A o ame elay in o ma ion sys em in he epide mal g ow h ac o
ecep o –d ug complexes e eals clues o new pa adigm in p o ein
con o ma ional change
Ta eq Hameduh
a
, Michal Mok y
a,b
, And ew D. Mille
a,c,d
, Voj ech Adam
a,b
, Zbynek Hege
a,b
,
Yazan Haddad
a,b,
⇑
a
Depa men o Chemis y and Biochemis y, Mendel Uni e si y in B no, Zemedelska 1, CZ-613 00 B no, Czech Republic
b
Cen al Eu opean Ins i u e o Technology, B no Uni e si y o Technology, Pu kyno a 656/123, 612 00 B no, Czech Republic
c
Ve e ina y Resea ch Ins i u e, Hudco a 70, CZ-62100 B no, Czech Republic
d
KP The apeu ics (Eu ope) s. .o., Pu kyn
ˇo a 649/127, B no CZ-61200, Czech Republic
a icle in o
A icle his o y:
Recei ed 2 Ma ch 2021
Recei ed in e ised o m 13 Sep embe
2021
Accep ed 24 Sep embe 2021
A ailable online 27 Sep embe 2021
Keywo ds:
EGFR
NSCLC
Tumou esis ance
Ty osine kinase inhibi o
Ro ame
P o ein s uc u e
P o ein olding
abs ac
Cance cells can escape he e ec s o chemo he apy h ough mu a ions and up egula ion o a y osine
kinase p o ein called he epide mal g ow h ac o ecep o (EGFR). In he pas wo decades, ou gene -
a ions o y osine kinase inhibi o s a ge ing EGFR ha e been de eloped. Using compa a i e s uc u e
analysis o 116 EGFR-d ug complex c ys al s uc u es, clus e analysis p oduces wo clans o 73 and 43
s uc u es, espec i ely. The i s clan o 73 s uc u es is la ge and is comp ised mos ly o he C-helix-
IN con o ma ion while he second clan o 43 s uc u es co ela es wi h he C-helix-OUT con o ma ion.
A deep o ame analysis iden i ies 43 esidues (18%) o he o al o 237 esidues spanning he kinase
s uc u es unde in es iga ion wi h signi ican o ame a ia ions be ween he C-helix-IN and C-helix-
OUT clans. The loca ions o hese o ame a ia ions ake on he appea ance o side chain con o ma ional
elays ex ending ou om poin s o EGFR mu a ion o di e en egions o he EGFR kinase. Acco dingly,
we p opose ha key EGFR mu a ions ac singly o oge he o induce d ug esis an con o ma ional
changes in EGFR ha a e communica ed ia hese side chain con o ma ional elays. Acco dingly, hese
side chain con o ma ional elays appea o play a signi ican ole in he de elopmen o umou esis-
ance. This phenomenon also sugges s a new pa adigm in p o ein con o ma ional change ha is media ed
by suppo i e elays o o ame s on he p o ein su ace, a he han h ough con en ional backbone
mo emen s.
Ó2021 The Au ho s. Published by Else ie B.V. on behal o Resea ch Ne wo k o Compu a ional and
S uc u al Bio echnology. This is an open access a icle unde he CC BY license (h p://c ea i ecommons.
o g/licenses/by/4.0/).
1. In oduc ion
Epide mal g ow h ac o ecep o (EGFR) p o ein was
disco e ed in he la e 1970s, and o his day i emains a p ima y
a ge o an icance he apy [1,2]. Mu a ions and up egula ion o
EGFR a e known mechanisms o bo h oncogenesis and he apeu ic
esis ance. Ac i e EGFR mu an s escape he e ec s o chemo he -
apy ia con inuous p oli e a ion and e asion o apop osis, espe-
cially in he case o non-small cell ca cinoma o he lung (NSCLC)
[3]. Fu he mo e, EGFR is p e alen in o he solid umou s includ-
ing: b eas , colon, enal, o a ian, and head-and-neck cance s [4].
EGFR is a membe o he ecep o y osine kinases (RTKs) cell
su ace ecep o amily [5,6], and is also a membe o he E bB am-
ily, a ou membe amily esponsible mainly o egula ing cell
p oli e a ion [6]. The p ima y unc ion o EGFR is o media e sig-
nals o di e en ia ion, mo ili y, and apop osis [6]. The e o e, any
i egula EGFR beha iou can easily become p o-oncogenic [7].
P o-oncogenic mu a ions in EGFRs a e widesp ead and linked o
umou o e g ow h and esis ance o chemo he apy [8].
EGFR comp ises h ee p o ein domains, 1) an ex acellula
ligand-binding domain o 621 amino acid esidues, 2) a ansmem-
b ane domain o 23 amino acid esidues, and 3) a cy oplasmic
domain comp ising 542 amino acids [9-11]. The ex acellula
domain con ains he ligand-binding pocke ha media es he
open-close s a us o he ecep o o signal ansduc ion. The ans-
memb ane domain is a single
a
helix. Finally, he cy oplasmic
domain con ains a jux amemb ane cy oplasmic subdomain, a
h ps://doi.o g/10.1016/j.csbj.2021.09.026
2001-0370/Ó2021 The Au ho s. Published by Else ie B.V. on behal o Resea ch Ne wo k o Compu a ional and S uc u al Bio echnology.
This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/).
⇑
Co esponding au ho .
E-mail add ess: [email p o ec ed] (Y. Haddad).
Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
jou nal homepage: www.else ie .com/loca e/csbj
y osine kinase subdomain and a C- e minal subdomain [12]. EGFR
ligands include epide mal g ow h ac o (EGF), amphi egulin
(AREG), ans o ming g ow h ac o (TGF) and epigen [13]. O he
ligands a e hepa in-binding EGF (HB-EGF), epi egulin (EPR) and
be acellulin (BTC) [14]. Ligand binding esul s in EGFR ac i a ion
by ei he homodime iza ion o he e odime iza ion (EGFR-
E bB
2/3/4
) ha leads o he in acellula modula ion o di e en sig-
nalling pa hways (MAPK, PI3K/Ak /mTOR o JAK/STAT) [15,16].In
cance , EGFR ac i a ion leads o he downs eam ini ia ion o di -
e en key cellula e en s in ol ed in cellula g ow h, p oli e a ion,
in asion, me as asis and angiogenesis [17]. O e p oduc ion o
EGFR ligands can also d i e cance p og ession, howe e , ha
can be in luenced by he cance mic oen i onmen [10,15,18],
hype sensi iza ion o y osine kinase subdomain [19], and EGFR
o e exp ession [20]. In he es o his a icle, he wo d ‘‘ligand”
will be exclusi ely used o desc ibe he compounds o d ugs bind-
ing o he in acellula kinase domain, unless men ioned
o he wise.
Chemo he apeu ic app oaches ha e ocused on se e al aspec s
o he EGFR s uc u e and mechanism o ac ion. Fo example, mon-
oclonal an ibodies, such as ce uximab, ha e been de ised o a ge
he ex acellula domain, as epo ed in ea men s o bowel o
head and neck cance [21]. Inhibi o s o he in acellula y osine
kinase subdomain ac o a enua e y osine kinase ac i a ed sig-
nalling pa hways, and he eby igge cance cell dea h [22], wi h
epo s o success in NSCLC pa ien s in e ms o ea men and
quali y o li e [23]. Un o una ely, i has become clea now ha
bo h monoclonal an ibody-based agen s and y osine kinase inhi-
bi o s (TKIs) a e s uggling o keep pace wi h he eme ging o
new EGFR mu a ions. While monoclonal an ibodies a e e ec i e
agains wild ype EGFR [24], hey a e much less e ec i e agains
EGFR mu an s (i.e., exon 19 dele ions and L858R mu a ion) de ec -
able in 10–15% o Caucasian NSCLC pa ien s and 50% o Asian
pa ien s [25,26]. The i s gene a ion TKIs such as e lo inib and
ge i inib we e o iginally hough o be mo e obus agains EGFR
mu a ions. Howe e , he eme gence o he T790M mu a ion, in
he EGFR ATP-binding si e o he y osine kinase subdomain, was
su icien o cu ail he e icacies o e e sible i s gene a ion TKIs
[27]. The T790M mu a ion is p esen in abou 50%–60% o pa ien s
ha de elop chemo esis ance o TKIs [28]. Fi s gene a ion TKIs
we e quickly eplaced by i e e sible second gene a ion EGFR TKIs
(pan-HER inhibi o s), such as a a inib, dacomi inib and ne a inib,
ha unc ionally inhibi wild ype and T790M EGFR mu an s [29].
The ea e , hi d gene a ion EGFR TKIs we e de eloped, such as
osime inib and ocile inib, o o e come his pa icula chemo e-
sis ance, wi h mo e e icacy and less side e ec s han obse ed
wi h i s and second gene a ion inhibi o s pa icula ly owing o
co alen binding o he C797 esidue [26]. Un o una ely, he
appea ance o a C797S mu a ion in he umou s o pa ien s ea ed
wi h hi d gene a ion TKIs apidly cu ailed he e icacy o hese
d ugs [30,31]. Acco dingly, s a e-o - he-a ou h gene a ion allos-
e ic EGFR inhibi o s, like EAI045 and EAI001, we e c ea ed o a -
ge a di e en binding si e in he EGFR kinase. These d ug en i ies
a oid he p oblems o bo h T790M and C797S mu a ions. Howe e ,
he e idence sugges s ha such ou h gene a ion inhibi o s a e
insu icien ly e ec i e alone, o example EAI045 is only p ope ly
e ec i e in combina ion wi h ce uximab [32,33].
Clea ly, esis ance o chemo he apy is no jus simply a unc ion
o EGFR mu a ions alone. EGFR-independen ac o s include he
o e exp ession o addi ional g ow h ac o ecep o s such as
HER2, MET and FGFR, educed NF1 exp ession, and o e ac i i y
o PI3K o B-Ra [28,34-36]. Ha ing said his, EGFR mu a ions a e
dominan , and hus EGFR emains a p ima y ups eam a ge o
cance chemo he apy.
Recen ly exploi ed concep s in cance like he game heo y and
he biological in o ma ional heo y could help o unde s and his
p oblem in mo e sophis ica ed way. The game heo y in cance ,
desc ibes how cance can ake ad an age o he dynamics o he
umou mic oen i onmen o i s own su i al by di e en ways,
o ins ance, umou he apy esis ance [37]. On he o he hand,
he biological in o ma ion heo y in cance , could help us o unde -
s and how cance cells can main ain hei signal ansduc ion
speci ici y and he amoun o in o ma ion ansmi ed wi h di e -
en mu a ions. Fu he , i explains how cance de elops esis ance
o hold he in eg i y o i s in o ma ional sys em, which will be o
in e es o cance cells su i al [38]. Acco dingly, he e is a majo
unme need o unde s and he molecula mechanisms o EGFR
mu a ion-induced esis ance o chemo he apy. In he ligh o
inc easing numbe s in deposi ed EGFR kinase 3D s uc u es, i
has become a challenge in s uc u e-based d ug design o make a
choice wi hou ully unde s anding he a ia ions in ol ed a he
molecula le el. In a p e ious s udy, we ha e iden i ied C-helix in
he N-lobe o EGFR kinase domain as he majo s uc u al a ia ion
occu ing in EGFR kinase complexes wi h ligands based on analysis
o backbone mo emen s [39]. The objec i e o his wo k is o
in es iga e he ela ionship be ween EGFR kinase domain o ame
a ia ions, mu a ions, C-helix mo emen , and kinase ac i a ion
(DFG domain mo emen s). By aking ad an age o ou ecen ly
de eloped code o o ame analysis [40], he e, we a emp ed o
shed he ligh on he biological ele ance o global o ame ic
changes in he EGFR kinase.
