CHO-S cells (Thermo Fisher Scientific) were cultured in serum-free Power-CHO medium (Lonza) supplemented with 8 mM L-glutamine (Thermo Fisher Scientific) and 1% penicillin and streptomycin mix (Biowest). The cells were subcultered every 2–3 days at a seeding density of 0.2–0.3 × 10⁶ cells·mL⁻¹. The cells were grown at 37°C with 8% CO₂ in a shaking incubator (140 rpm).

NCI-H520 (lung squamous cell carcinoma, FGFR1-positive), NCI-H1581 (lung large cell carcinoma, FGFR1-positive), and HCC95 (lung squamous cell carcinoma, FGFR1-negative) were obtained from the American Type Culture Collection (ATCC). NCI-H520 and NCI-H1581 were cultured in RPMI 1640 (Gibco) with 10% fetal bovine serum (Biowest) and 1% penicillin and streptomycin mix (Biowest); HCC95 cells were cultured in RPMI 1640 (Gibco) with 10% fetal bovine serum (Biowest), sodium bicarbonate (Gibco), and 1% penicillin and streptomycin mix (Biowest). The cancer cell lines were cultured at 37°C with 5% CO₂.

The extracellular domain (ECD) of FGFR1 fused to the Fc domain of human IgG1 was produced as described previously by our group (Sokolowska-Wedzina et al., 2014). The Fc domain was expressed and purified in the same manner.

FGFs, namely, FGF1, FGF2, and FGF12, were expressed in E. coli and purified by affinity chromatography as described before (Szlachcic et al., 2016; Krzyscik et al., 2017; and Sochacka et al., 2020). FGF16 was produced in the E. coli (BL21) expression strain. Bacteria were grown in the LB medium with 100 µg/ml amplicilin and 0.003% chloramphenicol to OD₆₀₀ = 0.8 at 37°C with shaking (180 rpm). Then the protein expression was induced by adding 0.1 mM IPTG (Irish Biotech GMBH), and the culture was incubated at 30°C for 16 h. Next, the bacteria were harvested by centrifugation at 6000xg, resuspended in lysis buffer (20 mM Tris-HCl, 0.5 M NaCl, 1 mM EDTA, 0.1 mM PMSF, and pH 7.4) and homogenized by sonication. The lysate was centrifuged at 15,000xg at 4°C for 45 min. The supernatant was diluted in binding buffer (20 mM Tris-HCl, 0.5 M NaCl, 1 mM EDTA, 0.1 mM PMSF, and 1 mM DTT) and loaded on sepharose–heparin resin. The column was washed with washing buffer (20 mM Tris-HCl, 0.7 M NaCl, 1 mM EDTA, 0.1 mM PMSF, and 1 mM DTT), and proteins were eluted with elution buffer (20 mM Tris-HCl, 2 M NaCl, 1 mM EDTA, 0.1 mM PMSF, and 1 mM DTT). The elution fractions containing FGF16 were dialyzed to PBS pH 7.2 at 4°C overnight. FGF6 (#554224) and FGF8 (#SRP4053) were purchased from Biocourse and Sigma Aldrich, respectively. Recombinant FGF4, FGF5, FGF7, and FGF10 were provided by Marta Minkiewicz and Martyna Sochacka from Protein Engineering Group.

ECD_FGFR1-Fc and Fc domain were fluorescently labeled with HiLyte488 dye according to manufacturers’ protocol (#225402, Thermo Fisher Scientific, United States).

Identified peptides’ sequences from FGF4 were cloned into the pLEV113 vector containing the sequence of Fc domain and expressed in CHO-S cells (Chinese Hamster Ovary). The production and purification of peptibodyF4_1, peptibodyF4_2, and peptibodyF4_3 were based on the protocol described previously for peptibodyF2 and FGFR1-Fc fusion protein (Sokolowska-Wedzina et al., 2014and Jendryczko et al., 2020) with a production time of 10 days for peptibodyF4_1 and 7 days for peptibodyF4_2 and peptibodyF4_3. Production of new peptibodies during the time was visualized by Western blotting using antihuman IgG (Fc) antibody conjugated with HRP (#ab97225) from Abcam.

