Available via license: CC BY 4.0
Content may be subject to copyright.
1Searching deeply into the conformational space of glycoprotein
2hormone receptors. Molecular dynamics of the human follitropin and
3lutropin receptors within a bilayer of (SDPC) poly-unsaturated lipids.
4
5Eduardo Jardón-Valadez1,¶,*, Alfredo Ulloa-Aguirre2,3,¶,*
6
71Departamento de Recursos de la Tierra, Unidad Lerma, Universidad Autónoma
8Metropolitana, Lerma de Villada, Estado de México, Mexico.
9
10 2Red de Apoyo a la Investigación, Universidad Nacional Autónoma de México. Ciudad
11 de México, Mexico.
12 3Instituto Nacional de Ciencias Medicas y Nutrición “Salvador Zubiran”. Ciudad de
13 México, Mexico.
14
15 *Correspondence:
16 E-mail: h.jardon@correo.ler.uam.mx (E. Jardón-Valadez)
17 and aulloaa@unam.mx (A. Ulloa-Aguirre)
18
19 Keywords. Glycoprotein hormone receptors, molecular dynamics, conformational
20 energy landscape.
21
22
23
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
24 Abstract
25 Glycoprotein receptors are a subfamily of G-protein coupled receptors, including the
26 follicle hormone (FSH) receptor (FSHR), thyroid-stimulating hormone receptor (TSH), and
27 luteinizing/chorionic gonadotrophin hormone receptor (LHCGR). These receptors display
28 common structural features such as a prominent extracellular domain, with a leucine-rich
29 repeats (LRR) stabilized by -sheets, a long and flexible loop known as the hinge region
30 (HR), and the transmembrane (TM) domain with seven helices interconnected by intra-
31 and extracellular loops. Binding of the ligand to the LRR resembles a hand coupling
32 transversally to the and subunits of the hormone, with the thumb being the HR. The
33 structure of the complex of FSHR-FSH suggests an activation mechanism in which Y335
34 at the HR binds into a pocket between the and chains of the hormone, leading to
35 an adjustment of the extracellular loops. In this study, we performed molecular dynamics
36 (MD) simulations to identify the conformational changes for the FSHR and LHCGR. We
37 set up an FSHR structure as predicted by AlphaFold (AF-P23945); for the LHCGR
38 structure we took the cryo-electron microscopy structure for the active state (PDB:7FII)
39 as initial coordinates. Specifically, the flexibility of the HR domain and the correlated
40 motions of the RLL and TMD were analyzed. From the conformational changes of the
41 LRR, TMD, and HR we explored the conformational landscape by means of MD
42 trajectories in all-atom approximation, including a membrane of polyunsaturated
43 phospholipids. The distances and procedures here defined may be useful to propose
44 reaction coordinates to describe diverse processes such as the active-to-inactive
45 transition, to identify intermediaries suited for allosteric regulation, and biased binding to
46 cellular transducers in a selective activation strategy.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
47 Author summary
48
49 In the present study, we describe the results from a computational microscopy
50 perspective (also known as molecular dynamics simulation) at the atomistic resolution for
51 the two gonadotropin hormone receptors, the follicle-stimulant hormone receptor and the
52 luteinizing/chorionic gonadotropin hormone receptor, which are essential for reproduction
53 in humans. Several dysfunctional mutations in these receptors, leading to reproductive
54 failure, have been detected in the clinical arena. To better understand the process
55 whereby these two receptors perform their signaling tasks, triggering an intracellular
56 response upon binding of their cognate agonist at the extracellular side, we assembled
57 the receptor structures in a membrane bilayer of phospholipids with water molecules as
58 solvent at both sides of the membrane. The systems included nearly 200 thousand atoms,
59 each moving around at 300 kelvin and 1 bar given the interactions (attractive or repulsive
60 forces) from each other. As the motion equations are solved in each time step (at
61 femtoseconds time scale), the system evolves over time during hundreds of nanoseconds
62 (millions of time steps) for three independent replicates. The receptor conformation,
63 therefore, may display non-random motions due to the stability of specific structures in
64 the complex molecular environment, including the hydrophobic membrane core, the
65 bilayer interfaces, and the aqueous medium. From analysis of simulation trajectories and
66 structural changes of the receptors, we could identify the main conformational changes
67 exhibited by each receptor explored in a model cellular environment. We discussed the
68 roll of the hinge domain at the extracellular domain in triggering the receptor
69 conformational changes, as well as differences in the dynamics between these receptors
70 in terms of the flexibility of the structures. Importantly, we proposed relative distances
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
71 among the different receptor domains as parameters to characterize conformational
72 intermediaries along a transition of states. Understanding of the signaling process in
73 gonadotropin hormone receptors could be useful to explore new strategies for the
74 modulation of the receptor functions, the bias of signaling pathways, or the selective
75 binding of agonists.
76
77
78
79
80
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
81 INTRODUCCIÓN
82 G protein-coupled receptors (GPCRs) are a large and functionally diverse superfamily of
83 plasma membrane receptors that respond to a widely variable of endogenous and
84 exogenous stimuli of diverse structures, from photons, odorants, and ions to lipids,
85 neurotransmitters, peptide hormones and complex protein hormones. G protein-coupled
86 receptors consist of single polypeptide chains of variable lengths which traverse the lipid
87 bilayer seven times, forming characteristic transmembrane (TM) -helices, connected by
88 alternating extracellular and intracellular loops, with an extracellular NH2-terminus or
89 ectodomain (ECD) and an intracellular COOH-terminus or Ctail [1, 2]. These membrane
90 receptors currently represent an important therapeutic target for several diseases in
91 humans; in fact, 30-40% of GPCRs of approved drugs target this family of membrane
92 receptors [3, 4].
93
94 The receptors for the glycoprotein hormones (GPHR) [follicle-stimulating hormone
95 (FSH) receptor or follitropin receptor (FSHR), lutropin receptor or luteinizing
96 hormone/chorionic gonadotropin (LH and CG, respectively) receptor (LHCGR), and
97 thyrotropin receptor or thyroid stimulating hormone (TSH) receptor (TSHR)] belong to a
98 conserved subfamily of a GPCR family, the so-called Rhodopsin-like family, and more
99 specifically, to the -group of this large class of GPCRs [1]. Structural features of the
100 GPHRs include a large extracellular NH2-terminus, where recognition and binding of their
101 corresponding ligands (FSH, LH and CG, or TSH) occur. This ECD exhibits a central
102 structural motif of imperfect leucine-rich repeats (LRR; 12 in the FSHR; 8-9 in the LHCGR;
103 10 in the TSHR) [5-8], that is shared with other plasma membrane receptors involved in
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
104 selectivity for ligand and specific protein-to-protein interactions (Fig. 1A)[9]. The carboxyl-
105 terminal end of the ECD exhibits a domain critical for all GPHRs function: the so-called
106 hinge region (HR); this particular region is involved in high affinity hormone binding,
107 receptor activation and intramolecular signal transduction, and/or silencing of basal
108 receptor activity in the absence of a ligand [10].
109
110 The gonadotropin receptors (FSHR and LHCGR) play a critical function in
111 reproductive function. They regulate spermatogenesis and ovarian follicular maturation
112 (FSHR) as well as steroidogenesis (testicular Leydig cells, ovarian follicle, and placenta)
113 as well as ovulation (LHCGR) [7, 11-13]. We have been interested in analyzing the effects
114 of point mutations on function and tridimensional structure of the FSHR [7, 12, 14]. We
115 have particularly focused on the understanding of the response of these receptors to their
116 cognate agonists departing from changes in the conformational dynamics related to the
117 biological function in dysfunctional variants [14, 15]. In principle, an extracellular signal is
118 transmitted when the receptor is stabilized in an active conformation, which allows
119 activation of the intracellular transducers coupled to the receptor. Nevertheless, the
120 transition processes among different conformational states of the receptor triggered by
121 the agonist are still incompletely understood [16, 17]. For example, the FSHR may
122 activate several intracellular signaling cascades in function of coupling to distinct
123 pathways mediated by several kinases and/or arrestins, that is to say, the receptors
124 are more than simple binary interruptors, but rather function as allosteric microprocessors
125 that respond to the ligand stimulus with different affinities to distinct transducers: a given
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
126 ligand may favor activation of a particular signal transducer and, in turn, the transducer
127 may increase the affinity of the receptor to the ligand [18].