2. Ma e ial and me hods
2.1. Da ase p ocessing
Using he keywo d ‘‘EGFR”, RCSB p o ein da abank (www. csb.
o g) da abase sea ch esul ed in 260 s uc u es. All en ies ha
did no co e he kinase domain o EGFR we e excluded. En ies
wi hou inhibi o s we e also excluded excep o ou wild ype
en ies: 1M14, 2GS2, 3GOP, 4TKS. Only chain A was e ie ed
( he numbe o chains igno ed we e 28 om all s uc u es). In
o al, 116 EGFR kinase 3D s uc u es spanning 714–950 (Unip o
ID P00533-1) we e immed and u he s udied. The wo lobes
o he kinase domain we e iden i ied as N-lobe (spanning 714–
795) and C-lobe (spanning 796–950), and he ligands we e
sal aged.
The 83 ligands sal aged in hese s uc u es we e p e iously
classi ied by us acco ding o he p ima y and seconda y he e o-
cyclic g oup in o he ollowing classes: 1 An ibio ic, 2 Benzimida-
zoles, 6 Fu opy imidines, 2 Indoloca bazoles, 7 Pu ines, 1
Py azine, 7 Py azolopy imidines, 3 Py idones, 34 Py imidines, 1
Py imidopy idone, 8 Py olopy imidines, 7 Quinazolines, 1 Quino-
lines, and 2 Thiazoles.
2.2. Visualiza ion
Visualiza ion o P o ein and ligand 3D S uc u es was done in
UCSF Chime a ( e sion 1.10.2). The ma chmake plugin was used
o supe posi ion o all hea y a oms ia BLOSUM-62 sco ing ma ix
and Needleman-Wunsch alignmen algo i hm. S uc u e ende ing
and anima ion we e done in UCSF Chime a using he command
line. G aphics we e p ocessed using MS Powe poin and Adobe
Pho oshop.
2.3. S uc u e i ing and clus e analysis
S uc u e i ing and clus e analysis we e done in R language
(Ve sion 3.6.1, The R ounda ion o S a is ical Compu ing, Aus ia)
using msd() unc ion om Bio3D lib a y (G an lab, Uni e si y o
Cali o nia, San Diego, USA) and agnes() unc ion om Clus e
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5444
lib a y (Ma in Machle , ETH Zu ich, Swi ze land) using he Wa d
me hod o oo mean squa e de ia ions (RMSD) dissimila ma ix,
espec i ely. The Wa d me hod o clus e ing, which is also known
as he minimum a iance me hod, is gene al pu pose me hod o
clus e ing ha s a s wi h nclus e s (each con aining a single
s uc u e), hen hese clus e s a e combined in each s ep (minimiz-
ing he a iance) un il all s uc u es a e con ained wi hin a single
clus e . RS udio Ve sion 1.2.5001 (RS udio, Inc.) was used o cod-
ing and ob aining esul s.
2.4. C-helix and DFG domain analysis
C-helix mo emen and DFG domain clus e ing we e measu ed
by wo echniques. 1) he angles o he helix axis ( esidues 756–
767) be ween each s uc u e and a e e ence s uc u e (Re e ence
PDB IDs: 3gop and 1m14) we e measu ed using a command line
in UCSF Chime a a e supe posi ion o 3D s uc u es. 2) Ac i e
and inac i e kinase s uc u es based on clus e s o DFG domain
o sions we e es ima ed acco ding o he new nomencla u e o
Modi and Dunb ack [41]. B ie ly, he DFG clus e is comp ised o
he Ramachand an egions (A, alpha; B, be a; L, le ) o he
Ty 854, Asp855 and Phe856 in addi ion o he i s side chain angle
o Phe856 (minus, plus, ans). The mos common ac i e DFGin
clus e is BLAminus (be a Ty 854, le Asp855, alpha Phe856 and
minus Phe856 side chain), whe eas he mos common inac i e
DFGou clus e is BBAminus (be a Ty 854, be a Asp855, alpha
Phe856 and minus Phe856 side chain).
2.5. Compa a i e o ame analysis
Compa a i e o ame analysis was done in R language acco d-
ing o ou p e iously published me hod [40]. B ie ly, using Bio3D
lib a y, s uc u es we e loaded ia ead.pdb() unc ion and o -
sional angles we e calcula ed ia o sion.pdb() unc ion. Classi ica-
ion o o ame s was done acco ding o he Richa dson’s
Penul ima e o ame lib a y [42], using IF/ELSE s a emen s as p e-
iously desc ibed [40]. Ro ame nomencla u e is based on he side
chain o sional angles (
1 o sion be ween N, C
a
,Cband C
c
a oms,
2 angle be ween C
a
,Cb,C
c
and Cda oms, e c.). The angle modes
we e used o classi y he h ee main classes as plus/ ans/minus
(i.e.,p, ,m o +60°, 180°, and -60°, espec i ely). Fo a oms in side
chains o he han ca bon, he angles modes we e shi ed and we e
eplaced wi h explici angle mode (e.g., m-80 o ame o Asn
desc ibed
1 = -60°and
2 = -80°modes). In his manusc ip ,
he deg ee symbol o explici angles was con enien ly emo ed
om nomencla u e o coding pu poses. The R sc ip s, which a e
s ill no op imized as a package, a e a ailable o academic pu -
poses upon eques om he co esponding au ho .
Chi squa e es o independence was used o cons uc c oss abs
be ween g oups o s uc u es and esidue o ame s and o de e -
mine i he e was any associa ion be ween he a iables in IBM
SPSS S a is ics 21 p og am (IBM Co po a ion, A monk, New Yo k,
USA). A p- alue below 0.05 was conside ed signi ican .
3. Resul s and discussion
3.1. Ro ame ic di e ences be ween EGFR kinase s uc u es
In he pas wo decades, an imp essi e wide ange o TKIs a -
ge ing EGFR ha e been de ised [39]. A he same ime, a la ge
numbe o EGFR-ligand complexes ha e been deposi ed in he
da abase e e y yea , al hough hese s uc u es a e no necessa ily
ep esen a i e o he la es ends in EGFR inhibi o design due o
he ime equi ed o X- ay c ys allog aphy expe imen s. Ne e -
heless, he a ailabili y o hund eds o simila such s uc u es does
p o ide s a is ical con idence in he alidi y o compa a i e s uc-
u e analysis ega ding con o ma ional changes.
