2. Doxorubicin_Resistance_Up 0.7836 1.6533 0.0175

3. Antigen_Response 0.6726 1.7603 0.0236

4. Thyroid_Carcinoma_Anaplastic_Up 0.5938 1.7737 0.0276

5. T_Lymphocyte_Up 0.7554 1.6582 0.0320

6. Epithelial_Mesenchymal_Transition_Up 0.6731 1.7110 0.0333

7. M_G1_Transition 0.6640 1.6491 0.0345

8. Core_Serum_Response_Up 0.6037 1.6801 0.0352

9. Cell_Cycle_Mitotic 0.6362 1.6307 0.0369

10. Synthesis_Of_DNA 0.6648 1.6498 0.0381

11. Mitotic_G1/G1_S_Phases 0.5985 1.6321 0.0386

12. CDC20_Mediated_Degradation_Of_Mitotic_Proteins 0.6666 1.6343 0.0423

13. DNA_Replication 0.6606 1.6724 0.0437

14. Mitotic_M/M_G1_Phases 0.6589 1.6693 0.0437

15. Cervical_Cancer_Proliferation_Cluster 0.6992 1.6660 0.0483 aAbbreviations: ES, enrichment score; NES, normalized enrichment score; NOM p-value,

nominalized p-value. bReactomes or pathways with NOM p-values < 0.05 were considered as significantly relevant

with mutations in top-four core CDH genes.

Figure S1. Double-system/single-box setup. (A) Two branches of a thermodynamic cycle are

placed in one simulation box. The different boxes in the scheme are indicated by the solid (𝜆

0) and broken (𝜆 1) lines. (B) An example of a simulation setup for the catenin binding free

energy calculation. The free energy difference corresponds to a double free energy difference:

ΔΔG = ΔG1 – ΔG2.

Figure S2. Mutations in nectin and nectin-like genes in skin cutaneous melanoma. (A)

Mutation frequencies and (B) mutational profiles of nectin and nectin-like genes in skin

cutaneous melanoma were analyzed by using NGS data from the TCGA cohort (TCGA,

PanCancer Atlas; n = 448).

Figure S3. Validation of mutation frequencies in key adherens junction (AJ) genes in non-

TCGA cohort. Mutation frequencies of key AJ genes, including (A) top-five highly-mutated

cadherin genes identified in TCGA cohort, (B) nectin and (C) nectin-like genes, were analyzed

by using NGS data from non-TCGA cohort (n = 396). The non-TCGA cohort include patients

from Broad/Dana Farber (n = 26, Nature 2012), MSKCC (n = 64, NEJM 2014), Broad (n = 121,

Cell 2012), Yale (n = 147, Nat Genet 2012), and UCLA (n = 38, Cell 2016).

Figure S4. Dual-color immunofluorescence staining of CDH6 and -catenin on melanoma

tissues from Taiwanese patients. Tissue sections were prepared from Taiwanese patients with

melanoma. Dual-color immunofluorescence staining, red for CDH6 and green for -catenin,

was performed on melanocyte from normal skin epithelium counterpart (upper), melanoma

with wild type CDHs (middle), and melanoma with D724N mutation (lower). Cell nuclei were

counterstained with DAPI.

Figure S5. Dual-color immunofluorescence staining of CDH9 and -catenin on melanoma

tissues from Taiwanese patients. Tissue sections were prepared from Taiwanese patients with

melanoma. Dual-color immunofluorescence staining, red for CDH9 and green for -catenin,

was performed on melanocyte from normal skin epithelium counterpart (upper), melanoma

with wild type CDHs (middle). Cell nuclei were counterstained with DAPI.

Figure S6. Dual-color immunofluorescence staining of CDH10 and -catenin on

melanoma tissues from Taiwanese patients. Tissue sections were prepared from Taiwanese

patients with melanoma. Dual-color immunofluorescence staining, red for CDH10 and green

for -catenin, was performed on melanocyte from normal skin epithelium counterpart (upper),

melanoma with wild type CDHs (middle), and melanoma with E682K mutation (lower). Cell

nuclei were counterstained with DAPI.

Figure S7. Dual-color immunofluorescence staining of CDH18 and -catenin on

melanoma tissues from Taiwanese patients. Tissue sections were prepared from Taiwanese

patients with melanoma. Dual-color immunofluorescence staining, red for CDH18 and green

for -catenin, was performed on melanocyte from normal skin epithelium counterpart (upper),

melanoma with wild type CDHs (middle), and melanoma with G782E mutation (lower). Cell

nuclei were counterstained with DAPI.

Figure S8. The predicted CDH6/p120-catenin (-catenin) complex structures and their

binding free energy differences (ΔΔG) between WT and MT CDH6. (A) The predicted

structure of JMD (residues 660-677, in blue) in complex with p120-catenin (-catenin, in green)

was generated by using the MODELLER program. The known mutated sites, D661 and D662,

on the JMD (in orange) were found to interact with K444 and K574 on the p120-catenin,

respectively. (B) A double-system/single-box setup was used to estimate the binding free energy

differences (ΔΔG) caused by mutations.

Figure S9. The predicted CDH6/-catenin complex structures and their binding free

energy differences (ΔΔG) between WT and MT CDH6. (A) The predicted structure of CBD

(residues 687-784, in blue) in complex with -catenin (green) was generated by using

MODELLER program. The known mutated sites were highlighted in orange spheres. A double-

system/single-box setup was used to estimate the binding free energy differences (ΔΔG) caused

by (B) R689Q, (C) D691N, (D) E722K, (F) E726N, and (G) R773Q substitutions. Cartoon

representations indicated the binding interfaces between mutated residues on CBD (yellow) and

the interacting residues on -catenin.