The interaction measurements were performed using a Biacore 3000 instrument (GE Healthcare) at 25°C in PBS with 0.05% Tween20, 0.1% BSA, 0.02% NaN₃, and pH 7.4. The extracellular domains of FGFR1 in Fc fusions (in 10 mM sodium acetate, pH 5.0) were immobilized on the CM5 sensor chip surface (GE Healthcare) at 8800 RU using an amine coupling protocol. To determine kinetic constants of the interaction between peptibodies F4_1 or F4_2 or F4_3 and FGFR1, a set of dilutions of peptibodies at concentrations ranging from 0.15 to 4.8 μM or FGF4 (concentration from 75 to 600 nM) were injected at a flow of 30 μl/min. The association and disassociation were monitored for 180 and 240 s, respectively. Between injections, 10 mM glycine (pH 1.5) was applied to regenerate the sensor chip surface. The data were analyzed using the BIAevaluation 4.1 software (GE Healthcare). SPR measurements showed that the peptibodies did not interact 1:1 with the receptor, and therefore, the data did not fit to the standard 1:1 Langmuir binding model. For this reason, we used steady-state affinity analysis (which is particularly suitable for measurements of weaker interactions) to determine the K_D value without computing kon and koff (Suh et al., 2014). Response values from the last 10 s of the association phase were averaged and used to determine the K_D.

Experiments were performed on HCC95 and NCI-H520 cell lines (10⁴ cells per well). Before experiments, cells seeded on 96-well plates were starved for 4 h.

The cells were incubated with peptibodyF4_1 (4 µg per well) or peptibodyF4_3 (4 µg per well), and after 10- and 30-min, plates were cooled to stop internalization. The cells were washed with PBS and fixed with 4% paraformaldehyde for 15 min at room temperature, permeabilized in 0.3% Triton X-100 at 4°C for 10 min, and blocked with 2% BSA and 0.3 M glycine in PBS-T for 1 h at room temperature. The cells were incubated overnight at 4°C with primary antibodies: rabbit anti-EEA1 antibody (#2411S Cell Signalling), rabbit anti-LAMP1 antibody (#ab24170, Abcam), or anti-FGFR1 antibody (#3472 Cell Signaling), diluted in PBS-T with 2% BSA, followed by further 1 h incubation with Donkey anti-Rabbit (H + L) ReadyProbes secondary antibody, Alexa Fluor 594 (#R37119, Thermo Fisher Scientific), diluted in 2% BSA in PBS-T. Peptibodies were labeled with Zenon™ Human IgG Labeling Kit (#Z25402, Invitrogen, Thermo Fisher Scientific) for 1 h at room temperature and fixed with 4% paraformaldehyde for 15 min at room temperature. Nuclei were labeled with NucBlue reagent (#R37605, Thermo Fisher Scientific).

Wide-field fluorescence microscopy was performed with a Zeiss Axio Observer Z1 fluorescence microscope using a LD-Plan-Neofluar 40/0.6 objective and Axiocam 503 (Zeiss, Germany). Images were processed with Zeiss ZEN 2.3 software (Zeiss, Germany) and Adobe Photoshop CS6 (Adobe, San Jose, CA, United States).

Serum-starved NCI-H520 cells were incubated with increasing concentrations (1, 10, 100, and 1,000 nM) of DyLite650-labeled peptibodyF4_1, labeled according to manufacturers’ protocol, and incubated for 30 min at 37°C. In order to stop the receptor trafficking, cells were placed on ice and washed with ice-cold PBS. Non-internalized peptibody was removed from the cell surface by washing in acid stripping buffer. The cells were detached with 10 mM EDTA-PBS and washed in FACS buffer. Flow cytometric analysis was performed using a NovoCyte 2060R instrument and NovoExpress software (ACEA Biosciences, San Diego, CA, United States).