Acco dingly, we ob ained and immed 116 EGFR s uc u es o
i ing and clus e ing analysis using RMSD o all a oms.
Clus e analysis showed wo main clans o s uc u es (Fig. 1)
simila o ou p e ious analysis o he N-lobe o EGFR kinase
[39]. The i s clan o 73 s uc u es was la ge and comp ised
mos ly o he C-helix-IN con o ma ion and wo o less mu a-
ions. While he second clan o 43 s uc u es was co ela ed wi h
he C-helix-OUT con o ma ion and h ee o less mu a ions. In
ou p e ious wo k, he ocus was on RMSD alues calcula ed
o C
a
-backbone mo emen s, he e we chose o s udy bo h back-
bone and side chain mo emen s. Ini ially, in o de o quan i y
he C-helix axis mo emen , wo e e ence s uc u es we e used
o de ine C-helix-IN (PDB ID 1m14) and C-helix-OUT (PDB ID
3gop) con o ma ions. In e es ingly, he C-helix axis angles o
1m14, and o he C-helix-IN con o ma ions, we e in he ange > 0°
o 6°, while C-helix angles o 3gop, and o he C-helix-OUT con-
o ma ions, we e in he ange o 6° o 33°deg ees
(a e age 25°deg ees) (Supplemen a y Table 1), al hough in
di e en ou wa d di ec ions.
Fo mo e insigh , we hen pe o med a deep o ame analysis
(Supplemen a y Table 1). Nea ly 43 esidues (18%) o he o al
o 237 esidues spanning he s udied kinase s uc u es showed
signi ican o ame a ia ions be ween he C-helix-IN and C-
helix-OUT clans (Table 1 and Fig. 2). In pa icula , six esidues
spanning he C-helix (namely, Asn756, Ile759, Glu762, Ty 764,
Se 768, and Val769) exhibi ed signi ican o ame a ia ions
be ween C-helix-IN and C-helix-OUT con o ma ions. In addi ion,
wo esidues o he DFG domain (Asp855 and Phe856) also dis-
played signi ican o ame a ia ions be ween he C-helix-IN and
C-helix-OUT clans.
In gene al, h ee ypes o o ame a ia ions we e obse ed
be ween he C-helix-IN and C-helix-OUT clans: Fi s ly, a ia ions
comp ising unique o ame s. Secondly, whe e one o ame is
common in bo h clans bu he second anked o ame is di e -
en . Thi dly, whe e one o ame is common in 100% o cases
in one clan, while o he o ame s domina e in he o he clan.
Ro ame di e en ia ions we e mos clea wi h iple mu an s
(las en ows a he bo om o Figs. 1 and 2). Al hough, iple
mu an s we e seen o sha e simila o ame s in hei DFG
domains, pa icula ly a Lys852 and Phe856, and a o he mo e
dis an esidues (A g889 and Lys913). Mos impo an ly, he
loca ions o hese o ame a ia ions ake on he appea ance o
side chain con o ma ional elays ex ending ou om poin s o
mu a ion o di e en egions o he EGFR kinase (Figs. 3 and 4,
Supplemen a y Mo ie 1–3).
This poses he ob ious ques ion which is, can EGFR mu a ions
ac singly o oge he o induce d ug esis an con o ma ional
changes in EGFR ha a e communica ed ia hese side chain
con o ma ional elays? Fu he mo e, can o ame a ia ions on
he p o ein su ace also play ole in allos e ic mechanisms? In
ac , gi en ha mos o he side chains in he 43 esidues men-
ioned abo e a e ac ually acing he p o ein su ace (wi h access
o wa e sol en ), hen any po en ial ‘‘ low o in o ma ion” om
o ame o o ame in such a dynamic en i onmen mus also
in ol e o sional angle mo emen s in bo h backbone and side
chains in addi ion o su ace ne wo k H-bonds o sal b idges
wi h wa e molecules. Indeed, nea ly 20 esidues (47%) in he
elay a e loca ed in he smalle N-lobe and hus a e mo e
exposed o wa e sol en . Fu he mo e, o ame a ia ions a e
obse ed ac oss mos ypes o amino acids (excep o Cys and
T p, pe haps due o hei gene al in equency in p o eins), and
no only long chains amino acids (e.g. such as A g and Lys) ha
may display al e na i e o ame s on he su ace due o hei
lexibili y alone [43,44].
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5445
Fig. 1. Clus e analysis o RMSD be ween EGFR kinase s uc u es spanning 714–950 esidues (nume a ion acco ding o Unip o ID P00533-1). The C-helix-IN con o ma ion
clan o s uc u es is shown on op in anned colou whe eas he C-helix-OUT con o ma ion clan o s uc u es is shown below in ligh blue colou . C-helix-IN clan showed
mos ly he ac i e C-helix-IN (DFGin/BLAminus) con o ma ions. The iple mu an s in he C-helix-OUT clan showed mos ly he inac i e C-helix-OUT (DFGou /BBAminus)
con o ma ions. C-helix o ien a ion was es ima ed by he helix axis angle agains he helix axis o a e e ence C-helix-IN (PDB ID 1m14). (Fo in e p e a ion o he e e ences o
colo in his igu e legend, he eade is e e ed o he web e sion o his a icle.)
