Identification of GDF15 peptide fragments inhibiting GFRAL receptor signaling

Growth differentiation factor 15 (GDF15) is believed to be a major causative factor for cancer-induced cachexia. Recent elucidation of the central circuits involved in GDF15 function and its signaling through the glial cell-derived neurotrophic factor family receptor α -like (GFRAL) has prompted the interest of targeting the GDF15-GFRAL signaling for energy homeostasis and body weight regulation. Here, we applied advanced peptide technologies to identify GDF15 peptide fragments inhibiting GFRAL signaling. SPOT peptide arrays revealed binding of GDF15 C -terminal peptide fragments to the extracellular domain of GFRAL. Parallel solid-phase peptide synthesis allowed for generation of complementary GDF15 peptide libraries and their subsequent functional evaluation in cells expressing the GFRAL/RET receptor complex. We identified a series of C -terminal fragments of GDF15 inhibiting GFRAL activity in the micromolar range. These novel GFRAL peptide inhibitors could serve as valuable tools for further development of peptide therapeutics towards the treatment of cachexia and other wasting disorders.


Introduction
Growth differentiation factor 15 (GDF15) was first described in 1997 [1][2][3] as a 25 kDa dimeric secreted hormone bearing common three-dimensional structural characteristics with the transforming growth factor β (TGF-β) superfamily. GDF15 is expressed in various tissues such as liver, lung, kidney, and placenta [4], and elevated serum levels of GDF15 are associated with a variety of physiological conditions including age, pregnancy, and exercise [5][6][7]. Additionally, serum GDF15 positively correlates with inflammation [8] in chronic diseases such as cancer [9], cardiovascular disease [10], chronic kidney [11] and liver diseases [12], as well as in viral infections [13]. Accumulating experimental evidence has linked GDF15 with energy intake regulation and weight maintenance while increased circulating GDF15 levels in obesity have been suggested as a compensatory molecular mechanism to reduce energy intake [14].
Notably, GDF15 has been proposed as a causative factor for development of cancer cachexia [15]. Cachexia describes a metabolic wasting syndrome characterized by unintended weight loss accompanied by impaired regulation of energy homeostasis and progressive depletion of skeletal muscle [16,17]. Impaired energy homeostasis is triggered by both the tumor and its treatment (e.g. radiation, chemotherapy) and involves combinations of metabolic abnormalities in various signaling pathways [18,19]. The multifactorial metabolic aetiology of cachexia hampers the efficacy of current treatments based on nutritional supplements and appetite stimulants, leaving a large and unmet medical need for more efficacious and targeted therapies [20,21].
Until recently, the molecular mechanism through which GDF15 promotes body weight loss has remained elusive. In 2017, four research groups independently identified the GDNF family receptor α-like (GFRAL) as the principal receptor mediating the metabolic effects of GDF15 [22][23][24][25]. Consistent with the inhibitory effects of GDF15 on appetite function, GFRAL expression was found to be highly restricted to the brainstem, specifically in neurons of the area postrema (AP) and the nucleus of the solitary tract (NTS), brain regions highly associated with central control of energy homeostasis [23,25]. GDF15 binds specifically and with high affinity to GFRAL, while recruitment of co-receptor RET, a receptor tyrosine kinase, and formation of the GDF15/GFRAL/RET ternary complex is required for intracellular signaling. Conversely, metalloproteinase-mediated inactivation of GFRAL has recently been reported to negatively regulate GDF15-GFRAL signaling [26]. Interestingly, GFRAL knock-out mice are resistant to chemotherapy-induced weight loss, supporting a role of GFRAL signaling in cancer cachexia [22]. Moreover, GDF15-induced appetite suppression in rats was prevented by co-administration of an anti-GFRAL antibody [25]. Collectively, these findings strengthen the concept of inhibiting GDF15-GFRAL activity for the treatment of cachexia.
To date, the focus of targeting GFRAL function has been largely related to the treatment of obesity. Accordingly, several pharmaceutical companies have developed clinical relevant GDF15-based peptide therapeutics for obesity management including recombinantly expressed long-acting GDF15 analogues [27,28]. Despite that currently no FDA-approved medications for the indication of cancer cachexia are available, to our knowledge, limited efforts have been made to identify modalities for inhibiting GDF15 function in cachexia. The clinical relevance of targeting the GDF15-GFRAL axis for cachexia is supported by several preclinical studies. For example, administration of monoclonal antibodies neutralizing GDF15-induced signaling have been demonstrated to reverse the cachectic phenotype in transgenic mice overexpressing GDF15 [15] and prevent weight loss in tumor-bearing mice [29][30][31]. Here, we report the identification of GDF15 peptide fragments inhibiting GFRAL signaling and providing templates for the development of novel peptide-based therapeutics of cachexia targeting GDF15/GFRAL/RET signaling.
Dried cellulose membrane discs were transferred to Eppendorf tubes and treated with side chain deprotection solution (80 % TFA, 12 % DCM, 5 % H 2 O, and 3 % TIPS) for 1.5 h at RT. Deprotection solution was then removed and a solvation mixture (88.5 % TFA, 4 % trifluoromethanesulfonic acid (TFMSA), 5 % H 2 O, and 2.5 % TIPS) was added for overnight solubilization of the discs. Peptide-cellulose conjugates were then precipitated with cold diethyl ether and pelleted at 3000 rpm for 15 min, followed by an additional wash with cold ether. DMSO stocks of the conjugates were prepared and transferred to a 384well plate for printing (in duplicates) on white coated CelluSpots blank slides (76 × 26 mm, Intavis AG) using a SlideSpotter robot (Intavis AG, Tübingen, Germany).
Printed slides were washed (2 ×) with PBS and incubated with 1 % bovine serum albumin (BSA) in PBS for 4 h. Slides were then washed with 1 % BSA in PBS (5 ×) and incubated with 0.1 µM GFRAL His-tagged extracellular domain (ECD) (Catalog# 9647-GR-050, R&D systems) in 1 % BSA in PBS for 1 h. Washing (4x) with 1 % BSA in PBS was followed by incubation with HRP anti-6x His-tag Ab (1:5000 dilution, Catalog# ab184607, Abcham) for 30 min. Lastly, slides were washed (4x) with 1 % BSA in PBS, (2 ×) with PBS and the detection substrate was (Super-Signal™ West Femto Maximum Sensitivity Substrate, Catalog# 34095, Thermo Scientific) added to the slides. Slides were visualized with a Syngene PXi image recorder (exposure time = 5 s). The resulting blots were analyzed using the Array Analyze Software (Active Motif), which defines the error range of each data set by comparing the intensities of each peptide duplicate on the analyzed array.