128
129 Departing from the elucidation of the LHCG receptor in the active and inactive state
130 by cryo-electron microscopy (3.5 Å), it is possible to identify the structural features of the
131 GPHR subfamily [19]. The large extracellular ECD encompass the LRR, similar to a
132 boxing glove, with the thumb formed by the HR. The -helix (P272-N280) and the loop
133 P10 (F350-Y359) of the HR conform a communication interface between the LRR and
134 the TM domain. For example, the -helix Q425-T435 of the EL1 and P10 exhibit
135 interactions in the inactivating S277I and activating E254K mutations, suggesting that the
136 positions K354 and K605 are important regulators of the activation or inhibition of the
137 receptor [20]. On the other hand, the C304-C353 disulfide bond was not observed in the
138 structure, in contrast to the C292-C338 bond in FSHR (Fig. 1A). Due to the absence of a
139 crystal structure of the FSHR, our group has proposed a model for this receptor
140 considering only the TM domain (FSHR-TM) [15, 21]. From the structural alignments of
141 the LHCGR vs FSHR and FSHR vs FSHR-TMD, it is possible to identify that our FSHR-
142 TM model corresponds to the active state of the FSHR (Fig. 1B). Among other findings
143 derived from the trajectory analyses of molecular simulations of the FSHR and D408Y
144 and I423T inactivating mutants, it was identified that helix 2 of the TM domain is an
145 important communication hub for the propagation of intracellular signaling [14, 15].
146
147 As a continuation of previous studies [14, 15] in the present work we performed
148 all-atom simulations for a FSHR model, but now including the corresponding LRR, HR
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
149 and TM, as disclosed by the AlphaFold server (AF2; Fig. 1). Examination of the
150 conformational energy landscape was performed for a FSHR model and the structure of
151 the LHCGR (PDB:7FII). Both GPCRs were relaxed in similar conditions through a
152 simulation box that included water molecules as solvent, a polyunsaturated phospholipid
153 membrane, and monovalent ions for charge balance. A difference to highlight between
154 both structures is the presence of the disulfide bond C292-C338 at the HR of the FSHR,
155 which is not observed in the equivalent position (C304-C336) in the LHCGR [19, 20].
156
157 In the study of protein folding, it has been proposed the notion of the “energy
158 funnel” that stretches down as the energy decreases [22]. When a protein is unfolded,
159 without a well-defined structure, its possible conformations show high variability that is
160 represented by the top width of the funnel; as the molecular interactions favor the
161 formation of contacts among residues, (e.g. hydrogen bonds formation), the free energy
162 of the protein will decrease and the overall, three dimensional shape of the protein (i.e.
163 its stable set of conformations) will be better defined, moving towards the lower portions
164 of the funnel. When the protein reaches its native state, the overall conformation will move
165 towards the end of the funnel exhibiting a minimal free energy. The progression of
166 changes from the unfolded to the native (folded) state is rather a rough surface, with local
167 minima and metastable states that should evolve to the more stable conformations. In a
168 given GPCR, as the FSHR, the surface of the conformational energy landscape may
169 show more than one stable state that represent either the inactive state or distinct active,
170 signaling states that may lead to full or biased signaling [23]. In this scenario, knowing the
171 transition route to favor a particular signaling over other(s), represent, on one side, the
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
172 understanding of the conformational changes occurring during the signal propagation
173 process and, on the other, the possibility to control the selective signaling of a GPCR
174 through different agonists, antagonists or allosteric modulators. The present study
175 explores the energy landscape of the LRR-HR-TM in gonadotropin receptors, to identify
176 the configurations compatible with the opening of their intracellular domains during
177 activation.
178
179 From the computational point of view, the identification of critical sites for the
180 allosteric regulation of protein function has been analyzed employing different
181 approaches including machine learning, bioinformartics to detect allosteric sequences,
182 site-directed mutagenesis, and molecular dynamics [24-26]. A widely employed
183 technique to describe conformational changes of a protein is the analysis of the principal
184 component (PC), which consists in the calculation of the covariance matrix whose
185 diagonalization gives rise to a coordinate transformation [27]. The eigenvalues contain
186 the mean square fluctuations associated to each eigenvector, ordered in decreasing
187 order. Through the trajectory projection on eigenvectors, the set of PC discloses the low
188 frequency motions captured in the simulation. Typically, the first PCs accumulate the
189 larger variability with respect to a reference structure (e.g. average structure). There are
190 different indicators to establish whether the PC corresponds to sampling with a
191 representative variability of conformational changes, for example, the calculation of the
192 autocorrelation function, the contribution of the thermic fluctuation (content of the
193 temporal periodicity of the cosine function in a PC), and the calculation of the total
194 fluctuation as the square root of the sum of eigenvalues divided by the number of atoms,
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
195 among others [28]. The calculation of the probability distribution of the first components
196 (e.g. CP1-4) is of particular interest since a projection of the free energy (∆G) can be
197 calculated form the corresponding histograms in 1 to 3 dimensions. As a result, it is
198 possible to obtain conformations that cluster together as function of the PC values,
199 allowing identification of those conformations with the higher probability as well as the
200 energy barriers associated with a transition. In addition, it is possible to detect
201 intermediate states through which the transition is carried out, which is quite informative
202 for the description of the transition states and the landscape of the conformational energy.
203 Nevertheless, there are effects that may interfere in the analysis: i. The adjusting effect
204 to a reference structure (initial, final or average structure) to eliminate translational or
205 rotational displacements; ii. The contribution of highly flexible zones that may affect the
206 adjustment with the reference structure; and iii. The time scale of the simulation that may
207 not be sufficient for sampling convergence [29, 30]. Some strategies to mitigate these
208 limitations are: i. The use of internal coordinates instead of cartesian coordinates, that is,
209 the analysis can be performed using dihedral angles and distance between atoms or
210 domains; ii. To restrict the analysis on rigid regions; and iii. To carry out bias simulations
211 through the definition of a reaction coordinate (collective variable), implementation of an
212 accelerated sampling method as the replica exchange, trajectories at high temperature,
213 metadynamics, among some [31]. In the present study the exploration of the
214 conformational surface of the GPCRs was performed through a combination of strategies,
215 such as the use of the distances or angles between the LRR, HR and TM, and PC
216 analysis. This study has the perspective of evaluating different coordinates of reaction
217 that allow to capture the transition routes through the intermediates compatible with the
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
218 conformational switches known to be involved in GPCR activation, namely the D/ERY
219 ionic motif, the displacement of TM helix 6, and the rearrangement of helices 5 and 7 [2,
220 32]. In addition, a trajectory of the LHCGR in 1-palmitoyl-2-oleoyl-sn-glycero-3-PC
221 (POPC) was generated as the continuation of the initial processing in CHARMM-GUI for
222 further comparisons.