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5446
3.2. Resis ance, mu a ions and EGFR elay sys ems
Tumou s a e known o de elop wo ypes o d ug esis ance, i.e.,
he inna e and he acqui ed. Inna e esis ance is de ined as he ail-
u e o ini ial he apy due o a ious umou mechanisms. Acqui ed
esis ance is de ined as p og ession (e.g., due o mu a ions) o he
disease a e a pe iod o ‘‘clinical bene i ” [45]. Acqui ed EGFR-
TKIs esis ance mechanisms a y acco ding o TKI ypes, mu a ion
ypes, and o he ac o s. Be o e we highligh hose EGFR mu a ions
in ol ed in acqui ed esis ance mechanisms, i is impo an o
emphasize on he na u e o wild ype EGFR kinase in ee and
d ug-bound o ms. The ca aly ic ac i i y o EGFR is egula ed by
h ee mechanisms: phospho yla ion, au oinhibi ion, and allos e ic
binding [46]. Wild ype EGFR kinase is in insically au oinhibi ed in
a simila way o he S c and CDK p o eins, and in e es ingly, e en
hough EGF ligand- ee wild ype EGFR is no phospho yla ed (no
ac i a ed) his p o ein adop s he ac i e o m con o ma ion in c ys-
als. Indeed, unlike o he y osine kinases, EGFR kinase is no as
‘‘ igh ly” au oinhibi ed, and can main ain an ac i e con o ma ion
and basal ac i i y a high concen a ions o EGFR o E bB2 he e o-
dime [15,47]. In ou p e ious s udy [39], we men ioned h ee
wild ype PDB s uc u es o ligand- ee EGFR, namely, PDB IDs
1m14, 2gs2, and 4 ks all showing he C-helix-IN (DFGin/BLAminus)
con o ma ion. Hence why 1m14 was selec ed as e e ence s uc-
u e o he C-helix-IN clan. The C-helix-OUT con o ma ion is ound
in 3gop which was used he e as e e ence s uc u e o he C-helix-
OUT clan (e en hough his does in ac comp ise a K745M
mu a ion).
James & Ve khi ke , [46] desc ibed he ollowing possible s a es
o ac i a ion o wild ype EGFR ( ollowing examples wi h com-
pounds complexed): (1) an inac i e s a e (C-helix-IN/DFGou ). (2)
a Cdk/S c inac i e con o ma ion 1 (C-helix-OUT/DFGin), e.g.,
PDB ID 1xkk. (3) a Cdk/S c inac i e con o ma ion 2
(C-helix-OUT/DFGou ), e.g., PDB ID 2 9. And (4) an ac i e s a e
(C-helix- IN/DFGin), e.g., PDB ID 2i x. In ou C-helix-IN clan, 19
complexed s uc u es we e iden i ied wi h wild ype EGFR in he
ac i e C-helix-IN (DFGin/BLAminus) con o ma ion (Supplemen-
a y Table 1). In he C-helix-OUT clan, nea ly 7 complexed wild-
ype EGFR s uc u es we e ound in he inac i e C-helix-OUT
(DFGin/BLBplus) con o ma ion (Supplemen a y Table 1).
The i s and mos common ype o mu a ion in EGFR, he
T790M mu a ion, occu s a exon 20 o he EGFR gene which is
esponsible o > 60% o he acqui ed esis ance cases in NSCLC.
He e, we obse ed T790M single mu an s wi h he DFGin con o -
ma ion in bo h C-helix-IN (10 s uc u es) and C-helix-OUT (10
s uc u es) clans, wi h nea ly 12 BLAminus con o ma ions
Table 1
Top equen di e en ial o ame s be ween C-helix-IN and C-helix-OUT clans (coun is shown a e each o ame ).
Residue C-helix-IN clan C-helix-OUT clan IN OUT To al p- alue
123123
Lys 714 m p 28 m 21 m 2 m p 11 m mm 11 m 6 52 32 84 <0.001
Ile 715 m 52 mm 11 mp 1 m 24 p 5 mm 5 64 36 100 0.012
Se 720 35 p 22 m 9 p 18 m 11 7 66 36 102 0.003
Phe 723 p90 30 m-85 1 m-30 1 m-85 9 p90 7 m-30 3 32 20 52 <0.001
Lys 728 m 24 p 7 6 23 p 4 p 3 41 33 74 <0.001
Glu 734 mm-40 13 0 9 m -10 2 0 10 m -10 6 p -20 3 27 21 48 0.016
Val 738 m 65 4 p 2 p 18 m 15 7 71 40 111 <0.001
P o 741 exo 52 exo 21 endo 17 52 38 90 <0.001
Lys 745 37 m 4 p 2 12 m 10 p 5 43 30 73 0.001
Th 751 m 45 p 7 2 p 8 m 7 5 54 20 74 <0.001
Se 752 p 41 16 m 2 p 10 8 m 6 59 24 83 0.005
Asn 756 m-80 19 m120 15 m-20 13 m-20 20 m-80 3 m120 3 49 27 76 0.001
Ile 759 43 p 6 m 5 m 28 mm 5 p 3 59 41 100 <0.001
Glu 762 0 52 m -10 9 p10 5 0 3 52 17 69 <0.001
Ty 764 80 57 80 14 m-85 10 m-30 3 57 27 84 <0.001
Se 768 m 64 p 30 m 6 p 2 64 38 102 <0.001
Val 769 m 72 m23 17 p 1 72 41 113 <0.001
A g 776 -85 14 180 6 85 4 p85 18 -85 5 p 85 2 34 31 65 0.001
Ile 780 p 72 mm 1 38 p 3 m 1 73 42 115 <0.001
Gln 791 0 62 0 17 m -30 11 62 28 90 <0.001
Me 793 mmm 67 mmp 2 mm 1 mmp 21 mm 9 m m 1 70 31 101 <0.001
Asn 808 m-80 28 m-20 6 m120 6 m120 12 m-20 5 30 5 41 26 67 <0.001
A g 832 m p180 29 m 180 6 m p85 4 m p85 11 m m-85 7 m p180 6 41 28 69 <0.001
Leu 833 p 16 12 m 11 m 31 4 pp 1 39 37 76 <0.001
A g 836 m p-105 56 m m105 1 m m180 1 m 85 9 m m105 7 m p-105 5 58 28 86 <0.001
Asn 842 m-20 73 m-20 28 m120 15 73 43 116 <0.001
