somatic mutational landscapes of adherens junctions and their … · 1 1 somatic mutational...
TRANSCRIPT
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)
exon8:c.G1279A:p.E427K
CDH1 9.09% exon8:c.C1082T:p.A361V
exon10:c.G1483A:p.V495M
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
0% None detected 9.09%
(2/22)
CDH5 9.09% exon4:c.C556T:p.H186Y
exon11:c.G1798A:p.A600T
0% None detected 9.09%
(2/22)
CDH7 9.09% exon3:c.G433A:p.D145N
exon11:c.G1789A:p.E597K
exon11:c.G1864A:p.V622M
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
0% None detected 9.09%
(2/22)
CDH11 9.09% exon3:c.C65T:p.A22V
exon5:c.G607A:p.E203K
0% None detected 9.09%
(2/22)
CDH13 4.54% exon5:c.C557T:p.S186F
exon6:c.G760A:p.V254I
9.09% exon14:c.C2143T:p.P715S
exon14:c.G2170A:p.V724I
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
D726N CDH6 CBD -catenin 5.98 ± 1.77
D733N CDH9 CBD -catenin 15.62 ± 1.83
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
wild type scoremutant type score
wild type scoremutant type score
A
B
C
amino acid substitutions in CDH9
amino acid substitutions in CDH10
amino acid substitutions in CDH18
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
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
WT CDHs MT core WT CDHs MT core WT CDHs MT core WT CDHs MT core
WT CDHs MT core WT CDHs MT core WT CDHs MT core WT CDHs MT core
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