223
224
225 RESULTS
226
227 Detection of conformational changes from domain distances
228
229 Describing the conformational states for the glycoprotein hormone receptors, obtained
230 either by cryo-electron microscopy or computational models, is relevant for understanding
231 of the mechanisms associated with receptor function. In cases when the resolved
232 structure is incomplete, modeling strategies emerge as an alternative approach to unveil
233 processes at the atomic level. To convey the molecular complexity by which membrane
234 receptors link an extracellular stimulus with an intracellular response, in Fig. 1A we show
235 the relaxed structures of the FSHR (AF-P23945) and LHCGR (PDB:7II) in a SDPC
236 membrane environment, identifying the LRR, HR, and TM domains; in addition, Fig. 1B
237 shows the TM domain of FSHR-TM model in comparison to the full FSHR model. From a
238 structural alignment of the helices, the comparison suggests consistent predictions on the
239 relative positions of the TM helices as well as the location of residues, shown by the
240 cysteine residues as an example (Fig 1B). In previous studies, the FSHR-TM model,
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
241 including D355 to N678, was developed before the breakthrough of the AF2 method for
242 predicting native structures. The good agreement of the FSHR structures obtained
243 provides an additional criterium to validate our former FSHR-TM model [15]. By structural
244 alignment of the LRR of the FSHR (A48-T270) and the LHCGR (T52-T274), a RMSD of
245 2.40 Å was calculated, whereas for the HR domain of the FSHR (P276-T345) and the
246 LHCGR (P276-P345), the calculated RMSD was 17.9 Å. By aligning the sequence of the
247 TM of the FSHR (Y362-F630) and the LHCGR (D360-T630), we calculated a RMSD of
248 6.1 Å. The relative displacement between the FSHR and the LHCGR at the intracellular
249 side of helix 6 was 9 Å, which is close to the 14 Å displacement between the reported
250 active and inactive states of the LHCGR [19]. In rhodopsin, the opening of the intracellular
251 sides of TM6, along with rearrangement of helices 5 and 7, favors the open conformation
252 for coupling of G proteins, G protein-coupled receptor kinases (GRK), and arrestins [32].
253 Therefore, from the position of the TM5 and TM6, which is consistent with an open
254 intercellular conformation, the FSHR model obtained corresponds to the active state. The
255 largest difference between the FSHR and LHCGR was observed in the HR domain, which
256 deserves a detailed analysis because its role on triggering the activation process of these
257 receptors [20, 33]. In this study, we explored the motion of the LRR, HR, and TM
258 consistent with the active state. For this purpose, we defined positions at each domain
259 and calculated the relative distances, which may serve as reaction coordinates to follow
260 the transition among conformational states. In table 1 the atoms selected to measure
261 relative distances of the receptor domains, and their values at the starting conformation
262 are shown. For the LRR we chose a residue at the middle of the first -strand, for the HR
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
263 a residue at the middle of the -helix, and for the TMD a residue at the middle of TM helix
264 3.
265
266 Table 1. Relative distances among the LRR, HR, and TM
267 domains of the FSHR and LHCGR
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
Distance
RHLCG
RHEF
RLL-BR
(S55-C304)
66 Å
(L31-Y303)
63 Å
TM-BR
(L452-C304)
32 Å
(L460-Y303)
56Å
RLL-TM
(S55-I431)
32 Å
(L31-L460)
56Å
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
283
284 Fig. 1. A. Simulation boxes for the FSHR (orange structure) and the LHCGR (purple
285 structure). The leucine rich region (LRR), hinge region (HR) and transmembrane (TM)
286 domain are indicated. The SDPC bilayer is depicted with the lipid tails as white spheres,
287 and the phosphorus atoms of the lipid heads as olive green spheres. Solvent water
288 molecules are depicted as small spheres in blue purple. Side chains of cysteine residues
289 are depicted in licorice, green for the FSHR and yellow for the LHCGR. Disulfide bonds
290 are identified for C292-C338, C276-C356 for FSHR, and C279-C336 of LHCGR. The
291 bond C292-C338 at the HR of FSHR was not defined between C336 and C304, which is
292 the homologous positions in the LHCGR. B. Structural alignment between FSHR (cyan
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
293 structure) and the FSHR-TM (orange structure); for reference, C of cysteine residues
294 are depicted as spheres in yellow and in blue for the FSHR-TM and FSHR, respectively.
295 The open intracellular conformation suggests that the models corresponds to the
296 activated conformation. Models were developed using different methods and the
297 agreement provided an additional criterium for validation of the previously reported model
298 [15].
299
300
301 Fig. 2. Distances among structural domains of the FSHR. Numbers in parenthesis
302 indicate the correlation coefficients for every pair of distances. Color code: R1-blue, R2-
303 magenta y R3-gray.
304
305 The comparison among relative distances LRR-TM, LRR-HR, and TM-HR are shown in
306 Fig. 2, including the correlation coefficient for each replicate. The overlapping areas
307 represent conformations with the same values of relative distances (internal coordinates)
308 and larger colored areas represent broader distributions. In R1, distributions were mainly
309 unimodal centered at 70 Å, 85 Å, and 93 Å, for LRR-HR, LRR-TM, and TM-HR distances,
310 respectively. Positive correlations calculated for distances LRR-TM and TM-HR were of
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
311 0.40, 0.55, and 0.4, for R1, R2, and R3, respectively (Fig. 2C). The increase of the LRR
312 and HR distance relative to TM was consistent with a shortening of the LRR and HR
313 distance, as can be observed in R1 by the negative correlation between LRR-HR and
314 LRR-TM. From the distributions of relative distances among the LRR, HR, and TM
315 domains, covering broader intervals (e.g. 80 to 100 Å between TM and HR, or 72 to 93 Å
316 between LRR and TM (Fig 2C), they can be used as reaction coordinates to identify
317 conformational intermediates. Interestingly, in replicate R3 the distance variability was
318 smaller than in replicates R1 and R2, which can be related to the sampling of a metastable
319 state. Metastable states may prevent transitions among intermediaries and alternative
320 strategies must be implemented for an extensive sampling of the conformational space.
321 For example, trajectories can be generated with configurations harvested from a previous
322 run, as it is described below.
323
324 The correlations for distances in LHCGR are shown in Fig. 3 for replicates R1-3 and for
325 the trajectory in POPC. From the trajectory in POPC, receptor configurations were
326 harvested at t=0 ns (R1), t=100 ns (R2), and t =180 ns (R3). Given the initial
327 configurations of R2 and R3, these trajectories displayed no overlaps with replicate R1.
328 Instead, the trajectory in POPC connected the explored regions by R2 and R3 with R1.
329 Interestingly, the trajectory in POPC showed multimodal and broader distributions than
330 replicates R1-3 for distances LRR-HR and TM-HR (Fig. 3B). All trajectories consistently
331 showed negative correlations between LRR-HR vs TM-HR (Fig. 3B), which indicated that
332 the HR was moving away from the LRR, approaching to TM. Correlations of distances
333 LRR-TM vs TM-HR in the LHCGR (Fig. 3C) showed different trends than those in the
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
334 FSHR (Fig. 2C); the concerted motion of the LRR and HR relative to TMD in FSHR was
335 not detected in the LHCGR. In summary, from the sampling of receptor conformations in
336 independent replicate trajectories, starting from the same initial configuration (with
337 restarted velocities) or from conformations harvested from a previous trajectory, it was
338 possible to prevent sampling of metastable states. In fact, the weighted ensemble (WE)
339 method takes advantage of generating multiple, short trajectories departing from different
340 initial configurations and restarting trajectories, whenever the reaction coordinate
341 populates a new bin along the state’s transition [34]. In Fig. S1 of SI_info, it is shown a
342 plot of the RMSD for the active state of the LHCGR obtained from 100 iterations
343 implementation of the WE method. The RMSD can be used, in fact, as reaction coordinate
344 to track the transition between the active to inactive states.