His 850 -160 67 -80 6 -80 23 -160 17 m80 3 73 43 116 <0.001
Lys 852 m 60 m p 17 m 15 m pp 1 60 33 93 <0.001
Asp 855 70 23 0 21 m-20 4 0 14 m-20 10 70 6 48 30 78 0.006
Phe 856 m-85 53 p90 3 m-85 14 p90 9 80 7 56 30 86 <0.001
Lys 860 m 13 mm m 9 mm 7 8 mm m 6 mm 6 36 24 60 <0.001
Leu 861 m 13 p 9 mp 8 p 26 m 2 mp 1 36 30 66 <0.001
Glu 866 0 13 m -10 2 m-20 1 m -10 11 mm-40 1 p -20 1 17 13 30 <0.001
Ile 878 mp 30 mm 17 m 10 mm 29 mp 4 m 4 64 37 101 <0.001
A g 889 m 180 23 m 85 7 m -85 2 m 180 8 m 85 8 mm -85 4 33 25 58 0.043
Lys 913 m 48 m p 4 m p 3 m 12 m p 7 m pp 5 55 30 85 <0.001
Asp 916 70 47 0 25 0 26 70 15 72 41 113 0.003
P o 919 endo 54 exo 13 exo 29 endo 9 67 38 105 <0.001
Se 924 30 p 29 m 1 p 38 2 60 40 100 <0.001
Se 925 m 28 19 p 13 p 18 11 m 10 60 39 99 0.025
Ile 926 m 50 p 5 mm 3 m 18 p 14 pp 7 59 39 98 <0.001
Ile 941 61 p 1 p 19 10 m 5 62 35 97 <0.001
Me 945 pp 37 p 4 m 2 mmp 7 pp 6 m m 5 44 29 73 <0.001
Me 947 m p 28 m 13 m m 3 m m 10 m p 7 m 3 44 20 64 <0.001
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5447

(Supplemen a y Table 1). Hence, we would sugges ha he
T790M mu a ion is no c i ical in igge ing he elay o a ‘‘ low
o in o ma ion” om o ame o o ame o desensi ize EGFR
kinase o TKIs. Such a s a emen is no su p ising gi en ha
T790M mu a ion in ol es jus he eplacemen o h eonine – a
local ga ekeepe esidue and impo an de e minan o inhibi o
speci ici y in he ATP-binding pocke – o me hionine. This esidue
eplacemen dec eases i s and second gene a ion TKI binding
a ini ies o he ATP-binding pocke , owing o inc eased s e ic hin-
d ance, and inc eases he ATP-binding a ini y; hus, enhancing
Fig. 2. Di e en ial Colo map o o ame s in esidues o C-helix-IN and C-helix-OUT clans. Rows ep esen 116 EGFR s uc u es while columns ep esen 43 esidues o he
elay in o ma ion sys em. Skewed clus e ee is shown o con enience.
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5448
compe i ion o binding be ween ATP and TKIs [45,48,49]. Acco d-
ing o Yun e . al.,[48] who s udied he T790M mu a ion in bo h
ac i e and inac i e EGFR s uc u e s a es, he mu a ion is hypo h-
esized o al e di ec ly he con o ma ion o he DFG moie y in he
ATP-binding pocke om an inac i e o ac i e o m ia a ou able
hyd ophobic in e ac ions be ween M766 and L777, ha could lead
o changes in he posi ions o he DFG loop o C-helix.
In e ms o compa a i e binding da a, he ATP-binding a ini y
is highe o he T790M/L858R double mu an han he L858R
single mu an . The di e ence co ela es di ec ly wi h highe
esis ance owa ds ge i inib and e lo inib [50]. He e, we epo ed
nea ly 32 s uc u es o he T790M/L858R double mu an
(wi h 23 belonging o he C-helix-IN clan wi h C-helix-IN
(DFGin/BLAminus) con o ma ion, 5 s uc u es belonging o C-
helix-IN clan wi h C-helix-IN (DFGin/BLAplus) con o ma ion), and
9 s uc u es o he L858R single mu an ound in he C-helix-IN
clan wi h he C-helix-IN/DFGin/BLAminus con o ma ion (Supple-
men a y Table 1). Acco dingly, we would sugges ha he L858R
mu a ion could ac o igge he elay o a ‘‘ low o in o ma ion”
om o ame o o ame o sensi ize EGFR kinase o TKIs. Indeed,
Fig. 3. Schema ic d awing o he EGFR kinase domain showing he equencies o he op di e en ia ing o ame s be ween C-helix-IN (A) and C-helix-OUT (B) clans. In some
cases, he low pe cen age is ac ually ep esen ing he second op ank o ame as s a is ically di e en ia ing whe e he i s op ank o ame was simila (e.g., in Ile715 he
i s op ank o ame was m in bo h elays).
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5449
his L858R single mu an is a single missense mu a ion in exon 21
ha is also one o he mos equen EGFR al e a ions ound in
NSCLC umou s [51]. Mo eo e , L858R is e y equen ly mu a ed
o T790M/L858R double mu an s in cance pa ien s, such ha
o he L858R double mu an s a e ound wi h a bes only 5% inci-
dence [52-54]. Fu he mo e, i he T790M single mu an and he
T790M/L858R double mu an a e compa ed, al hough hey main-
ain he same low nanomola a ini y o ge i inib as he L858R sin-
gle mu an , he T790M single mu an exhibi s a highe ATP-
binding a ini y han he L858R single mu an . Acco dingly, he
T790M/L858R double mu an ep esen s an ac i a ed enzyme ha
becomes esis an o ATP-compe i i e TKIs [48].
Mo ing on o he C797S mu a ion, his is also loca ed in he ATP
binding pocke and p e en s co alen binding o co alen TKIs.
