Myomedin variants developed for in vitro PD-L1 diagnostics in tissue samples of non-small cell lung carcinoma patients

Abstract

Background

The treatment of non-small cell lung cancer (NSCLC) patients is correlated with the efficacy of immune checkpoint blockade therapy (ICB) targeting programmed cell death ligand 1 (PD-L1) or its cognate receptor (PD-1) on cancer cells or infiltrating immune cells. Analysis of PD-L1/PD-1 expression in tumor tissue represents a crucial step before PD-L1/PD-1 blocker usage.

Methods

We used directed evolution of protein variants derived from a 13 kDa Myomedin loop-type combinatorial library with 12 randomized amino acid residues to select high-affinity binders of human PD-L1 (hPD-L1). After the ribosome display, individual clones were screened by ELISA. Detailed analysis of binding affinity and kinetics was performed using LigandTracer. The specificity of Myomedins was assessed using fluorescent microscopy on HEK293T-transfected cells and cultured cancer cells in vitro, formalin-fixed paraffin-embedded (FFPE) sections of human tonsils, and FFPE tumor samples of NSCLC patients.

Results

Seven identified PD-L1 binders, called MLE, showed positive staining for hPD-L1 on transfected HEK293T cells and cultured MCF-7 cells. MLE031, MLE105, MLE249, and MLE309 exhibited high affinity to both human and mouse PD-L1-transfected HEK293T cells measured with LigandTracer. The diagnostic potential of MLE variants was tested on human tonsillitis tissue and compared with diagnostic anti-PD-L1 antibody DAKO 28-8 and PD-L1 IHC 22C3 pharmDx antibody. MLE249 and MLE309 exhibited an excellent overlap with diagnostic DAKO 28-8 (Pearson´s coefficient (r) = 0.836 and 0.731, respectively) on human tonsils on which MLE309 exhibited also excellent overlap with diagnostic 22C3 antibody (r = 0.876). Using three NSCLC tissues, MLE249 staining overlaps with 28-8 antibody (r = 0.455–0.883), and MLE309 exhibited overlap with 22C3 antibody (r = 0.534–0.619). Three MLE proteins fused with Fc fragments of rabbit IgG, MLE249-rFc, MLE309-rFc and MLE031-rFc, exhibited very good overlap with anti-PD-L1 antibody 28-8 on tonsil tissue (r = 0.691, 0.610, and 0.667, respectively). Finally, MLE249-rFc, MLE309-rFc and MLE031-rFc exhibited higher sensitivity in comparison to IHC 22C3 antibody using routine immunohistochemistry staining system Ventana, which is one of gold standards for PD-L1 diagnosis.

Conclusions

We demonstrated the development of MLE Myomedins specifically recognizing hPD-L1 that may serve as a refinement tool for clinical PD-L1 detection.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06699-6.

Introduction

The discovery of immune checkpoint molecules, programmed cell death-1 (PD-1, CD279) and its cognate programmed cell death ligand 1 (PD-L1, also known as CD274 or B7-H1), and unveiling their potential use as targets in anti-cancer therapy or as prognostic markers in various types of cancer have opened new treatment opportunities [1–3]. PD-L1, a type I transmembrane glycoprotein, is a member of the B7 protein family, consisting of 290 amino acids. Under normal conditions, PD-L1 is expressed on vascular endothelium and antigen-presenting cells (APCs) in organs including the lung, liver, and kidney [4]. However, PD-L1 is also identified in various cancer cells to inhibit T cell activation, including interaction via the PD-1 receptor guiding T cell exhaustion and induction of T cell apoptosis. Therefore, it enables cancers to bypass the immune checkpoint control [5–7]. It results in the generation of an immunosuppressive tumor microenvironment (TME) that allows tumor cells to evade immunosurveillance [7, 8].

Expression of PD-L1 on tumor tissue sections detected by immunohistochemistry staining (IHC) has been established as a diagnostic biomarker to monitor the response to anti-PD-L1/anti-PD-1 therapy [9, 10]. IHC allows PD-L1 detection on tumor cells and tumor-infiltrating cells such as B cells, dendritic cells (DCs), macrophages, myeloid-derived suppressor cells (MDSCs), T cells, and neutrophils [10, 11]. Typically, in immune checkpoint blockade therapy (ICB), therapeutic antibodies blocking the interaction between PD-1 and PD-L1 are used in line with the diagnostic assays approved by the Food and Drug Administration (FDA) and the European Medicines Agency (EMA). Assays utilize diagnostic antibodies that detect the PD-L1 expression on tissue samples. Based on the expression of PD-L1 levels, these antibodies may help to indicate the most beneficial therapeutic approach for the patient.