345
346
347 Fig. 3. Distances between structural domains of LHCGR. Numbers in parenthesis are the
348 correlation coefficients for every pair of distances. Color code: R1 gray, R2 magenta, and
349 R3 blue, POPC carbon gray.
350
351 Conformational analysis of LHCGR
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
352
353 Fig. 4 shows RMSD matrices for the LHCGR replicates R1-3 and the POPC; insets also
354 show cluster conformational analysis with a cut-off criterium RMSD <0.5 Å. For the RMSD
355 calculation, only the TM helices were fitted in time frames at t and t+∆t: TM1, 360-387;
356 TM2 392-420; TM3, T437-470; TM4, 480-504; TM5, 522-552; TM6, 565-597; and TM7,
357 602-625. The red solid line in the plots was drawn to identify self-similar groups. Not
358 surprising, similar structures were found along the diagonal as conformations in
359 consecutive frames differed within the cutoff criterium. RMSD values in the TM1-7 helices
360 were lower than 3.1 Å in SDPC and lower than 2.8 Å in POPC. In Fig. 4A, self-similar
361 groups were detected at the ~10 ns time scale in R1. In R2, a group was detected in the
362 first 50 ns and other after 100 ns (Fig. 4B). In R3, RMSD values showed closer differences
363 in comparison with R1 and R2. In the cluster analysis, a structure was added whenever
364 it showed a RMSD within the cut-off of any of the members in that cluster. For example,
365 with a cut-off of 1 Å only one cluster was detected; therefore, we set a 0.5 Å for detection
366 of larger number of clusters. Clusters were calculated for each replicate in SDPC and the
367 POPC trajectory, altogether encompassing ~800 ns of conformational sampling.
368
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
369
370
371 Fig. 4. RMSD matrix analysis (gray scale) for the TM helices of LHCGR. (A) replicate R1,
372 (B) replicate R2, (C) replicate R3, and (D) trajectory in POPC. Red solid lines along the
373 diagonal were drawn for visual identification of self-similar groups. The cluster analysis is
374 shown in the insets. Conformations in clusters include structures within 0.5 Å of RMSD
375 among each other. The cluster number is identified in the vertical axis. Horizontal
376 segments along the curve identify the time frames forming a given cluster.
377
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
378
379 PC analysis for the LHCGR runs is shown in Fig. 5 by means of 2D projections of
380 replicates R1-3 over the first four eigenvectors. For the largest clusters found in each
381 replicate, the set of conformations were also projected over the eigenvectors: cluster 6383
382 of R1, clusters 4193 and 2595 of R2, and clusters 1404 and 2964 of R3. Conformations
383 in clusters were plotted in the context of the full trajectory (Fig 5). Further, in Fig. 6 the
384 distributions for each PC of the replicates are shown, providing an additional context in
385 which the clusters were located. For example, conformations in cluster 6383 of R1 were
386 not at the maximum of -8 nm, but instead they were located at 0 nm near of the tail of the
387 distribution (Fig 5A). In R2, the PC1 varied from -50 nm to 10 nm, with maxima at -30 nm,
388 -8 nm, and 8 nm; cluster 4193 contributed with conformations at the 8 nm maxima,
389 whereas cluster 2595 did so with conformations at -30 nm. Cluster 2964 of R3 had
390 conformations at the maximum of 2.3 nm of PC1; cluster 1404, around the maximum of
391 22 nm (Figs. 5C, 6A). Because of the broad shape of the PC3 of R1 (Fig 6C), cluster 6383
392 showed values that spread from -5 nm to 5 nm, but right on maximum of PC4 at -2.7 nm
393 (Fig. 6D). Cluster 4193 of R2 showed a population at the maxima of PC3 (-5.4 nm) and
394 PC4 (-4.0 nm); and cluster 2595 at 7.5 nm of PC4 (Figs. 5E, 6D). Finally, clusters 2964
395 and 1404 of R3 (Figs. 5F, 6C) had conformations at maxima of PC3 (3.8 nm) and PC4
396 (3.3 nm; Fig 6D). By combining cluster and PC analysis for the TM conformations, we
397 could identify those conformations that most likely contributed to the fluctuations of the
398 low frequency motions, that is, the intermediary states in a free energy landscape.
399
400
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
401
402
403 Fig. 5. Principal component analysis for LHCGR. PC1 and PC2 (A-C) and PC3-PC4 (D-
404 F) of trajectories R1-3 (blue dots). Projections for cluster 6383 of R1 (A, D), 4193 of R2
405 (B,E) and 2964 of R3 (C,F; black dots); clusters 2595 of R2 (B,E), and 1404 of R3 (C,F;
406 purple points). TM domain conformations in clusters were identified with a RMSD <0.5 Å,
407 ie. a structure is included whenever its RMSD is lower than 0.5 Å from any of the members
408 in that cluster.
409
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
410
411
412
413 Fig. 6 Distributions of the trajectory projections on the first four eigenvectors for the
414 LHCGR in replicates R1 (brown), R2 (orange), and R3 (brown-orange). PC1 (A), PC2 (B),
415 PC3 (C), and PC4 (D).
416
417 Conformational analysis of FSHR
418
419 The motion of the receptor LRR and HR domains detected by the relative distances, can
420 be related to the fluctuations of the TM as a correspondence between the dynamics of
421 the aqueous and membrane environments. It seems evident that by including the
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
422 variability of the LRR and HR domains together with the fluctuations of the TM helices in
423 a conformational analysis, makes difficult to identify the transition states due to the
424 slightness of helices motions [32]. In Fig 7 the RMSD matrices and the cluster analysis
425 for replicates R1-3 of the FSHR are shown. Only the TM helices were fitted for structures
426 taken at time t and t +∆t: TM1, Y362 to Y392; TM2, P397 to Y432; TM3, N437 to T472;
427 TM4, A487 to G507; TM5, M532 to R557; TM6, D567 to L597; and TM7, A607 to Y626.
428 Calculated values of RMSD were lower than 2.5 Å, 2.2 Å and 1.9 Å, for R1, R2, and R3,
429 respectively. The cluster analysis is included in insets of Fig. 7, using a RMSD cut-off
430 <0.5 Å; a structure is included whenever the RMSD was <0.5 Å in any of the members of
431 that cluster.
432
433
434
435 Fig. 7. RMSD matrix analysis (gray scale) for the TM helices of FSHR. A. replicate R1,
436 B. replicate R2, and C. replicate R3. The cluster analysis is shown in the insets.
437 Conformations in clusters include structures within 0.5 Å of RMSD among each other.
438 The cluster number is identified in the vertical axis. Horizontal segments along the curve
439 identify the time frames forming a given cluster.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
440
441 The 2D projections of replicates R1-3 on the first four eigenvectors are shown in Fig. 8:
442 PC1 vs PC2 and PC3 vs PC4. The RMSF of PC1 represents ~70% of total fluctuation,
443 whereas the RMSF of PC1 to PC4 ~95% (Fig. S2 in SI_info). Clusters 3405, 1171, and
444 0733 were identified for R1, R2, and R3, respectively, with the largest number of
445 conformations within the RMSD cut-off. To identify the most populated intermediaries in
446 the context of the full trajectory, projections of the clusters on eigenvectors were included
447 in Fig. 8. In addition, Fig. 9 shows the distribution of the PC1 to PC4 for each replicate.
448 Cluster 3405 of R1 showed values of -0.5 Å in both PC1 and PC2 (Fig 8A), which was
449 consistent with the maxima of their distributions (Figs. 9A, 9B). Due to the broad
450 distributions of PC3 and PC4 (Figs 9C and D), conformations in cluster 3405 were as
451 disperse as the full trajectory. Cluster 1171 of R2 showed conformations around the
452 maxima of PC1 to PC4 (Figs. 8E and F), according to its corresponding distributions (Figs.
453 9C and D). Identification of conformations at the maxima of the main PC may underpin
454 an important intermediary. In R3, distribution of PC1 displayed a prominent maximum at
455 -16 nm and a secondary maximum at -25 nm (Fig 9A); PC2 showed a broad distribution
456 with four maxima from -20 nm to 15 nm (Fig. 9B). Cluster 0733 of R3 showed a
457 conformation at maxima of all PC, albeit not in the principal maximum of PC3. In
458 summary, the most populated clusters whose conformations of the TM domain exhibited
459 a RMSD <0.5 Å, populated the PC distributions with conformations showing high
460 probabilities. By identifying less populated clusters, the intermediates in values at
461 shoulders or minimum of the PC distributions could be useful to establish possible
462 transition states.