When cys eine is subs i u ed wi h se ine a codon 797, c oss-
esis ance is gained wi h espec o i e e sible hi d gene a ion
TKIs, such as Osime inib. In his ins ance, 2 s uc u es, double
mu an T790M/C797S and iple mu an T790M/C797S/V948R,
appea in he C-helix-OUT clan. The o me double mu an om
PDB ID 5xgn p esen s a DFGin/BLAminus con o ma ion wi h he
C-helix axis a equal angles (13°) o bo h 3gop and 1m14 e e -
ence s uc u es (Supplemen a y Table 1). The la e iple mu an
om PDB ID 5zwj is i mly wi h C-helix-OUT (DFGin/BLBplus)
con o ma ion (Supplemen a y Table 1). Acco dingly, he C797S
mu a ion could ac o igge he elay o a ‘‘ low o in o ma ion”
om o ame o o ame o desensi ize EGFR kinase o TKIs. In his
espec i is wo h no ing abou T790M/C797S double mu a ions
ha he e a e in ac h ee well desc ibed esis ance s a es: 1)
he cis T790M/C797S allelic s a e, whe e bo h mu a ions occu in
he same ecep o p o ein, which is esis an o all a ailable
EGFR-TKIs al hough sensi i e o ou h gene a ion, 2) he ans
T790M/C797S allelic s a e, whe e ei he o he wo exp essed
ecep o p o eins ha bou s one o bo h mu a ions, which is sensi-
i e o i s and hi d gene a ion TKIs, 3) a T790M mu a ion loss
combined wi h a C797S mu a ion gain which is sensi i e o i s
and he second gene a ion TKIs. E en hough he cis s a e domi-
na es, u he in es iga ion is needed o unde s and he s uc u al
di e ences in he ATP binding pocke ha a e associa ed wi h he
di e en mu a ional combina ions [55-59].
The al e na i e G719S mu a ion occu s in he phospha e-
binding loop (P-loop) which is conside ed TKI-sensi i e acco ding
o he Na ional Comp ehensi e Cance Ne wo k (NCCN, www.nc-
cn.o g) guidelines. He e, 5 cases o he mu an we e loca ed in
he C-helix-IN clan wi h ac i e C-helix-IN (DFGin/BLAminus) con-
o ma ion. In addi ion, a double mu an G719S/T790M is also
loca ed in he C-helix-IN (Supplemen a y Table 1. A guably, he
G719S mu a ion could ac o igge he elay o a ‘‘ low o in o ma-
ion” om o ame o o ame o sensi ize EGFR kinase o TKIs, in
his ins ance. Compu a ional s udies on he G719S mu a ion sug-
ges ha TKIs can en e he ATP-binding si e wi h ease. Indeed,
simula ions indica e ha he dis ance be ween esidues L718 and
G796 is inc eased widening he ATP-binding si e o TKIs o en e
(con e sely, he T790M mu a ion causes he dis ance be ween
L718 and G796 o dec ease) [60]. Mo eo e , he G719S mu a ion
Fig. 4. (A-D) EGFR kinase domain s uc u e (PDB ID 5ugc) showing he in o ma ion elay esidues as solid ligh blue su ace (C-helix in g een and mu a ions in ed). The es
o he kinase is shown in blue ibbon and anspa en sol en accessible su ace. Mos o he elay is connec ed and i s he low o in o ma ion om mu a ion o he es o
he elay (no e ha V948R mu a ion connec s o only ew esidues in he C-lobe, in which case he in o ma ion om V948R is ans e ed by o he means han o ame ic
mo es such as backbone, wa e and allos e ic e ec s). (Fo in e p e a ion o he e e ences o colo in his igu e legend, he eade is e e ed o he web e sion o his
a icle.)
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5450
des abilizes he inac i e con o ma ion and p omo es he ac i e
con o ma ion o he kinase, leading o mo e TKI sensi i i y [61-
65]. Howe e , when G719S is combined wi h T790M as a double
mu a ion, he seconda y T790M mu a ion o e u ns he impac
o G719S on he dis ance be ween he P-loop and ac i a ion loop
[60].
Finally, u ning o he T790M/C797S/L858R iple mu a ion,
s udies on a ge ed he apy – ia he new allos e ic inhibi o
EAI045 in combina ion wi h ce uximab – demons a e a di e en
mode o esis ance as compa ed o ha exhibi ed p e iously
[66,67]. EAI045 binds allos e ically o T790M ia a pocke ha is
acili a ed by ex e nal disloca ion o he C-helix. EAI045 is able o
achie e allos e ic binding o EGFR by binding o he mu an
M790 ga ekeepe esidue and o ming a hyd ogen bond wi h he
DFG mo i . A leas wo mechanisms accoun o he mu an -
speci ici y o he EGFR allos e ic inhibi o s: Fi s ly, he M790 ga e-
keepe esidue enhances he selec i i y o EAI045 o he T790M
mu an . Secondly, in he wild ype EGFR, EAI045 is unable o bind
e icien ly gi en he lack o allos e ic pocke in he kinase.
Ce uximab – a dime iza ion blocking agen - is usually used wi h
EAI045 o mimic he e ec o mu a ions ha dis up he asymme -
ic dime in EGFR. Basically, he allos e ic pocke in he L858R/
T790M mu an is accessible in he wo subuni s o he asymme ic
dime unlike in wild ype EGFR. The e o e, i is e y a ional o use
ce uximab o enhance he po ency o allos e ic agen s [67].
Ou knowledge o ano he iple EGFR mu an
T790M/C797S/V948R comes mainly om compa a i e binding
s udies be ween EAI001 and EAI045. EAI045 exhibi s a highe
a ini y o iple mu an s han i s p edecesso EAI001 o
T790M/V948R double mu an s [68]. This inc eased a ini y was
a ibu ed o he o ma ion o new hyd ogen bonds be ween
EAI045 and he backbone o F856. In his case, he C-helix is
pushed ou wa ds o accommoda e EAI045 binding and he o ma-
ion o mul iple hyd ophobic in e ac ions ia i s a oma ic ings
(pa icula ly wi h L747, I759, M766, L777, L788, M790, and
F856). On he o he hand, u he s udies a e equi ed o shed ligh
on he ole o T790M/L858R/V948R iple mu an s in EGFR esis-
ance o TKIs. In his ins ance, i is impo an o emphasize he ole
o dime iza ion dependency in unde s anding he s uc u e–
unc ion consequences o mu a ions. Fo example, while se e al
mu an s like L858R o G719S a e dime iza ion-dependen
( equi ing dime iza ion o oncogenic ac i a ion o EGFR), o he
mu a ions we e epo ed o be dime iza ion-independen . Indeed,
he V948R mu an ep esen s a su ace mu a ion in he C-lobe and
is known as a dime iza ion-de icien mu an , which is e y use ul
in unc ional s udies [69,70]. He e we ha e iden i ied 12 s uc u es
belonging o he C-helix-OUT clan wi h C-helix-OUT (DFGou /
BBAminus) con o ma ion (Supplemen a y Table 1). Mo eo e , all
he V948R mu an s (single, double and iple) belonged o he C-
helix-OUT clan, hus emphasizing i s ole in inac i a ion o EGFR
kinase. Acco dingly, we would sugges ha he V948R mu a ion
could ac o igge he elay o a ‘‘ low o in o ma ion” om o a-
me o o ame o desensi ize EGFR kinase o TKIs.