Peptide synthesis
Reagents for solid-phase peptide synthesis (SPPS) were purchased from Iris Biotech GmbH (Marktredwitz, Germany). Milli-Q water (Merck Millipore) was used for all experiments. Peptides were synthesized using fully automated Syro-II peptide synthesizer (MultiSynTech GmbH, Witten, Germany) by SPPS according to the 9-fluorenylmethyloxycarbonyl (Fmoc) strategy. Peptide synthesis was conducted on 0.008-0.2 mmol scale using TentaGel S RAM resin (0.24 mmol/g) as solid support. Fmoc-protected amino acids (4 eq.) were coupled using DIC (4 eq.) and ethyl Oxyma (4 eq.) in DMF, except Fmoc-Phe-OH, which was dissolved in NMP. All couplings were performed at 75 • C for 10 min, except His and Cys which were performed at 50 • C for 15 min, either as single or as double couplings. Fmoc deprotection was performed using 20 % piperidine in DMF and 0.1 M HOBt was added to avoid aspartimide formation. Release of peptide from the solid support and simultaneously removal of the acid-labile side chain protecting groups was performed by incubation with a trifluoroacetic acid (TFA):triethylsilane:H 2 O (95:2.5:2.5) mixture for 3 h at RT. The peptides were precipitated using cold diethyl ether.