The cytotoxicity of peptibodyF4_1vcMMAE was performed on FGFR1-negative cell line (HCC95) and FGFR1-positive cell lines (NCI-H520 and NCI-H1581). Cells were seeded on a 96-well plate (5,000 cells per well) and incubated for 24 h at 37°C with 5% CO2. Conjugate was added to the cells in different concentrations (from 0.05 to 200 nM) and incubated for 96 h. The cytotoxicity of peptibodyF4_1vcMMAE was determined by Alamar Blue reagent (Invitrogen) according to the manufacturer’s protocol. The fluorescence intensity with an excitation of 560 nm and an emission of 590 nm was measured using an Infinite M1000 PRO plate reader (Tecan). Every experiment was performed in triplicates. EC₅₀ values were calculated based on the Hill equation using Origin 7 software (Northampton, MA).

Interactome screening is one of the approaches for the identification of novel target-binding peptide sequences. For FGFR1, we chose its natural ligands—the family of fibroblast growth factors (FGFs). Identifying the FGF family members that display the ability to be bound by FGFR1 even in the unfolded state is the starting point to find the binding region and identify the linear peptide sequence. In many cases, binding sites in the proteins are scattered through their sequence and fully formed only in the properly folded protein. Therefore, our first step involves denaturing FGFs and incubating them with fluorescently labeled FGFR1—for the initial screening to verify if FGFs in their unfolded state can still bind to FGFR1 (Figure 1A). In the screening, we used a subset of FGFR1-binding proteins that are independent of co-receptors (e.g., β-Klotho), considering all FGFs’ subfamilies.

FGFs in the unfolded state immobilized on a PVDF membrane were incubated with fluorescently labeled FGFR1—recombinant protein composed of FGFR1 extracellular domain (FGFR1-ECD) fused with Fc (fragment crystallizable region of IgG1). Since we cannot fully control if the proteins bound to the membrane remain in the denatured state, to minimize the chance of refolding on the membrane, the assay involved fluorescently labeled FGFR1 for probing, and not primary and secondary antibodies, eliminating additional steps. We have used the PVDF membrane that binds proteins through hydrophobic interactions, which disrupted usually, making it impossible for the protein to properly form its hydrophobic core. To avoid detection of Fc-mediated binding to unfolded FGFs, the Fc alone was used as a negative control (Figure 1B). We have observed fluorescent signal corresponding to FGFR1 binding for FGF2, FGF4, FGF6, and FGF12. FGFs with highest signals from FGFR1 binding are FGF2, FGF12, and FGF4, successively. We eliminated FGF2 because of a strong signal of the Fc-domain binding, and we did not succeed in the identification of any peptide binders after proteolytic cleavage of FGF12. Here, we describe a process of FGFR1-binding site identification within FGF4 sequence, which in principle can be applied to other FGFs also.

To generate shorter peptidic fragments, we digested FGF4 with trypsin or chymotrypsin, but chymotryptic fragments were more appropriate; generated fragments were more uniform in length than tryptic peptides. Trypsin cleavage sites were clustered in few regions, yielding very short (2-5 amino acid) peptides; hence, we used chymotrypsin for further study. We used chymotrypsin immobilized on Sepharose resin, in order to minimize protease activity and carry on to next steps of peptide isolation, and to minimize the peptide purification step after digestion, which may potentially deplete the sample. In the next step, we incubated the peptide mixtures with FGFR1 or Fc-immobilized on ProteinA-Sepharose. Again, we used Fc as a negative control to exclude peptide binding to the Fc region, instead of the extracellular FGFR1 part. Finally, we analyzed the eluted peptides by mass spectrometry for the identification of potential peptidic FGFR1 binders. We obtained two peptides present in the eluate from the column with FGFR1, and not Fc (Figure 1C). They correspond to 136–150 and 179–206 regions of the FGF4 sequence.

In parallel to peptide isolation and mass spectrometry identification aforementioned, we have used a complementary approach based on peptide array immobilized on the membrane. This PepSpot customized peptide arrays consist of 15 amino-acid peptides spanning FGF4 sequence, with 10 amino acid overlap, and immobilized onto a cellulose membrane. After incubation with fluorescently labeled FGFR1 recombinant protein (extracellular domain of FGFR1 fused with Fc), we observed intense signals from peptide nos. 4, 10, 15, and 19–21. To eliminate unspecific Fc-mediated binding signal, we used a fluorescent-labeled Fc as a negative control. Incubation with Fc-HiLyte488 showed only one unspecific signal from peptide no. 4 (WAGRGGAAAPTAPNG) (Figure 2A). The fluorescence values were normalized to peptide no. 20, which had the highest fluorescence value after subtracting the Fc-derived background (Figure 2B).