Overexpression of PD-L1 is often found in many solid tumors including non-small cell lung cancer (NSCLC) [12]. Although overexpression of PD-L1 in NSCLC is a typical indicator of poor prognosis, a positive correlation between PD-L1 expression and response to the therapeutic agents has been shown in the case of pembrolizumab, atezolizumab, and nivolumab [13, 14]. Currently, several commercially available anti-PD-L1 antibody clones are used as diagnostic tools for scoring NSCLC patients [15], and three FDA-approved diagnostic assays are available for determination of the PD-L1 expression levels in the tested NSCLC tissue sections. [10]. For instance, Dako 22C3 (PD-L1 IHC 22C3 pharmDx) is assigned as a companion diagnostic to verify PD-L1 presence in a range of solid tumors, including tissue samples of NSCLC patients selected for treatment with pembrolizumab. Likewise, Ventana SP142 (PD-L1 SP142 Assay) is used as a companion diagnostic tool for the assessment of PD-L1 expression in NSCLC when atezolizumab is used for the treatment [10, 16]. Dako 28-8 (PD-L1 IHC 28-8 pharmDx) is also used as a complementary diagnostic tool together with ipilimumab and nivolumab for the treatment of NSCLC [14].

According to the FDA classification, the ‘companion’ diagnostic tool provides the information essential for implementing effective and safe personalized medicine, whereas the ‘complementary’ diagnostic tool gives a general view of the potential benefit/risk ratio of the therapy before administration of the drug [17]. These assays, however, differ in their sensitivity and reproducibility [18]. Data collected from various studies suggest the convergence of results for Dako 28-8, Dako 22C3, and Dako SP263 with the exclusion of Ventana SP142 [19, 20]. Therefore, it is important to generate small binding proteins as non-immunoglobulin alternatives with the required specificity and high affinity to be used as PD-L1 diagnostic tools before the treatment strategy is designed for NSCLC patients.

Recently, we have developed a set of high-affinity binding proteins called Myomedins, which were derived from the scaffold of domain 10 of a human contractile protein, myomesin-1, and selected from a highly complex Myomedin loop-type combinatorial library by ribosome display [21, 22]. To develop PD-1-specific Myomedin variants, named MBA, we used a concept of beta (β)-sheet Myomedin library with the randomization of 12 amino acid residues, thus providing the theoretical library complexity of 10¹⁵ protein variants. These high-affinity MBA proteins were used to stain the PD-1-positive cell populations in frozen sections of human tonsils and biopsy samples of patients with NSCLC and identified mouse PD-1-expressing cells in vivo using PET/CT imaging [23]. Here, we report on the generation of Myomedin variants, named MLE, with the arranged specificity to human PD-L1 using directed evolution. We demonstrate that several MLE proteins displayed specific staining of cell surface-expressed PD-L1 on cultured or paraffin-fixed cells, and in correlation to diagnostic antibodies DAKO 28-8 and 22C3, can detect PD-L1 on paraffin-fixed sections of a human tonsil and biopsy samples of NSCLC.