GDF15 peptide fragments bind to the GFRAL extracellular domain
The crystal structure of GDF15 in complex with the GFRAL extracellular domain (ECD) revealed direct interactions of the N-and C-terminal hairpins of GDF15 with the D2 domain of GFRAL [22]. This finding prompted us to explore whether peptide fragments of GDF15 could mimic binding to GFRAL ECD. For this purpose, we employed SPOT peptide arrays for the parallel synthesis of GDF15 peptide fragments on functionalized cellulose membranes. SPOT arrays displaying GDF15 fragments were subsequently screened for binding to the GFRAL ECD (Fig. 1A).
First, the GDF15 sequence (GDF15 201-308 ) was screened by synthesizing on SPOT 15-mer GDF15 peptides (SPOT array 1). The array was designed by fragmenting GDF15 starting from the N-terminus and moving towards the C-terminus with shifting of 1 amino acid at a time (Table S1). Screening of SPOT array 1 against the GFRAL ECD revealed that the two strongest peptide binders (88 287 KTDTGVSLQTYDDLL 301 and 89 288 TDTGVSLQTYDDLLA 302 ) derived from the C-terminal region of GDF15 (Fig. 1B and Fig. S1A). From the X-ray crystal structure of GDF15 in complex with GFRAL ECD, it was observed that the binding sequence of 88 and 89 (residues 288-302) was part of the binding interface formed by the C-terminal hairpin of GDF15 and the D2 domain of GFRAL.
Next, we focused on the C-terminal region of GDF15, and a second array (SPOT array 2) was designed to enable a more thorough screening of the C-hairpin sequence (residues 273-308). Longer peptide fragments (16-23-mers) were synthesized, and C-hairpin was screened by shifting of 1 amino acid between peptides while moving towards the C-terminus (Table S3). Subsequent array screening showed various peptide binders to the GFRAL ECD (Fig. 1C, Fig. S1B). Interestingly, all binders shared overlapping sequences covering the complementary β-strands of the GDF15 C-hairpin (residues 279-301, Fig. 1D).
Following, an in vitro cell based functional assay (ERK phosphorylation) was established for the evaluation of the functional properties of the identified GDF15 peptide binders on GFRAL signaling. The two top  Table 1).

C-terminal GDF15 fragments inhibit GFRAL
After identifying GFRAL peptide binders using SPOT arrays, we decided to screen a complementary library of GDF15 fragments for their effect on GFRAL signaling. This allowed the synthesis of peptide libraries using solid-phase peptide synthesis (SPPS) and provided the additional advantage of investigating longer GDF15 peptide fragments than the ones displayed on SPOT arrays. A library of 192 peptides was designed where the GDF15 sequence (GDF15 201-308 ) was screened with peptides of various lengths (34, 32, 30, 28 and 26 amino acids), with shifting of 2 amino acids between peptides while moving from the N-to C-terminus (Table S5). Additionally, oxidation-prone methionine residues were substituted with norleucine (Nle) to optimize synthetic efficiency.
Peptide synthesis was followed by quality control analysis where 7-8 representative peptides of each length were selected for resolving their purity and concentration using liquid chromatography-mass spectrometry (LC-MS) and charged aerosol detector (CAD), respectively. An average peptide purity and concentration was calculated, based on which peptides were aliquoted for functional screening (Table S6). Next, all 192 peptides were analysed by LC-MS and peptides for which their mass was not detected (71 out of 192) were excluded from analysis (Table S5).
Functional screening was performed on GFRAL/RET cells, in a single-point determination (C peptide = 50 µM) in the presence of a fixed concentration of GDF15 (EC 85 = 0.3 nM) ( Fig. 2A). This screening procedure indicated that a subset of peptides deriving from the C-terminal region of GDF15 (271− 304) consistently inhibited GFRAL signaling (Fig. 2B). A series of fragments across the GDF15 sequence were then selected for individual LC-MS and CAD analysis and subsequently characterized in a 5-point crude peptide concentration-response assay. In agreement with the initial library screening, it was observed that 30-34 amino acid C-hairpin peptide fragments 311, 348 and 386 sharing the sequence 271-304 could inhibit GFRAL with a potency (EC 50 ) of up to 32 µM (Table S7).