3.3. Biological signi icance o EGFR elay sys em
EGFR is a pa o he signalling p ocesses in ol ed in cell- o-cell
communica ion sys em [71]. The e o e, his ecep o is cen al o
no mal as well as cance cell iabili y. Du ing chemo he apeu ic
in e en ions, EGFR becomes unde emendous selec i e p essu e
o main ain i s ac i i y, in o de o p omo e cell g ow h and
p oli e a ion. This is clea ly demons a ed by he de elopmen
and subsequen obsolescence o h ee gene a ions o TKIs h ough
he appea ance o a combina ion o inna e ( andom) and adap i e
(induced) EGFR mu a ions. Acco ding o ou analysis, EGFR has
unde gone an adap i e and cumula i e sequence o h ee poin
mu a ions which could ac singly o oge he o induce d ug esis-
an con o ma ional changes in EGFR ha a e communica ed by a
‘‘ low o in o ma ion” om o ame o o ame ia side chain con-
o ma ional elays (Figs. 4 and 5A)[72-75]. Each con o ma ional
elay ep esen s a chain o mu a ion-induced, linked changes
(domino-like) in amino acid esidue o ame con o ma ions, ha
we p opose cause he displacemen o a whole helix moie y wi hin
he y osine kinase subdomain (C-helix-OUT). The combined
e ec s o his con o ma ional elay p esen s a si ua ion whe e
TKI inhibi o s no longe ha e a sui able binding pocke o bind o
and inhibi EGFR, and mu an EGFRs hemsel es become mo e
agg essi e agen s o signal ansduc ion wi hou he need o yp-
ical y osine kinase ac i i y [76]. Indeed, such mu an EGFRs p e-
se e he ‘‘in o ma ional sys em” wi h sus ained p o-p oli e a i e
signalling ha is p o cance cell su i al, [38] leading o mo e
agg essi e umou p og ession han is possible wi h wild ype
EGFR, as obse ed in NSCLC [77].
The ailu e o h ee gene a ions o TKIs o inhibi he EGFR com-
munica ion sys em can be unde s ood using Shannon’s Biological
In o ma ion Theo y (Fig. 5B)[72-75]. B ie ly, EGFR wi h a mu an ,
inac i e y osine kinase is s ill capable o ansmi ing signals and
in o ma ion, hence he kinase unc ion is ela ed o g ow h no su -
i al [76]. The e o e, in he case o wild ype EGFR d ug ea men
(No mal sensi i i y, W > M = C-helix-IN > C-helix-OUT), he cell
will sea ch o ways o de elop d ug esis ance o main ain g ow h
ansmission [36]. Hence, e en i he wild ype EGFR has been a -
ge ed co ec ly his will only a ec umou g ow h empo a ily no
o e all su i al [78]. Cance cells exp essing wild ype EGFR do no
ely on he EGFR kinase ac i i y bu on EGFR o su i al [79,80].
The same is ue in he case o he mu an ype whe e EGFR is
al eady und uggable (Highe sensi i i y = M W = C-helix-OUT
C-helix-IN). In his case, eme ging mu a ions will con e he
o ame ic elay and u he desensi ise EGFR o TKIs. The e o e,
should physicians insis on using TKIs on EGFR un il mo e mu an
inac i e EGFR kinases a e ealized (Ul a sensi i i y = M >>>
W = C-helix-OUT >>> C-helix-IN) [81,82], hen he esul mus
be an und uggable p o ein ha will accele a e he signalling cas-
cade, p omo e cance cell su i al, and wo sen disease p ognosis
[83-85].
Seen om a di e en angle, he use o i s , second and hi d
gene a ion TKIs leading o he widesp ead appea ance o und ug-
gable mu an EGFRs can be seen and unde s ood h ough he lens
o game heo y (Fig. 5C)[86,87]. Cance cells a e known o hei
adap i i y, bu hey can nei he an icipa e no e ol e adap a ions
o ea men s ha he physician has no ye applied. The e o e,
a dis inc i e leade – ollowe (o ‘‘S ackelbe g”) dynamic mus
apply, in which he oncologis ‘‘leade ” plays i s and he cance
cell ‘‘ ollowe ” hen esponds and adap o he ea men egime.
The physician ‘‘plays” a ixed s a egy e en while he opposing can-
ce cells con inuously e ol e coun e measu es un il disease p o-
g ession is no longe hal ed [88]. Fu he mo e, by changing
ea men only when he umou p og esses, he physician aban-
dons leade ship o he cance cells and ea men ailu e becomes
nea ly ine i able. A he molecula le el, we obse e he end p o-
duc o his game heo y challenge. In he case o EGFR, obsessi e
use o one class o d ugs o one key a ge ende s he a ge
und uggable as a esul o only h ee mu a ions and hei linked
con o ma ional elays (Fig. 5D). Ou s uc u al and compu a ional
da a highligh he need o adop mo e sophis ica ed combina ion
app oaches o ea men in o de o o e whelm umou s be o e
hey can moun di ec adap i e changes a he molecula le el ha
lead o esis ance o ea men . While ou s uc u al and compu a-
ional da a accoun o TKI insensi i i y, we do no cu en ly ha e
an equi alen molecula le el unde s anding o how mu an
EGFRs possess heigh ened signal ansduc ion o cance cell
p oli e a ion.
T. Hameduh, M. Mok y, A.D. Mille e al. Compu a ional and S uc u al Bio echnology Jou nal 19 (2021) 5443–5454
5451