Materials and methods

Ribosome display selection

The Myomedin scaffold loop library was assembled as described previously [21] by the polymerase chain reaction (PCR) technique. The ribosome display consisted of the following steps: Coupled transcription and translation reactions, preselection, selection to human PD-L1, isolation of mRNA, reverse transcription, and PCR reactions to produce a new template for transcription and translation. All the above steps, except for the preselection and selection, were described earlier [21, 22]. The selection of a ternary complex was performed in two different formats: 5 selection rounds in the Maxisorp plate (Nunc A/S, Roskilde, Sjælland, Denmark) or 3 selection rounds on the streptavidin-coated magnetic beads (Dynabeads™ MyOne™ Streptavidin T1, Thermo Fisher Scientific, Waltham, MA, USA) in the solution. Preselection in streptavidin-coated wells (1 µg/ml) was applied in each round before the selection step and, starting from the 2^nd round, also to directly coated recombinant PD-1 (1 µg/ml). For the selection in plates, different concentrations (0.2, 0.2, 0.1, 0.05, and 0.02 µg/ml) of biotinylated human PD-L1 (PD-L1-His-Avi(biot), AVI9049-050, R&D Systems, Minneapolis, MN, USA) were immobilized on a streptavidin-coated Maxisorp plate (Nunc A/S, Roskilde, Sjælland, Denmark) in the subsequent selection rounds. After the 5^th round of ribosome display, transcribed cDNA of selected Myomedin variants was cleaved by NcoI and BamHI enzymes and introduced by ligation into the pET28b vector with terminal V5-tag, thus forming the final cDNA library for further screening. In the case of selection in solution, the stopped transcription/translation reaction was preselected on magnetic beads (Dynabeads™ MyOne™ Streptavidin T1, Thermo Fisher Scientific, Waltham, MA, USA) by 1 h incubation at 4 °C. Then, the supernatant was transferred to an empty tube, and 380 nM (or 100 nM and 36 nM in the 2^nd and 3^rd rounds, respectively) of biotinylated human PD-L1 was added and incubated at 4 °C for another 1 h. Then, streptavidin-coated magnetic beads were added to the mixture for binding. Starting from the 2^nd round, non-biotinylated PD-L1 was added to the incubation mixture to perform off-rate selection. The excess of non-biotinylated over biotinylated PD-L1 was 10 times and 500 times in the 2^nd and 3^rd rounds, respectively. After washing and elution of mRNA, reverse transcription to cDNA was performed. The generated cDNA was cloned into a pET28b plasmid vector containing a V5-tag at the C-terminus, thus forming the final plasmid library of the transcribed cDNA.

Identification of PD-L1 specific MLE variants

The generated plasmid cDNA libraries were used to transform cells of Escherichia coli (E. coli) XL1 strain, and purified plasmid DNA of selected bacterial colonies was sequenced. The bacterial clones with the verified sequence were grown in the E. coli BL21 (DE3) strain for protein production. Initially, individual colonies of E. coli BL21 (DE3) were cultured overnight at 37°C in 2 ml LB broth supplemented with kanamycin (60 µg/ml). The next day, the night culture was diluted 20 times in 10 ml or 200 ml LB media with kanamycin (60 µg/ml) and further incubated at 37°C until an absorbance (optical density, OD₆₀₀) of 0.8. Then, protein production was induced by 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) for 4 h at 32°C, and cell pellets were collected by centrifugation (5000×g) for 10 min at 4°C and stored at -20°C. To increase protein production, bacterial cell culture (OD₆₀₀ at 1.0) was induced with 1 mM IPTG, and protein expression was performed overnight at 25°C. For the identification of target-specific variants, cell lysates were prepared by dissolving the pellet in 200 µl of B-PER buffer (Thermo Fisher Scientific, Waltham, MA, USA) followed by 15 min incubation at room temperature (RT) and centrifugated at 14000×g for 10 min at 4°C. Cleared bacterial lysates were then 50 times diluted in protein-free blocking buffer (PFBB) (Thermo Fisher Scientific, Waltham, MA, USA) and used in the screening ELISA. On the other hand, cell pellets were sonicated for 5 min (10 sec on/10 sec off) by the ultrasonic disruptor Misonix S3000 sonicator in 150 mM NaCl, 50 mM Tris buffer at pH 8.0, centrifuged (18000×g) for 10 min at 4°C, and protein was purified using affinity chromatography (Ni-NTA agarose, Qiagen, Hilden, Germany), and serially diluted (≤ 2 µg/ml) purified protein was used in the binding ELISA assay. In ELISA, Maxisorp 96-well plates (Nunc A/S, Roskilde, Denmark) were coated with ≤ 2 µg/ml PD-L1 in the carbonate binding buffer (pH 9.6) for 1 h at RT, then washed three times with 300 µl PBST (PBS with 0.05% Tween 20), and blocked with 1% BSA in PBST or PFBB for at least 2 h at RT or 4°C overnight. Following, cell lysates or purified proteins were applied for 1 h and protein binding was detected using an anti-V5-tag horseradish peroxidase (HRP) conjugated antibody. Reaction with TMB-Complete 2 (3,3’,5,5’-Tetramethylbenzidine, TestLine, Brno, Czech Republic) substrate was performed for 15–30 min, stopped by 2 M sulfuric acid, and color intensity was recorded at 450 nm using the Epoch 2 microplate spectrophotometer (BioTek, Santa Clara, CA, USA).