In vitro validation of GFRAL inhibitors
Since screening was performed with crude peptides, validation of the findings was required. Three hit peptides of different lengths ( 271 APCCVPASYNPNleLIQKTDTGVSLQTYDDLLAKD 304 311, 269 CCVPA SYNPNleVLIQKTDTGVSLQTYDDLLAKD 304 348, 267 VPASYNPNleVLIQ KTDTGVSLQTYDDLLAKD 304 387) and a peptide fragment showing no inhibition ( 241 IGACPSQFRAANNleHAQIKTSLHRLKPDTVPAP 272 333) were selected for synthesis, purification, and in vitro validation of their functional properties (Table 1, Table S8).
Similar to the potency data obtained with crude peptides (Table S7), concentration-response curves of 311 and 348 showed that these were the most potent GFRAL inhibitors (EC 50 = 25.2 ± 3.7 and 19.1 ± 6.6 µM respectively, Fig. 2D), while 333 did not show receptor inhibition (Table 1). Additionally, the shortest analogue 387 with an N-terminal deletion of two cysteine residues, Cys273 and Cys274, resulted in loss of receptor potency (EC 50 >100 µM, Table 1). Finally, we evaluated whether increased potency of 311 and 348 could be attributed to peptide dimerization via disulphide bridge formation under the assay conditions. A serine substituted analogue of 348 ( 269 SSVPAS YNPNleVLIQKTDTGVSLQTYDDLLAKD 304 ) 469 showed similar antagonistic profile (EC 50 = 25.1 ± 15.7, Table 1) to 348 indicating that the GDF15 peptide fragment 348 is the most potent inhibitor of the GFRAL. Last, 348, its shorter analogue 195 and the inactive peptide 333 were evaluated for receptor agonism showing no GFRAL activation (Fig. S2).

Discussion
GDF15 is well documented for playing a critical role in driving uncontrollable and severe loss of body weight in patients with cancer [32,33]. Here, we report the application of peptide screening platforms for the discovery of novel GDF15 peptide fragments that inhibit GFRAL signaling, thus providing valuable tools for the in vivo elucidation of GFRAL inhibitory effects as well as the development of peptide therapeutics towards the treatment of cancer cachexia.
The binding properties of GDF15 peptide fragments to the GFRAL ECD were investigated utilizing SPOT peptide arrays. The initial screening of the GDF15 sequence revealed binding of two 15-mer peptides (88 and 89, residues 287-302) derived from the C-terminal hairpin (residues 274-308) of GDF15. Interestingly, GFRAL binders included residues (e.g. Thr290, Val292) reported to mediate intermolecular contacts (≤ 4.5 Å) between GDF15 C-hairpin and the D2 domain of GFRAL [22]. Subsequent screening of 16-23-mer fragments demonstrated consistent C-terminal binders sharing an extended overlapping sequence (residues 277-302). Collectively, these data suggested that C-terminal fragments covering residues of the complementary β-strands of the GDF15 C-hairpin could bind with increasing affinity to the GFRAL ECD. Functional evaluation of two selected binders (195 and 253) in cells co-expressing GFRAL/RET showed weak inhibition of GDF15-induced signaling.
SPOT peptide arrays offer a powerful method for the identification of binding epitopes and targeting of protein-protein interactions [34], though the synthetic efficiency is highly dependent on the peptide sequence, length and potential secondary structure. Low synthetic yields observed for C-terminal GDF15 fragments longer than 20 amino acids limited the application of SPOT arrays for the investigation of the binding properties of long GDF15 peptide fragments. For this purpose, we applied a parallel SPPS method that allowed a complementary screening of GDF15 peptide fragments ranging from 26 to 34 amino acids. Due to high diversity of peptide sequences and lengths, variation on synthesis yield of individual peptides was observed. Hence, thorough quality control analysis was performed allowing unbiased processing and interpretation of the peptide screening data. In agreement with SPOT data, functional screening of long GDF15 peptide fragments in Table 1 Sequences and potency data of GDF15 fragments. All peptides are C-terminally amidated. Data are presented as mean ± SD, n = 3; X indicates substitution of methionine with norleucine; N.D. for EC 50 > 200 µM.