Interestingly, two regions with high PepSpot scores [peptide 15 (SRFFVAMSSKGKLYG), and peptides 19–21 (KYPGMFIALSKNGKT, KNGKTKKGNRVSPTM, and RVSPTMKVTHFLPRL); Figure 2B] coincided with two regions identified with pull-down and MS approach above. Peptide no. 10 (LRRLYCNVGIGFHLQ), with positive readout in PepSpot analysis, was not detected in MS analysis. Peptide no. 4, which has shown Fc-binding, was excluded from further analysis. In order to visualize the location of these peptides within the FGF4 structure, and to check if identified peptides are positioned close to the predicted FGF–receptor interaction site, a homology model of the FGF4 molecule based on the template of FGF1–FGFR1 complex (PDB: 1EVT) was generated with the use of Swiss-Model server (Waterhouse et al., 2018) (Figure 2C). Identified peptides are not structurally clustered in one region of FGF4, and whereas peptideF4_3 is within the region predicted to be interacting with FGFR1, peptideF4_2 is located on the opposite side of the molecule. However, conformation of isolated peptides may differ from the one they adopt in a full-length folded protein, and we did not exclude peptides based on indirect structural information.

Therefore, we selected all three peptides (4_1: IALSKNGKTKKGNRVSPTMKVTHFLPRL, 4_2: FVAMSSKGKLYGSPF, and 4_3: LRRLYCNVGIGFHLQ) for further characterization in the peptibody format.

To overcome clinical limitations of peptidic agents (e.g., short life-time and fast renal clearance), we genetically fused a newly identified peptide F4_1, F4_2, and F4_3 at the C-terminus of crystalizable fragment of human IgG1 (Fc domain) and created peptibodyF4_1, peptibodyF4_2, and peptibodyF4_3, respectively. As aforementioned, sequences of peptides F4_1, F4_2, and F4_3 are derived from FGF4 (residues 179–206, 136–150, and 83–97, respectively) (Figure 3A). We introduced additional glycine residue as a spacer between the Fc domain and the targeting peptide.

We overproduced peptibodies in CHO-S cells and purified them using affinity chromatography based on the protocol previously developed in our group (Figure 3B) (Sokolowska-Wedzina et al., 2014 and Jendryczko et al., 2020). We obtained 20 mg of at least 95% pure peptibodyF4_1 and 15 mg of at least 95% pure peptibodyF4_2 from 1 L of cell culture. PeptibodyF4_3 was characterized with much lower levels of expression and yielded between 0.5 and 3 mg of 95% pure protein from 1 L of culture, depending on the expression batch. The identity and purity of peptibodies were confirmed by Coomassie Brilliant Blue staining and Western blotting.

The crucial feature of newly obtained peptibodies is their ability to efficiently bind FGFR1 and to be efficiently and specifically internalized into FGFR1-expressing cells. Therefore, new peptibodies were characterized with regard to their binding to FGFR1 with the use of SPR (surface plasmon resonance), and with fluorescence microscopy to evaluate the level of receptor binding and internalization into FGFR1-expressing cells.

For SPR analysis, recombinant FGFR1 was immobilized on the sensor, and association and dissociation profiles of newly developed peptibodies were measured. We observed binding signal for peptibodyF4_1, and titration experiment showed that the binding is dependent on the peptibody concentration (Figure 3C). PeptibodyF4_1 binds FGFR1 with K_D of 2.99 × 10⁻⁶ M, as estimated from steady-state values. For peptibodyF4_2, there were negligible signal levels, suggesting that this variant does not show FGFR1 binding (Figure 3D). PeptibodyF4_3 showed FGFR1 binding and the measured K_D equals to 1.09 × 10⁻⁶ M (Figure 3E), similar to peptibodyF4_1. Based on these results, we chose peptibodies F4_1 and F4_3 for further analysis. FGFR1 binding affinities were significantly low for FGF4, for which K_D equaled to 6.57 × 10^–8 M, as measured by SPR (Supplementary Figure S1A).