Figure S10. Reactomes/pathways associated with mutations in top-four highly mutated

cadherin (core CDHs) genes in skin cutaneous melanoma. Gene set enrichment analysis

(GSEA) was performed using mRNA expression data from the TCGA group (TCGA,

PanCancer Atlas; n = 448). Patients without mutations in any cadherin genes were utilized as

the controls. Concurrent up-regulated reactomes with p-values less than 0.05 were shown here.

Two pathways, Antigen Response and T-Lymphocyte Up, were also shown in Fig. 7C.

Figure S11. Neo-antigen potentials of mutations in core CDH genes in skin cutaneous

melanoma. Antigen-tolerance scores of known mutations in the extracellular domain of (A)

CDH9; (B) CDH10; and (C) CDH18 were analyzed by using NetMHCpan algorithm and

compared with the scores with wild type sequence. Substitutions with lower scores indicate

higher neo-antigen potentials. Two-way ANOVA was utilized to compare the differences

between wild type and mutant type. Statistical significance: *, p < 0.05; **, p < 0.01.

CDH18

1 2 3 4 5 6 7 8 910111213141516171819202122232425262728293031323334353637383940414243444546470

40

80

120WTMT

CDH10

1 2 3 4 5 6 7 8 910111213141516171819202122232425262728293031323334353637383940414243444546470

40

80

120WTMT

CDH9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 410

40

80

120WT

MT

G54

DG

86E

H11

1YE

120K

R12

7CR

133K

E14

0KS

143L

D15

1NE

156K

I201

KR

267Q

P26

9SG

310E

G32

7EE

375K

L446

FS

450L

P45

2LN

464D

V47

4FE

498K

D51

5NL5

33P

D57

4ND

592N

E64

3KY

659H

D66

1NE

662K

E66

6KE

708K

D73

3NS

734L

E74

1KG

742W

D74

4ND

748E

S74

9LR

773C

H50

RR

51C

N59

KG

69R

D81

NH

112Y

E12

1KI1

32N

D15

2NP

170S

E21

4KR

230K

G24

3SL2

79F

Y30

6NR

307Q

P35

7SG

363E

S38

7FF

391C

H39

4YS

421F

R42

4HA

462V

N46

6KL4

80V

P48

7LG

520S

D53

9YG

553R

G58

0SS

597F

P60

6SR

638Q

E64

8KD

649N

D65

9NG

663R

E68

3GI6

90N

P69

8SR

699W

D71

2YG

740E

L765

F

R50

CK

52E

H78

PD

82E

D10

5ND

109N

K11

4NE

144K

G17

5SG

192R

G20

4ER

226K

E23

0KM

303I

G31

4ER

325K

K34

0QD

376N

E37

9KE

394K

V40

2AE

426K

P45

2SD

467H

E49

8KG

520R

P53

5LE

543K

N54

5DR

555K

T55

9AD

562N

R59

8QR

652W

E65

3KG

662E

E66

6KP

679L

E69

4KD

711Y

E71

4KP

730L

D75

4ND

763N

G76

8E

p = 0.0030**

p = 0.0070**

p = 0.0148*

CDH9

CDH10

CDH18

Ant

igen

-tol

eran

ce s

core

Ant

igen

-tol

eran

ce s

core

Ant

igen

-tol

eran

ce s

core 120

80

40

0

120

80

40

0

120

80

40

0

wild type scoremutant type score



A

B

C

amino acid substitutions in CDH9



Figure S12. Changes of tumor microenvironments in skin cutaneous melanoma by

mutations in core CDH genes. Microenvironment immune types were analyzed in tumors with

mutations in (A) CDH6 or (B) core CDH genes, and compared with the data in tumors with

wild type CDHs. The immune scores include T cells, M1 macrophages, activated NK cells,

activated CD4+ memory T cells, CD8+ T cells, monocytes, follicular helper T cells and cytolytic

score.

-100

0

100

200

300

400

-100

0

100

200

300

-500

0

500

1000

1500 p = 0.0188*

-200

-100

0

100

200

300

400 p = 0.0498*

-200

-100

0

100

200

300 p = 0.1031

0

100

200

300p = 0.0011**

-50

0

50

100

150

200

250p = 0.0013**

-100

0

100

200

300

400p = 0.0096**

-100

0

100

200

300 p = 0.0953

-100

0

100

200

300

400 p = 0.0076**

T cells gam

ma delta

Macrophages M

1

T cells CD4 m

emory activated

NK cells activated

T cells CD8

Monocytes

Cytolytic Score

T cells follicular helper

WT CDHs MT core WT CDHs MT core WT CDHs MT core WT CDHs MT core


p = 0.0749 p = 0.1003 p = 0.0490* p = 0.0258*

p = 0.0145* p = 0.0414* p = 0.1942 p = 0.0285*

T cells gam

ma delta

Macrophages M

1

T cells CD4 m

emory activated

NK cells activated

T cells CD8

Monocytes

Cytolytic Score

T cells follicular helper



A CDH6

B core CDHs

-50

0

50

100

150

200

250

-100

0

100

200

300

400

-500

0

500

1000

1500

-200

-100

0

100

200

300

400

-200

-100

0

100

200

300

400

0

100

200

300

somatic mutational landscapes of adherens junctions and their … · 1 1 somatic mutational...

Documents