463
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
464
465
466
467
468 Fig. 8. Principal component (PC) analysis in the FSHR. Projections on PC1-PC2 (A-C) y
469 PC3-PC4 (D-F) of trajectories R1-3 (blue points). Projections for groups 3405 in R12,
470 1171 in R2, and 0733 in R3 are included (black points). The groups conform subclusters
471 of conformational states whose RMSD difference in the TMD region is <0.5 Å.
472
473
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
474
475
476 Fig. 9 Distributions of the trajectory projections on the first four eigenvectors for the FSHR
477 in replicates R1 (orange), R2 (brown-orange), and R3 (brown). PC1 (A), PC2 (B), PC3
478 (C), and PC4 (D).
479
480 Correspondence between membrane and aqueous dynamics of the receptor’s
481 domains
482
483 Clusters conformed by the cut-off criterium (RMSD <0.5 Å) may be related to motions of
484 the extracellular domains, LRR, HR, and TM. Fig. 10 shows distances for the clusters
485 identified in replicates R1-3 of the LHCGR. Cluster 6383 was detected for R1 (Fig. 10,
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
486 row R1); however, the cut-off criterium produced too many clusters with few
487 conformations (Fig. 4A). To circumvent this problem, the cut-off can be adjusted to values
488 0.5 <RMSD< 1.0 in order to conform more populated clusters. In fact, for R2 the RMSD
489 cut-off criterium allowed to identify distinct clusters in terms of the relative LRR, HR and
490 TM distances (Fig. 10, row R2); hence, each cluster corresponds to a different state of
491 relative distances. In R3, there were also distinguishable clusters, albeit in this case there
492 were overlapping conformations most likely because clusters 2407 and 2964 were close
493 in time (Fig. 10, row R3). In addition, in R3 it was possible to detect the negative
494 correlation between RRL-RB vs TM-RB showed in Fig. 3B. From the analysis of relative
495 distances in clusters, it was possible to identify conformational states of the LHCGR in
496 MD trajectories for independent replicates, with starting configurations at different times
497 of a previous run. Using both, the RMSD and any of the relative distances, it is possible
498 to follow transitions among conformational states (e.g., active—inactive) with these
499 reaction coordinates.
500
501
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
502
503 Fig. 10. Distances among the LRR, HR and TMD in the LHCGR (Å) in R1 (upper panel),
504 R2 (middle panel) and R3 (lower panel). Analysis for clusters 6383 of R1, 1129, 2596,
505 and 4193 of R2, and 1404, 2407, and 2964 of R3.
506
507 Figure 11 shows the relative distances for the FSHR and, for this particular case, for only
508 the most populated cluster of each replicate R1-3. Conformational states can be identified
509 in terms of the relative distances: in particular, R2 showed almost no overlapping
510 conformations with R1 and R3. Because the trajectories were independent, the clusters
511 explored different regions of the conformational energy landscape. More comparisons
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
512 can be made among replicates; for example, it is possible to project a trajectory over the
513 eigenvectors of a second trajectory. In Fig. S3 of SI_info, projection of R1 over the first
514 eigenvector of R2 provides information on the conformations of R1 that contributes to
515 PC1 of R2; conversely, a lack of overlap among distributions would represent that
516 trajectories show different dynamics in the conformational space. By the strategy applied
517 in this study, we could distinguish conformations of the FSHR in which the HR moved
518 relative to LRR, as suggested by the previously proposed activation mechanism [7].
519
520
521
522 Fig. 11. Distances among LRR, HR and TM domains for the FSHR (Å). Blue dots, group
523 3406 of R1; magenta dots, group 1171 of R2; and black dots, group 0733 of R3.
524
525 DISCUSSION
526 In this study we explored the conformational changes of the gonadotropin receptors
527 (subfamily of GPHR), in which the FSHR is the prototypical member, within the family A
528 of GPCR. One of the key physiological functions of GPCRs is signal transduction, that is,
529 the triggering of a cell response to a particular (agonist) stimulus. Nevertheless, these
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
530 receptors are not necessarily a switch that turns on and off in response to a stimulus, but
531 rather exhibit diverse responses [18]. Conformational variability might explain that a
532 particular receptor may be coupled to distinct signaling molecules, which in turn is related
533 with different concepts such as allosteric regulation, signal predisposition or selective
534 signaling [35, 36]. Given that the structural information to determine the intermediate
535 states during the activation process of GPCRs still is scarce; in silico molecular dynamics
536 techniques are quite useful to establish their structure-function relationship in such
537 dissimilar environs, from the phospholipid membrane core to the bulk aqueous medium.
538 In particular, we employed a FSHR model that included the LRR, HR and TM as it is
539 reported in the AF2 server [37, 38]; the LHCGR structure was obtained from the PDB
540 repository (access code 7FII) [19]. GPHRs were analyzed in a comparative manner
541 following the same computational protocols. In both systems, the internal coordinates
542 were defined to measure distances among the LRR, HR and TM domains to detect the
543 relative motion of the HR, which is recognized as a key region involved in the activation
544 of these receptors [20, 33, 39].
545
546 The HR exhibits an -helix segment, a P10 segment and a loop which extends to the
547 aqueous medium that resembles the thumb of a glove (Fig. 1). The activation
548 mechanisms proposed based on the LHCGR structure in the active state, consists in the
549 displacement of the LRR in vertical position with respect to the plasma membrane (the
550 TM-LRR distance increases) while the HR approximates to the membrane (TM-HR
551 distance decreases) [19]. In the LHCGR, in the present generated trajectories exhibited
552 relative motion in which the LRR-TM fluctuates between 59.7 and 88.6 Å, the LRR-HR
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
553 between 31.4 and 70.0 Å, and the TM-HR between 48.2 and 79.2 Å. Motions of the TM-
554 HR and LRR-HR consistently exhibit negative correlations, where the increase in the
555 former tends to decrease the latter. In the crystal structure, the presence of the hormone
556 of the LHCGR may prevent the proximity between the LRR and the HR. These extreme
557 values may be useful to implement weighted ensemble simulations [34].
558
559 Although the role of the HR as an activation switch in GPCRs like the FSHR has been
560 recently challenged [39], it has been proposed that during FSHR activation a
561 displacement of the HR that allows insertion of residue Y355 in the interface between the
562 and subunits of FSH occurs [7, 40]. In the active state, the LRR is detected in vertical
563 position (perpendicular to the membrane) and the HR must move to the LRR [19, 41].
564 That is, the HR movement consists in an increase in LRR-TM distance and a
565 corresponding decrease in the LRR-HR one, leading to a negative correlation between
566 those distances. In the FSHR the relative motions of the LRR-TM would fluctuate between
567 71.4 and 93.3 Å, those of the LRR-HR between 46.7 and 81.3 Å, and those of the TM-
568 HR between 78.9 and 101.2 Å. Hence, trajectories using configurations with increasing
569 TM-LRR and/or decreasing LRR-HR distances, along with fluctuations of RMSD at the
570 TM domain, could be useful to identify transition intermediaries.
571
572 Another important element associated to the motion of the HR is the rearrangement of
573 the EC2, which has an inhibitory effect on the TM. In our simulations, this effect on the
574 TM helices was not observed because its initial configuration was already in the active
575 position, with the intracellular portions of the helices 5 and 6 displaced out. Nevertheless,
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
576 by means of a distance criterion (RMSD <0.5 Å), clusters showed well differentiated
577 conformations in the extracellular domains, in particular, the relative position of the HR.