To check if in vitro FGFR1 binding translates to efficient binding and internalization into FGFR1-expressing cells, we used lung cancer cell lines characterized with regard to FGFR1 expression levels. Internalization into NCI-H520 cells, previously characterized as lung cancer cells with increased levels of FGFR1 expression, was compared to internalization into HCC-95 lung cancer cells with low levels of FGFR1 (Wynes et al., 2014). To evaluate if micromolar affinities to FGFR1 are sufficient to observe internalization into FGFR1-positive cells, NCI-H520 cells were incubated with various concentrations of the peptibody and analyzed by flow cytometry (Supplementary Figure 1B). We can observe internalization signal for submicromolar concentrations of peptibody, suggesting that relatively low in vitro affinities to FGFR1, measured by SPR, still allow for effective peptibody binding to the receptors expressed in the cells.

Since peptibodyF4_1 is a candidate for a cytotoxic drug carrier, we checked by fluorescence microscopy if internalized peptibodies colocalize first with endosomal proteins (with anti-EEA-1 antibodies), and then at later stage with lysosomal markers (anti-LAMP-1 antibodies). For the release of cytotoxic drugs, the molecule is required to reach the lysosomal compartment to allow for either protein degradation or drug-linker cleavage, both leading to the release of cytotoxic cargo.

PeptibodyF4_1 internalization studied with fluorescence microscopy showed that it is internalized much more efficiently to FGFR1-positive NCI-H520 cells, with negligible signal observed for receptor-negative HCC-95 cells, suggesting the FGFR1-specific internalization (Figure 4A). Green fluorescence signal corresponding to peptibody labeled with Zenon488 reagent partially colocalized with red fluorescence of visualized endosomes or lysosomes, suggesting that at least a fraction of internalized peptibodyF4_1 is trafficked via endosomes to lysosomes. PeptibodyF4_3 behaved very similarly, as shown by fluorescence microscopy–internalized readily into FGFR1-positive cells, and not into FGFR1-negative, and showed partial colocalization with both endosomal and lysosomal markers (Figure 4B). The colocalization experiment with FGFR1 (Supplementary Figure S2A) and the fact that peptibody internalization can be competed off with unlabeled FGF4 (Supplementary Figure S2B) indicate that peptibodyF4_1 cell uptake is mediated by FGFR1.

Since binding to FGFR1 can potentially cause FGFR1 dimerization and receptor activation, we also analyzed if peptibody can induce such responses. FGFRs are activated by binding their specific ligands—FGFs. Dimerization and changes in conformation within the structure of the receptor leads to activate downstream signaling cascades (Ornitz and Itoh, 2015). While short-term stimulation leads to the activation of metabolic response, long-term response causes cell proliferation (Zinkle and Mohammadi, 2018), and any proliferative stimulation in the case of cancerous cells is highly undesirable.

FGFR1 activation was tested on mouse fibroblast cells (NIH3T3), and FGFR-dependent signaling pathway activation was used as readout of short-term response. PeptibodyF4_1 caused moderate phosphorylation of FGFR1 and activation of downstream ERK1,2 kinases compared to natural FGFR1 ligand, FGF1, and much higher concentrations of peptibody were required to see any effect on cell signaling (Figure 4B). Not surprisingly, a long-term proliferation assay on starved NIH3T3 cells stimulated with peptibodies at different concentrations showed that no significant proliferative activity can be observed upon the addition of peptibodyF4_1 to the cells (Figure 4C).

FGFR1-binding peptibody can be used as a carrier of cytotoxic drug for its targeted delivery, analogously to antibody–drug conjugates (ADCs), well established in the clinic. The targeting molecule can specifically deliver cytotoxic compound to target-expressing cells, with negligible amounts of conjugate internalized to target-negative cells.