Competition ELISA

For the competition ELISA assays, 1 µg/ml of PD-L1-Fc protein (R&D Systems, Minneapolis, MN, USA) was coated in a Maxisorp 96-well plate (Nunc A/S, Roskilde, Denmark) using a coating carbonate/bicarbonate buffer (pH 9.6) for 1 h at RT. Next, the plate was washed three times with PBST (PBS + 0.05% Tween, pH 7.4) and overnight blocked with blocking solution (PBS + 1% BSA) at 4 °C. Then, serially diluted purified proteins were prepared and applied on a plate together with a constant 15 nM concentration of PD-1 protein. Following, the plate was incubated for 1 h at RT and then washed with PBST. Anti-PD-1 primary antibody and secondary rabbit anti-goat antibody conjugated with HRP were used for signal detection. Finally, the absorbance was recorded at 450 nm using the Epoch 2 microplate spectrophotometer.

Cell lines and growth conditions

Human embryonic kidney cells (HEK293T), MCF-7, A549, and NIH-3T3 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (BioSera, Cholet, France) supplemented with 10% fetal bovine serum (FBS) and streptomycin-penicillin solution (BioSera, Cholet, France).

Immunofluorescence staining

Staining of hPD-L1-HEK293T, mPD-L1-HEK293T cells, and NIH-3T3 cells on 24-well plates

To perform the human cells (HEK293T) transfectant-based assays, human PD-L1 (Uniprot; entry: Q9NZQ7) or mouse PD-L1 (Uniprot; entry: Q9EP73) DNA sequence synthesized at ThermoFisher Scientific was inserted into the pcDNA6 myc-His A plasmid (Invitrogen, Waltham, MA, USA). Restriction digestion was performed using the HindIII and XbaI endonucleases. Restriction sites for both enzymes were introduced at either of the ends of the synthesized hPD-L1 DNA sequence. A standard protocol as described earlier [21] was used for the transfectant HEK293T cells. Staining of transfected HEK293T cells with MLE proteins was performed using the previously described protocol [21]. Shortly, 48 h after the transfection, the culture medium was discarded, and transfected cells were washed three times with PBS, followed by treatment with MLE proteins (20 µg/ml) in a fresh DMEM medium. Cells were then incubated for 1 h at 37 °C in a CO₂ cell incubator. After the incubation, the medium with unbound MLE proteins was discarded, and hPD-L1-HEK293T cells were treated with a 200 µl mix of polyclonal goat anti-hPD-L1 (R&D Systems, Minneapolis, MN, USA) and monoclonal anti-V5-tag antibodies (Thermo Fisher Scientific, Waltham, MA, USA) prepared in PBS with 1.5% BSA. For the detection of mouse PD-L1, a mix of rat anti-mPD-L1 (Thermo Fisher Scientific, Waltham, MA, USA) and monoclonal anti-V5-tag antibodies (Thermo Fisher Scientific, Waltham, MA, USA) was used for staining of mPD-L1-HEK293T cells or NIH-3T3 cells. All the culture plates were then incubated in the dark for 1 h at RT, followed by three washes with PBS and treated with secondary donkey anti-goat antibody Alexa Fluor 568 conjugate diluted (1:1000) in PBS (1.5% BSA) for 1 h at RT under dark conditions. The antibody was then discarded; cells were washed three times with PBS and prepared for imaging in PBS.

Staining of 4% PFA fixed MCF-7 and A549 cells on 24-well plates

A similar immunofluorescence staining procedure as described above was applied for the paraformaldehyde (PFA)-fixed MCF-7 and A549 cells with minor modifications. Herein, cell fixation was performed with 4% PFA for 10 min at RT, followed by incubation with blocking solution (PBS + 1% BSA) for 30 min at RT. Additionally, 24 h before staining, cells were also treated with 100 ng/ml of interferon-gamma (IFN-γ) and tumor necrosis factor-alpha (TNF-α).

Staining of 4% PFA fixed MCF-7 cells on glass slides

For confocal microscopy, 7 × 10⁵ cells per well were grown on a 12-well plate (Nunc A/S, Roskilde, Denmark) with the inserted microscope 18 nm cover glasses (P-LAB, Prague, Czech Republic). After 24 h of seeding on the plate, cells were first stained to visualize the plasma membrane using Abberior Star Orange (Abberior GmbH, Gottingen, Germany), subsequently fixed with 4% PFA and stained to visualize the hPD-L1 and V5-tag on MLE variants. After staining, microscope cover glasses were transferred from the cell plates onto glass slides (Knittel Microscope Slides) and mounted with a medium containing DAPI for nuclear counterstaining (Vectashield with DAPI, Vector Laboratories, Newark, CA, USA).