578 In the present study, the extracellular and TM domains were initially studied uncoupled
579 due to the difference in the dynamics of the aqueous media and the membrane;
580 thereafter, it was possible to identify TM clusters with specific values of relative positions
581 of the LRR and the HR. Therefore, to explore the transitional states between the active
582 — inactive states, we propose that LRR-TM, or TM-HR relative distances may be useful
583 as reaction coordinates.
584
585 The purpose of exploring the energy landscape of the gonadotropin receptors is, among
586 others, to screen for structural variations that may respond to conformational changes
587 leading to well-known states, or alternative states; for example, those promoting binding
588 of new drugs, or selective coupling to intracellular transducers. A broad perspective could
589 include using the reactions coordinates, here explored, to bias the conformational
590 dynamics of the receptors upon binding allosteric modulators or agonists able to favor
591 particular signaling pathways, as required by a given therapeutic purpose.
592
593 MATERIAL AND METHODS
594
595 Two simulation boxes containing the receptor, a lipid bilayer, water molecules as solvent,
596 and Na+ or Cl- ions for charge balance were set up. For the FSHR initial coordinates, we
597 used the AF2 [37, 38] structure, generated with the full sequence of the human receptor
598 model (AF-P23945). Post-translational modifications were introduced at C644 and C646
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
599 by palmitoylation through thioester bonds [21]. Disulfide bonds between the cysteine side
600 chains of C18-C25, C23-C32, C275-C346, C276-C356, C442-C517, and C229-C338
601 were defined according to the S-S bond distance criterium. Protonation states of side
602 chains were set to those of the predominant species at neutral pH. The FSHR structure
603 was inserted in a preequilbrated bilayer of 1-stearoyl-2-docosahexaenoyl-sn-glycero-3-
604 phosphocholine (SDPC). Water molecules for solvation of the receptor and the lipid heads
605 were added in a rectangular box of 120X120X160 Å3, respectively, in the x, y and z
606 directions. A total of 191601 atoms were included: 48411 water molecules, 281 lipids, 3
607 Na+ ions and the FSHR with 695 residues. The second simulation box contained 190072
608 atoms: 48414 water molecules, 247 SDPC lipids, 3 Na+ ions and 7 Cl- ions, and the
609 LHCGR with 613 residues. The LHCGR structure (PDB:7FII) corresponds to the state
610 bound to the hormone and the Gs-protein [19], and it was processed according to the
611 CHARMM-GUI membrane builder [42, 43] using the following parameters [44]. The chain
612 R (segment PROD) of the pdb file was selected to be inserted in a POPC lipid bilayer,
613 with box dimensions of 100x100x168 Å3. For the missing residues T287 to W329 of the
614 HR, the CHARMM-GUI modeling scheme included the coordinates as predicted by the
615 GalaxyFill algorithm [45]. Disulfide bonds were defined between cysteines C279-C343,
616 C280-C353, C131-C156 and C439-C514. The LHCGR principal axis was aligned in the
617 z direction and displaced until the TM domains matched the hydrophobic core of the
618 bilayer. The system size was 191601 atoms: 248 1-palmitoyl-2-oleoyl-sn-glycero-3-
619 phosphocholine (POPC) lipids, 37479 water molecules, Na+Cl- ions (0.15 M), and the
620 LHCGR. From the assembly of the initial configuration, the systems were energy
621 minimized with conjugated gradient algorithm for 10 k steps, followed by a gradual
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
622 relaxation of the receptor atoms. In a first stage, the receptor backbone atoms were fixed,
623 then subsequent stages with positional constrains of 20, 15, 10, 5, 3, 2, and 1 kcal /mol
624 Å2, in short trajectories of 200 ps each were applied. The trajectory of LHCGR in POCP
625 was prolonged for 200 ns without constraints. Because of our interest in exploring the
626 receptor’s conformational landscapes, we generated independent trajectory replicates for
627 both the FSHR and the LHCGR in SDPC. In the case of the FSHR, three replicates were
628 generated using the last configuration after the relaxation procedure and restarted
629 velocities for each replicate. For the LHCGR replicates, we used configurations taken
630 from the trajectory in POCP at times 0 ns, 100 ns, and 180 ns, and relaxed the receptor
631 in the membrane environment of SDPC lipids.
632
633 All simulations were performed with the NAMD 2.14 software [46], version NAMD3.0
634 alpha, which was optimized for GPU-accelerated servers [47]. Simulation trajectories
635 were generated in the isothermal-isobaric ensemble (NPT) with Langevin dynamics to
636 maintain a constant temperature [46], and Nosé-Hoover Langevin piston to maintain a
637 constant pressure of 1 bar [48]. Anisotropic cell fluctuations in the x-, y- and z-axis were
638 allowed [49]. Non-bonding interactions were calculated with a cutoff of 11 Å, and a shifting
639 function starting at 10.0 Å. A multiple time step integration for solving the motion equations
640 was used with one step for bonding interaction and short-range nonbonding interactions,
641 and two steps for electrostatic forces, with 2 fs time step. All hydrogen atoms were fixed
642 using the SHAKE and RATLLE algorithms [50, 51]. Electrostatic interactions were
643 evaluated using PME [52], with a 4th order interpolation on a grid of ~1 Å in the x-, y- and
644 z-directions, and a tolerance of 10-6 for the direct evaluation of the real part of the Ewald
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
645 sum. CHARMM36 all-atom force field parameters were used for the lipid molecules [53],
646 and CHARMM36m for the protein atoms [54, 55], including CMAP correction [56, 57].
647 Water molecules were modeled using the TIP3P potential [58].
648
649 Trajectory analysis were performed in the TCL environment of VMD [59] for the
650 calculation of the root mean square deviation (RMSD), root mean square fluctuations
651 (RMSF), LRR-HR-TMD distances, among other structural and dynamical parameters. For
652 the principal component analysis, GROMACS 2020 [60] was employed with commands
653 gmx covar and anaeig for the calculation of covariance matrix and analysis of
654 eigenvectors, respectively,. For the first four eigenvectors we calculated the principal
655 components of the C atoms at the TM helices of the FSHR: TM1, Y362 to Y392; TM2,
656 P397 to Y432; TM3, N437 to T472; TM4, A487 to G507; TM5, M532 to R557; TM6, D567
657 to L597; and TM7, A607 to Y626; and the LHCGR: TM1, 360-387; TM2 392-420; TM3,
658 T437-470; TM4, 480-504; TM5, 522-552; TM6, 565-597; and TM7, 602-625. RMSD
659 matrices for comparison of frames at t and t+∆t were calculated for TM helices.
660 Correlations among domain distances also were calculated to detect concerted motions.
661 Visual molecular dynamics (VMD) was used for visualization, representation of 3D
662 structures and image generation [59].
663
664 ACKNOWLEDGMENTS
665 This study was supported by grant IN208323 from the Programa de Apoyo a Proyectos
666 de Investigación e Innovación Tecnológica (PAPIIT), UNAM, Mexico (to A.U.-A).
667
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
668 REFERENCES
669 1. Fredriksson R, Langerström MC, Lundin L-G, Schiöth HB. The G-Protein-
670 Coupled Receptors in the Human Genome Form Five Main Families. Phylogenetic
671 Analysis, Paralogon Groups, and Fingerprints. Mol Pharmacol. 2003;63:1256–72. doi:
672 10.1124/mol.63.6.1256a.
673 2. Gershengorn MC, Osman R. Minireview: Insights into G protein-coupled receptor
674 function using molecular models. Endocrinology. 2001;142(1):2-10. doi:
675 10.1210/endo.142.1.7919.
676 3. Insel PA, Sriram K, Gorr MW, Wiley SZ, Michkov A, Salmerón C, et al.
677 GPCRomics: an approach to discover GPCR drug targets. Trends in pharmacological
678 sciences. 2019;40(6):378-87. doi: 10.1016/j.tips.2019.04.001.