GFRAL/RET cells indicated inhibition of GFRAL signaling with peptides
deriving from the C-hairpin of GDF15. We observed that fragments 311, 348 and 387 sharing Asp304 as the C-terminal residue, consistently inhibited GFRAL. Consequently, these peptides were synthesized, purified and functionally evaluated, resulting on the identification of 348 (residues 273-304) as the most potent GFRAL inhibitor (EC 50 = 19.1 ± 6.6 µM). We hypothesize that 348 inhibits GFRAL by disrupting the GDF15/ GFRAL interface. This hypothesis is supported by previously reported structural information of GDF15 in complex with GFRAL ECD, indicating that the amino acid residues of 348 (273− 304) form an extensive binding interface between the C-hairpin of GDF15 and a hydrophobic pocket on the D2 domain of GFRAL ECD (Fig. 3) [22,35]. Importantly, Hsu et al. demonstrated that hydrophobic interactions formed by Val283 and Ile285 are critical for GDF15/GFRAL interaction as single point mutations in these residues (V83A and I85A) reduced GDF15-dependent signaling. Additionally, Li et al. reported a cryo-EM structure of the extracellular ternary complex of GDF15/GFRAL/RET revealing a binding interface between GDF15 and the C-terminal cysteine rich domain (CRD) of RET [35]. The RET/GDF15 interaction site appears opposite to the GFRAL/GDF15 site resulting in a 'sandwich' formation between RET, GDF15 and GFRAL. The RET-CRD interacts mainly through hydrophobic contacts with the N-and C-hairpin loops of GDF15 (Fig. 3). Interestingly, mutation of Tyr297 (Y297E) in GDF15 abolished the formation of the ternary complex and dramatically decreased phosphorylation of ERK. We therefore hypothesise that 348 could inhibit GFRAL signaling through disruption of both the GDF15/GFRAL and GDF15/RET interface.
In summary, this study reports the application of two complementary peptide technologies, SPOT peptide arrays and functional screening, for the focused identification of GDF15 peptide fragments inhibiting the GFRAL/RET receptor complex. We here describe, for the first time, peptide inhibitors of GFRAL, though the pharmacological potential of these peptides remains to be investigated, including binding affinity to the extracellular domains of GFRAL and RET. Also, identification of amino acid residues essential for receptor binding will potentially contribute to the optimization of the inhibitory potency of GDF15 peptides. It should be noted that the native structural conformation of the GFRAL/RET receptor complex could be essential for determination of peptide binding and receptor inhibition. Additionally, introduction of conformational constraints locking the peptide into a hairpin conformation may potentially improve the pharmacological properties of GFRAL peptide inhibitors. We hope that this study will encourage further investigations towards the development of GFRAL peptide antagonists, opening new directions for in vivo elucidation of GFRAL inhibition in various diseases, notably cancer cachexia and other anorectic conditions. While it is well-established that cancer cells overexpress and secrete GDF15, it should be noted that GDF15 may play a complex role in tumorigenesis as GDF15 has been reported to suppress growth of certain early tumor types while stimulating tumor growth in advanced cancer [36]. Future studies must aim to profile GFRAL peptide antagonists for effects on appetite function.

Data availability
Data will be made available on request.

Acknowledgments
This work was supported by a PhD grant (Flora Alexopoulou) from Innovation Fund Denmark (grant no. 9065-00138B) and Danish Diabetes Academy which is funded by the Novo Nordisk Foundation (grant no. NNF17SA0031406).