All the imaging experiments for live cells or PFA-fixed cells seeded on 24-well plates were performed using a fluorescence microscope, while imaging for PFA-fixed cells on glass slides was performed using a Carl Zeiss LSM 880 NLO confocal microscope. All the images were analyzed with ImageJ software.

Kinetics and binding affinity measured by LigandTracer

The binding affinity and kinetics experiments were performed using the LigandTracer Green Line instrument coupled with a BlueGreen: 488–535 nm (ex/em) detector (Ridgeview Instruments AB, Uppsala, Sweden). For this experiment, cell culture medium from cell dishes with hPD-L1-HEK293T and mPD-L1-HEK293T cells were exchanged with fresh culture medium (without any additives) immediately before the measurements. First, the baseline measurement was recorded only in a culture growth medium for a minimum of 15 min or till the stabilization of the basal fluorescence signal. During the baseline measurements, MLE proteins were incubated in DMEM medium with mouse monoclonal anti-V5-tag Alexa Fluor 488 conjugated antibody (Thermo Fisher Scientific, Waltham, MA, USA). After baseline signal collection, the MLE-antibody mixes were applied to the cell dish, and measurements of the fluorescence signal were recorded for a minimum of 30 min or until the fluorescence signal stabilized. Following, an increasing concentration of the protein-antibody mixes was used to collect the association phase measurements. Finally, the dissociation of the MLE proteins from PD-L1 was measured in the medium without protein-antibody complexes. In all cases, detection of the fluorescence signal was performed for 15 s with 3 s delay intervals. The fluorescence signal was detected using an anti-V5-tag antibody conjugated with Alexa Fluor 488 and recorded for at least 30 min or until the protein began to dissociate rapidly. Binding affinity evaluation was done using TraceDrawer 1.7.1 software, where ‘One-to-one’ or ‘One-to-one depletion corrected’ evaluation methods were used for the estimation of the kinetic parameters (ka, kd, KD).

Competition binding assay tested by LigandTracer

A competition LigandTracer assay was performed on hPD-L1-HEK293T cells. The experiment was performed like the assay for assessing the binding affinity described above with minor modifications. Herein, increasing concentrations (30, 50, and 150 nM) of PD-1 (R&D Systems, Minneapolis, MN, USA) were added every 30 min during the association phase measurements. The data collection without the addition of the competitor (PD-1) was also performed and marked as a control.

Generation of chimeric MLE variants with C-terminal IgG Fc domain

The rabbit IgG Fc fragment Pro96-Lys322 (rFc, Immunoglobulin gamma heavy chain, GenBank: AKM12456.1) was determined based on the alignment with the human IgG Fc fragment (Pro100-Lys330) and the crystal structure of the rabbit IgG C region (RCSB PDB database: 2VUO). MLE chimeric proteins were designed with a V5-tag introduced into a spacer between the Myomedin and rFc domain, and synthetic DNA optimized for E. coli expression was prepared (GeneArt GmbH, ThermoFisher Scientific, Regensburg, Germany). Synthetic cDNA coding for MLE proteins and Myomedin wild-type control was inserted into the pET28b vector by cloning with NcoI and XhoI restriction endonucleases. Finally, 42 kDa recombinant chimeric proteins with N-terminal His-tag (His₆-MLE(MyoWT)-V5tag-rFc) were expressed in Shuffle E. coli strain culture in LB medium amended with kanamycin (30 µg/ml) at 30 °C, and protein expression was induced by 1 mM IPTG at optical density 0.6. After 16 h of cultivation, the bacterial cells were collected by centrifugation (6000×g, 10 min). Specific MLE proteins and MyoWT were either directly purified on Ni-NTA agarose (50 mM TRIS pH 8.0, 300 mM NaCl, 20 mM imidazole) or extracted in 6 M urea solution (50 mM TRIS pH 8.0, 300 mM NaCl, 20 mM imidazole, 6 M urea) followed by refolding on Ni-NTA agarose during purification.

Thermostability estimated by differential scanning fluorometry

Purified MLE clones dialyzed to PBS were loaded into capillaries, and the thermal stability of protein variants was measured in the Prometheus equipment (NanoTemper). The temperature range was adjusted from 20 °C to 85 °C with an increase of 1 °C/min.