679 4. Hauser AS, Attwood MM, Rask-Andersen M, Schiöth HB, Gloriam DE. Trends in
680 GPCR drug discovery: new agents, targets and indications. Nature reviews Drug
681 discovery. 2017;16(12):829-42. doi: 10.1038/nrd.2017.178.
682 5. Arey BJ, Dias JA. the Physiology and Pharmacology of leucine-rich repeat
683 GPCrs. Frontiers in Endocrinology. 2016;7:56. doi: 10.3389/fendo.2016.00056.
684 6. Miller-Gallacher J, Sanders P, Young S, Sullivan A, Baker S, Reddington SC, et
685 al. Crystal structure of a ligand-free stable TSH receptor leucine-rich repeat domain.
686 Journal of Molecular Endocrinology. 2019;62(3):117-28. doi: 10.1530/JME-18-0213.
687 7. Ulloa-Aguirre A, Zariñán T. The follitropin receptor: matching structure and
688 function. Molecular pharmacology. 2016;90(5):596-608. doi: 10.1124/mol.116.104398.
689 8. Ascoli M, Fanelli F, Segaloff DL. The lutropin/choriogonadotropin receptor, a
690 2002 perspective. Endocrine reviews. 2002;23(2):141-74. doi: 10.1210/edrv.23.2.0462.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
691 9. Bogerd J. Ligand-selective determinants in gonadotropin receptors. Molecular
692 and cellular endocrinology. 2007;260:144-52. doi: 10.1016/j.mce.2006.01.019.
693 10. Mueller S, Jaeschke H, Gunther R, Paschke R. The hinge region: an important
694 receptor component for GPHR function. Trends Endocrinol Metab. 2010;21(2):111-122.
695 doi: 10.1016/j.tem.2009.09.001.
696 11. Casarini L, Huhtaniemi IT, Simoni M, Rivero-Müller A. Gonadotropin receptors.
697 Endocrinology of the Testis and Male Reproduction Endocrinology Springer, Cham.
698 2017:123-68. doi: 10.1007/978-3-319-44441-3_4.
699 12. Ulloa-Aguirre A, Zariñán T, Jardón-Valadez E, Gutiérrez-Sagal R, Dias JA.
700 Structure-Function Relationships of the Follicle-Stimulating Hormone Receptor.
701 Frontiers in Endocrinology. 2018;9:707. doi: 10.3389/fendo.2018.00707.
702 13. Althumairy D, Zhang X, Baez N, Barisas G, Roess DA, Bousfield GR, et al.
703 Glycoprotein G-protein coupled receptors in disease: luteinizing hormone receptors and
704 follicle stimulating hormone receptors. Diseases. 2020;8(3):35. doi:
705 10.3390/diseases8030035.
706 14. Zariñán T, Mayorga J, Jardón-Valadez E, Gutiérrez-Sagal R, Maravillas-Montero
707 JL, Mejía-Domínguez NR, et al. A novel mutation in the FSH receptor (I423T) affecting
708 receptor activation and leading to primary ovarian failure. The Journal of Clinical
709 Endocrinology & Metabolism. 2021;106(2):e534-e50. doi: 10.1210/clinem/dgaa782.
710 15. Jardón-Valadez E, Castillo-Guajardo D, Martínez-Luis I, Gutiérrez-Sagal Rn,
711 Zariñán T, Ulloa-Aguirre A. Molecular dynamics simulation of the follicle-stimulating
712 hormone receptor. Understanding the conformational dynamics of receptor variants at
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
713 positions N680 and D408 from in silico analysis. Plos One. 2018;13(11):e0207526. doi:
714 10.1371/journal.pone.0207526.
715 16. Hauser AS, Kooistra AJ, Munk C, Heydenreich FM, Veprintsev DB, Bouvier M, et
716 al. GPCR activation mechanisms across classes and macro/microscales. Nature
717 structural & molecular biology. 2021;28(11):879-88. doi: 10.1038/s41594-021-00674-7.
718 17. Latorraca NR, Venkatakrishnan AJ, Dror RO. GPCR dynamics: structures in
719 motion. Chemical reviews. 2017;117(1):139-55. doi: 10.1021/acs.chemrev.6b00177.
720 18. Smith JS, Lefkowitz RJ, Rajagopal S. Biased signalling: from simple switches to
721 allosteric microprocessors. Nature reviews Drug discovery. 2018;17(4):243-60. doi:
722 10.1038/nrd.2017.229.
723 19. Duan J, Xu P, Cheng X, Mao C, Croll T, He X, et al. Structures of full-length
724 glycoprotein hormone receptor signalling complexes. Nature. 2021;598(7882):688-92.
725 doi: 10.1038/s41586-021-03924-2.
726 20. He X, Duan J, Ji Y, Zhao L, Jiang H, Jiang Y, et al. Hinge region mediates signal
727 transmission of luteinizing hormone and chorionic gonadotropin receptor. Computational
728 and Structural Biotechnology Journal. 2022;20:6503-11. doi:
729 10.1016/j.csbj.2022.11.039.
730 21. Melo-Nava B, Casas-González P, Pérez-Solís MA, Castillo-Badillo J, Maravillas-
731 Montero JL, Jardón-Valadez E, et al. Role of Cysteine Residues in the Carboxyl-
732 Terminus of the Follicle Stimulating Hormone Receptor in Intracellular Traffic and
733 Postendocytic Processing. Frontiers in Cell and Developmental Biology. 2016;4:1-13.
734 doi: 10.3389/fcell.2016.00076.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
735 22. Wales DJ, Bogdan TV. Potential energy and free energy landscapes. The
736 Journal of Physical Chemistry B. 2006;110(42):20765-76. doi: 10.1021/jp0680544.
737 23. Landomiel F, De Pascali F, Raynaud P, Jean-Alphonse F, Yvinec R, Pellissier
738 LP, et al. Biased signaling and allosteric modulation at the FSHR. Frontiers in
739 Endocrinology. 2019;10:148. doi: 10.1016/j.csbj.2022.11.039.
740 24. Ribeiro JML, Filizola M. Allostery in G protein-coupled receptors investigated by
741 molecular dynamics simulations. Current opinion in structural biology. 2019;55:121-8.
742 doi: 10.1016/j.sbi.2019.03.016.
743 25. Leander M, Liu Z, Cui Q, Raman S. Deep mutational scanning and machine
744 learning reveal structural and molecular rules governing allosteric hotspots in
745 homologous proteins. Elife. 2022;11:e79932. doi: 10.7554/eLife.79932.
746 26. Thirumalai D, Hyeon C, Zhuravlev PI, Lorimer GH. Symmetry, rigidity, and
747 allosteric signaling: from monomeric proteins to molecular machines. Chemical reviews.
748 2019;119(12):6788-821. doi: 10.1021/acs.chemrev.8b00760.
749 27. Hess B. Convergence of samplig in protein simulations. Phys Rev E.
750 2001;65:031910-10. doi: 10.1103/PhysRevE.65.031910.
751 28. Hayward S, De Groot BL. Normal modes and essential dynamics. Molecular
752 Modeling of Proteins: Springer; 2008. p. 89-106.
753 29. Lange OF, Grubmüller H. Can principal components yield a dimension reduced
754 description of protein dynamics on long time scales? The Journal of Physical Chemistry
755 B. 2006;110(45):22842-52. doi: 10.1021/jp062548j.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
756 30. Sittel F, Stock G. Perspective: Identification of collective variables and
757 metastable states of protein dynamics. The Journal of chemical physics.
758 2018;149(15):150901. doi: 10.1063/1.5049637.
759 31. Mori T, Miyashita N, Im W, Feig M, Sugita Y. Molecular dynamics simulations of
760 biological membranes and membrane proteins using enhanced conformational
761 sampling algorithms. Biochimica et Biophysica Acta (BBA)-Biomembranes.