As a cytotoxic drug, we used monomethyl auristatin E (MMAE)—a synthetic derivative of toxin dolastatin-10. This highly cytotoxic agent with anti-mitotic properties inhibits microtubule formation (Waight et al., 2016). To ensure that the cytotoxic cargo will be released only into the cancer cells, we used valine–citrulline (vc) linker cleaved by cathepsin B in the lysosomal compartments, and well characterized in clinical studies (Bargh et al., 2019).

Our conjugation strategy based on maleimide–thiol reaction using TCEP-reduced cysteine residues in the hinge region of the Fc domain, analogously to antibody modification with drugs (Figure 5A). Unfortunately, the standard protocol used for the antibodies and Fc domain–containing proteins yielded low efficiency of conjugation and precipitate formation when the reaction was scaled up. To find the optimal conditions of conjugation of peptibodyF4_1 and peptibodyF4_3, we analyzed the effect of the various levels of excess of cytotoxic drug over protein, protein concentration, and buffer composition, including additives such as urea and glycerol (Supplementary Figures S3A,B). To increase the solubility of conjugated protein, we have also verified if a PEGylated form of vcMMAE performs better. For peptibodyF4_1 conjugation yields for PEGylated MMAE were much lower than those for vcMMAE, and final conjugation protocol included 15-fold excess of vcMMAE over reduced peptibodyF4_1 and buffer containing 5% glycerol and 1 M urea. Additives in the reaction buffer prevented protein precipitation during the reaction. PeptibodyF4_1 conjugates were purified using ion exchange chromatography to remove any remaining free vcMMAE, and the resulting conjugate purity was >80% (Figure 5B). The drug-to-protein ratio was determined spectrophotometrically by analyzing the absorbance values at 248 and 280 nm, and equaled 2.1.³³

For peptibodyF4_3, we have also observed the highest efficiency of conjugation for vcMMAE (Supplementary Figure S4A). Reactions with PEGylated MMAE, either with short PEG chains (PEG4vcMMAE) or longer PEG chains (PEG27vcMMAE), proceeded with precipitation were not clearly visible in small-scale reactions but apparent from SDS-PAGE analysis of conjugation reactions (Supplementary Figures S4B,C). Scaling up the reactions lead to substantial precipitation, and we were not able to purify sufficient amounts of conjugate to perform any cell studies (Supplementary Figure S4D). Interestingly, the conjugate gel separation pattern for vcMMAE peptibodyF4_3 conjugation reactions was slightly different from the predicted two bands for Fc cysteine modifications, with the third band present and suggesting modification of additional cysteine. PeptideF4_3 is the only one of the three analyzed containing cysteine residues within its sequence, which poses the risk of drug modification in the targeting region.

As the linker used by us is cleaved by lysosomal proteases, we have checked cellular localization of peptibodyF4_1vcMMAE conjugate in both FGFR1-positive and FGFR1-negative cells (Figure 5C). PeptibodyF4_1vcMMAE conjugate, visualized Zenon488 reagent with green fluorescence, colocalizes with LAMP-1 staining, suggesting that drug release can occur in lysosomal compartment.

Finally, we tested the peptibodyF4_1-vcMMAE conjugate for its ability to specifically deliver drug to FGFR1 positive cancer cell lines as the ultimate prerequisite for cytotoxic peptibody conjugates is their ability to cause cell toxicity dependent on the presence of molecular target, in our case, FGFR1.

We performed cytotoxicity tests on a set of non–small-cell lung carcinoma cells (NCI-H520, NCI-H1581, and HCC-95). The NCI-H520 and NCI-H1581 cells represent FGFR1-positive cell lines, whereas HCC-95 was used as a negative control. We observed a strongly decreased viability of cell lines overexpressing FGFR1 in response to increasing concentration of peptibodyF4_1vcMMAE, while HCC-95 cells were insensitive of conjugate (Figure 5D). Strong and specific cytotoxic effect is confirmed by EC50 values for peptibodyF4_1vcMMAE: 18.1 and 13.4 nM for NCI-H520 and NCI-H1581 cell lines, respectively (Figure 5E).