Immunofluorescence staining of human tissue samples

Tonsil and NSCLC tissues were fixed in 10% formalin and paraffin-embedded (FFPE). Next, paraffin blocks were cut into 5 μm sections by microtome and mounted on positively charged glass slides. After deparaffinization and rehydration, the antigen retrieval was performed in an 800 W microwave with two cycles of 3.5 min in 1 mM EDTA (pH 9.0) for antibody Alexa Fluor 488 anti-PD-L1 antibody [28-8] (Abcam, Cambridge, United Kingdom) or citrate buffer (pH 6.0) for antibody PD-L1 IHC 22C3 pharmDx (Agilent Technologies, CA, USA). Tissues were blocked in 5% horse serum (Merck, Darmstadt, Germany). Binders were added in the final amount of 1 µg/100 µl and kept overnight at 4 °C inside the humidified chamber. The next day, MLE proteins were detected by an anti-V5-tag monoclonal antibody, Alexa Fluor 647 (Thermo Fisher Scientific, MA, USA). MLE-rFc binders were detected by an anti-rabbit secondary IgG antibody, Alexa Fluor 555 (Proteintech, Rosemont, USA). Next, Alexa Fluor 488 Anti-PD-L1 antibody [28-8] was applied and left overnight at 4 °C. PD-L1 IHC 22C3 pharmDx was applied for 40 min at 37 °C. A mounting medium with DAPI (Merck, Darmstadt, Germany) was used for the final preparation of the specimen for microscopy. All the sections were scanned using a fluorescent microscope Leica DM6 (Leica Biosystems, Germany) and processed by LAS X Life Science Microscope Software Platform (Leica Biosystems, Germany). Co-localization was quantified using ImageJ software (LOCI, University of Wisconsin) supplemented with the plugin JACoP (Fabrice P. Cordelières, Bordeaux Imaging Center, France).

Immunohistochemistry staining of human tissue samples

Immunohistochemistry was performed on FFPE tumor tissues according to the ESP Lung EQA 2018 scheme recommended protocols for PD-L1 immunohistochemistry (http://lung.eqascheme.org/content/files/public/Protocol_PDL1_22C3_Dako_2).

Results

Selection of PD-L1 targeted MLE variants

To select ligands specific to the PD-L1 protein from the combinatorial library, we performed a ribosome display based on the Myomedin-loop scaffold randomization [21]. We conducted five rounds of display with plate-based selection (P-library), where PD-L1 was immobilized, and three rounds with solution-based selection, where PD-L1 was captured on magnetic beads (M-library). In the case of the P-library, out of 125 individual clones sequenced, 27 had a correct sequence (22%), and 3 clones were duplicated. In the case of M-library, from 131 sequenced clones, 34 clones were correct (26%), and 4 clones were duplicated. All sequences were designated as MLE clones, and those from the M-library were numbered from 1 to 200, while the P-library clones were given numbers from 201 onward. Clones with correct sequences (total 54) were assayed in ELISA. MLE proteins significantly binding to PD-L1 (Supplementary Fig. S1) were purified, and ELISA binding curves were measured (Fig. 1). The most promising candidates from these assays are MLE031, MLE098, MLE105, MLE249, MLE264, MLE270, and MLE309 (sequences in Supplementary Table S1), and their melting temperatures were also determined (Supplementary Table S2).

Fig. 1.

The binding of selected MLE variants to hPD-L1-Fc measured by ELISA. MaxiSorp plate was coated with chimeric recombinant hPD-L1-Fc (1.5 mg/ml) and blocked with Protein-Free (PBS) Blocking Buffer (PFBB). The protein variants in the form of His₆-MLE-V5tag or His₆-MyoWT-V5tag as a negative control were produced in E. coli BL21 strain, purified using Ni-NTA agarose and assayed in ELISA. Anti-V5-tag-HRP conjugated monoclonal antibody was used for the detection of the bound Myomedin variants. Each point is shown as the mean value of triplicates with standard deviation error bars

Verification of the inhibitory potential of selected MLE proteins

As a part of the functional characterization of the generated Myomedins MLE, the ability to compete with soluble PD-1 for binding to PD-L1 was estimated by competition ELISA. In this assay, 7 MLE variants were tested (MLE031, MLE098, MLE105, MLE249, MLE264, MLE270, and MLE309). ELISA was done in the presence of PD-1 (15 nM), which was added to test whether it competes for the binding to PD-L1 with MLE variants. Results suggest that only MLE105 and MLE270 exhibit the inhibitory potential (Fig. 2).