762 2016;1858(7):1635-51. doi: 10.1016/j.bbamem.2015.12.032.
763 32. Smith SO. Mechanism of Activation of the Visual Receptor Rhodopsin. Annual
764 Review of Biophysics. 2023;52:301-17. doi: 10.1146/annurev-biophys-083122-094909.
765 33. Agrawal G, Dighe RR. Critical involvement of the hinge region of the follicle-
766 stimulating hormone receptor in the activation of the receptor. Journal of Biological
767 Chemistry. 2009;284(5):2636-47. doi: 10.1074/jbc.M808199200.
768 34. Zuckerman DM, Chong LT. Weighted ensemble simulation: review of
769 methodology, applications, and software. Annual review of biophysics. 2017;46:43-57.
770 doi: 10.1146/annurev-biophys-070816-033834.
771 35. Hilger D, Masureel M, Kobilka BK. Structure and dynamics of GPCR signaling
772 complexes. Nature structural & molecular biology. 2018;25(1):4-12. doi:
773 10.1038/s41594-017-0011-7.
774 36. Wootten D, Christopoulos A, Sexton PM. Emerging paradigms in GPCR
775 allostery: implications for drug discovery. Nature reviews Drug discovery.
776 2013;12(8):630-44. doi: 10.1038/nrd4052.
777 37. Varadi M, Anyango S, Deshpande M, Nair S, Natassia C, Yordanova G, et al.
778 AlphaFold Protein Structure Database: massively expanding the structural coverage of
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
779 protein-sequence space with high-accuracy models. Nucleic Acids Research.
780 2022;50(D1):D439-D44. doi: 10.1093/nar/gkab1061.
781 38. Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, et al. Highly
782 accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583-9. doi:
783 10.1038/s41586-021-03819-2.
784 39. Nagel M, Moretti R, Paschke R, von Bergen M, Meiler J, Kalkhof S. Integrative
785 model of the FSH receptor reveals the structural role of the flexible hinge region.
786 Structure. 2022;30(10):1424-31. doi: 10.1016/j.str.2022.07.007.
787 40. Jiang X, Liu H, Chen X, Chen P-H, Fischer D, Sriraman V, et al. Structure of
788 follicle-stimulating hormone in complex with the entire ectodomain of its receptor. Proc
789 Nat Acad Sci USA. 2012;109:12491–6. doi: 10.1073/pnas.1206643109.
790 41. Jiang X, Dias JA, He X. Structural biology of glycoprotein hormones and their
791 receptors: Insights to signaling. Molecular and Cellular Endocrinology. 2014;382:424–
792 51. doi: 10.1016/j.mce.2013.08.021.
793 42. Jo S, Lim JB, Klauda JB, Im W. CHARMM-GUI membrane builder for mixed
794 bilayers and its application to yeast membranes. Biophysical Journal. 2009;97:50-8. doi:
795 10.1016/j.bpj.2009.04.013.
796 43. Jo S, Kim T, Iver VG, Im W. CHARMM-GUI: A web-based graphical user
797 interface for CHARMM. J Comput Chem. 2008;29:1859-65. doi: 10.1002/jcc.20945.
798 44. Li Y, Liu J, Gumbart JC. Preparing membrane proteins for simulation using
799 CHARMM-GUI. Structure and Function of Membrane Proteins: Springer; 2021. p. 237-
800 51.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
801 45. Coutsias EA, Seok C, Jacobson MP, Dill KA. A kinematic view of loop closure.
802 Journal of computational chemistry. 2004;25(4):510-28. doi: 10.1002/jcc.10416.
803 46. Phillips JC, Braun B, Wang W, Gumbart J, Tajkhorshid E, Villa E, et al. Scalable
804 molecular dynamics with NAMD. Journal of Computational Chemistry. 2005;26:1781-
805 802. doi: 10.1002/jcc.20289.
806 47. Phillips JC, Hardy DJ, Maia JDC, Stone JE, Ribeiro JV, Bernardi RC, et al.
807 Scalable molecular dynamics on CPU and GPU architectures with NAMD. The Journal
808 of chemical physics. 2020;153(4):044130. doi: 10.1063/5.0014475.
809 48. Martyna GJ, Tobias DJ, Klein ML. Constant-pressure molecular-dynamics
810 algorithms. JChemPhys. 1994;101:4177-89. doi: 10.1063/1.467468. PubMed PMID:
811 19004.
812 49. Yu T-Q, Alejandre J, López-Rendón R, Martyna GJ, Tuckerman ME. Measure
813 preserving integrators for molecular dynamics in the isothermal-isobaric ensemble
814 derived from the Lioville operator. Chem Phys. 2010;370:294-305. doi:
815 10.1016/j.chemphys.2010.02.014.
816 50. Miyamoto S, Kollman P. An analytical version of the SHAKE and RATTLE
817 algorithm for rigid water models. Journal of Computational Chemistry. 1992;13:952-62.
818 doi: 10.1002/jcc.540130805.
819 51. Andersen HC. Rattle: A Velocity Molecular Version Dynamics of the Shake
820 Calculations. Journal of Computational Physics. 1982;52:24-34. doi: 10.1016/0021-
821 9991(83)90014-1.
822 52. Essmann U, Perera L, Berkowitz ML, Darden T, Lee H, Pedersen LG. A smooth
823 particle mesh Ewald method. JChemPhys. 1995;103:8577-93. doi: 10.1063/1.470117.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
824 53. Lee S, Tran A, Allsopp M, Lim JB, Hénin J, Klauda JK. CHARMM36 United Atom
825 Chain Model for Lipids and Surfactants. The Journal of Physical Chemitry B.
826 2014;118:547-56. doi: 10.1021/jp410344g.
827 54. MacKerell AD, Jr., Bashford D, Bellott M, Dunbrack RL, Jr., Evanseck JD, Field
828 MJ, et al. All-atom empirical potential for molecular modeling and dynamics studies of
829 proteins. JPhysChemB. 1998;102(18):3586-616. doi: 10.1021/jp973084f
830 55. MacKerell AD, Jr., Feig M, Brooks CL, II. Extending the treatment of backbone
831 energetics in protein force fields: Limitations of gas-phase quantum mechanics in
832 reproducing conformational distributions in molecular dynamics simulations.
833 JComputChem. 2004;25:1400-15. doi: 10.1002/jcc.20065.
834 56. Best RB, Zhu X, Shim J, Lopes PEM, Mittal J, Feig M, et al. Optimization of the
835 additive CHARMM all-atom protein force field targeting improved sampling of the
836 backbone phi, psi and side-chain chi1 and chi2 dihedral angles. Journal of Chemical
837 Theory and Computation. 2012;8:3257-73. doi: 10.1021/ct300400x
838 57. MacKerell AD, Feig M, Brooks CL. Improved Treatment of the Protein Backbone
839 in Empirical Force Fields. Journal of the American Chemical Society. 2004;126(3):698-
840 9. doi: 10.1021/ja036959e.
841 58. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Comparison
842 of simple potential functions for simulating liquid water. JChemPhys. 1983;79(2):926-35.
843 doi: 10.1063/1.445869.
844 59. Humphrey W, Dalke W, Schulten K. VMD: Visual molecular dynamics. J. Mol.
845 Graphics. 1996;14(1):33-8. doi: 10.1016/0263-7855(96)00018-5.
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
846 60. Páll S, Zhmurov A, Bauer P, Abraham M, Lundborg M, Gray A, et al.
847 Heterogeneous parallelization and acceleration of molecular dynamics simulations in
848 GROMACS. The Journal of Chemical Physics. 2020;153(13):134110. doi:
849 10.1063/5.0018516.
850
851
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted August 12, 2023. ; https://doi.org/10.1101/2023.08.09.552573doi: bioRxiv preprint