Fig. 2.

Inhibitory potential of MLE variants tested by competition ELISA. Of the 7 tested MLE variants, two of them (MLE105 and MLE270) exhibit an inhibitory effect. For variant MLE105, p-values for points with the three highest concentrations (2000 nM, 667 nM, and 222 nM) are less than 0.0001 (p < 0.0001). For MLE270, p-values for these three concentrations are p = 0.00969 (2000 nM), p = 0.0001 (667 nM), and p = 0.00997 (222 nM). Analysis was done using the ANOVA test (p < 0.05), which was considered statistically significant. In the figure, the green square represents absorbance for PD-1 only, i.e., the positive control, and the purple square corresponds to the signal for BSA as a negative control. PD-1 was applied in 15 nM concentration. Absorbance at 450 nm was detected using an anti-PD-1 antibody and a secondary rabbit anti-goat antibody conjugated with HRP

Binding of MLE proteins to cell-surface expressed PD-L1

To test the specificity of MLE variants, transiently transfected HEK293T cells with a gene coding for PD-L1 (hPDL1-HEK293T) were used. First, the specificity of the anti-PD-L1 antibody was tested on transfected versus mock-transfected cells (Supplementary Fig. S2A). Next, following the same transfection protocol, cells expressing hPD-L1 were treated with 7 MLE variants (together with MyoWT as a negative control), and the immunofluorescence staining on cells expressing PD-L1 was performed. An overlap of immunofluorescence signals for antibodies detecting PD-L1 or V5-tag on the merged images suggests that the binding of MLE variants is specific to human PD-L1. Additionally, the lack of fluorescence signal for the detection of the V5-tag in the case of the MyoWT protein confirms that only Myomedin variants with introduced mutations can bind to PD-L1. All tested MLE variants exhibited binding to human PD-L1 (Fig. 3A).

Fig. 3.

Binding of MLE variants to hPD-L1 on (A) transfected HEK293T and (B) double-stimulated and PFA-fixed MCF-7 cells. Double immunofluorescence staining was performed to detect PD-L1 protein on the cell surface using a polyclonal goat anti-PD-L1 antibody recognized by a secondary anti-goat antibody conjugated with Alexa Fluor 568. For the detection of MLE proteins, a monoclonal antibody against V5-tag conjugated with Alexa Fluor 488 was used. Magnification is 100x

As an independent verification of the PD-L1 specificity of MLE proteins, the MCF7 cell line, which endogenously expresses PD-L1 [24, 25], was tested for PD-L1 expression in various conditions (Supplementary Fig. S3A). Then, to enhance PD-L1 expression, double immunofluorescence staining of 4% PFA-fixed MCF7 cells was performed after 24 h treatment with 100 ng/ml IFN-γ and TNF-α. As shown in Supplementary Fig. S3A, while IFNγ had a supporting effect on PD-L1 stimulation, TNFα substantially increased cell-surface PD-L1 with at least 50 ng/ml concentration or in combination with IFNγ. Then, the binding of 7 MLE proteins to double-stimulated fixed MCF-7 cells was tested, thus further confirming the binding specificity of MLE variants to PD-L1 (Fig. 3B).

In a parallel study, the A549 cell line, described as PD-L1 positive [26, 27], was tested for the presence of PD-L1 by anti-PD-L1 antibody staining. Due to an increased background in staining on live cells, or a very low fluorescence signal on PFA-fixed cells, the A549 cell line was not used in further experiments, although it proved to express hPD-L1 (Supplementary Fig. S3B).

To co-localize hPD-L1 and MLE proteins in cell compartments, staining of 4% PFA-fixed MCF-7 cells was performed on glass slides and used for confocal microscopy. In addition to the antibodies detecting PD-L1 and Myomedins via the V5-tag, which were used for immunofluorescence (Fig. 3B), plasma membrane staining and counterstaining of nuclei were performed. As shown in Fig. 4 and merged images in Supplementary Fig. S4, all tested MLE variants demonstrated co-localization of signal with anti-PD-L1 antibody and membrane stain binding, thus supporting anti-PD-L1 specificity of the developed MLE variants.

Fig. 4.

Detection of hPD-L1 expression and MLE binding on MCF-7 cells fixed with 4% PFA using confocal microscopy. Staining of plasma membrane was done using the stain in which cholesterol is the anchor on the membrane (Abberior, Star Orange). Detection of PD-L1 and V5-tag was performed using goat anti-human polyclonal PD-L1 antibody with secondary donkey anti-goat antibody (Abberior STAR 635P conjugate) and anti-V5-tag antibody conjugated with Alexa Fluor 488, respectively. Magnification is 630x

Binding of MLE proteins to murine PD-L1

For potential use of the developed Myomedin variants for in vivo diagnostics of PD-L1 positive cell populations and tumors, MLE proteins were investigated for binding to the murine version of PD-L1 (mPD-L1). HEK293T cells transfected with the mPD-L1-pcDNA6 vector presented cell expression of mPD-L1 (Supplementary Fig. S2B). Therefore, double immunofluorescence staining for detection of mPD-L1 and Myomedins via V5-tag detection was performed on HEK293T cells transiently expressing mPD-L1. All selected MLE proteins presented binding to the murine version of PD-L1 either when tested on transfected HEK293T cells (Supplementary Fig. S5A) or on fibroblast cells of the NIH-3T3 line described to naturally express mPD-L1 [28, 29] (Supplementary Fig. S5B).

Measuring of binding kinetics of MLE proteins

To estimate the kinetic parameters of MLE proteins, measurements on transfected HEK293T cells expressing the human or murine version of PD-L1 (Fig. 5) were performed. The interaction of fluorescently labelled MLE proteins with the surface of transiently transfected HEK293T cells was measured in real-time mode using the LigandTracer Green line system. On and off-rates (ka and kd values) were evaluated for MLE proteins based on the calculations using TraceDrawer software. While the variant MLE105 exhibited the lowest KD (7.59 nM) to hPD-L1 (Fig.S6A), the slowest off-rate was measured for MLE105 (3.52 × 10^− 4 [1/s]) and MLE249 (6.03 × 10^− 4 [1/s]) (Supplementary Table S3A). Kinetic parameters for murine PD-L1 were tested for MLE105 and MLE249 (Supplementary Table S3A). The summary of experimentally characterized 5 MLE proteins is presented in Table S4.

Fig. 5.

Representative images with binding curves presenting binding of MLE variants to PD-L1 performed using LigandTracer Green. (A-B, D-E and G-H) The assay was done with hPD-L1-HEK293T cells for evaluation of the binding of (A) MLE249, (B) MLE249-rFc, (D) MLE309, (E) MLE309-rFc, (G) MLE031 and (H) MLE031-rFc variants to the human version of the PD-L1. The detection of MLE proteins was done using anti-V5-tag monoclonal antibody conjugate with AlexaFluor 488. (C, F, I) The binding of MLE proteins to the murine version of PD-L1 on transfected mPD-L1-HEK293T cells was monitored by anti-V5-tag monoclonal antibody conjugate with AlexaFluor 488. (I-J) Binding of three monoclonal antibodies to hPD-L1 on transfected HEK293T cells. A fluorescence signal was detected by secondary antibody conjugates of Alexa Fluor 488 (for 28-8 and 22C3) or FITC (for durvalumab)

To allow MLE protein detection in clinical praxis that relies on PD-L1 expression examination using automated systems based on paraffin-embedded tumor tissue sections and established diagnostic procedures developed for rabbit-anti-human PD-L1 monoclonal antibodies, chimeric MLE-rabbit IgG-Fc proteins (His₆-MLE/MyoWT-V5tag-rFc, 42 kDa as a monomer) were generated and optimized for E. coli expression. The binding kinetics of these chimeric MLE variants were assessed using LigandTracer and the binding curves are presented for variants MLE031, MLE249 and MLE309 (Fig. 5B, E, H, Table S3B). Interestingly, estimated KD values for both MLE-rFc demonstrate higher affinity to murine PD-L1 (5.65 and 1.86 nM for MLE031 and MLE249, respectively) (Fig. S6B, C) than to human PD-L1 (60 and 35.5 nM for MLE031 and MLE249, respectively). Additionally, we measured the binding of two diagnostically used antibodies 28-8 and 22C3 as well as the therapeutic antibody durvalumab to human PD-L1 expressed on transfected HEK293T cells using LigandTracer (Fig. S7A, B). In this experimental setup, the estimated KD value for 28-8 is 25.20 nM (Fig. 5, J) and for 22C3, it is 0.04 nM (Fig. 5, K). For the therapeutical monoclonal antibody durvalumab, KD value estimated by LigandTracer is 0.05 nM (Fig. 5, L).

As the MLE105 variant inhibited the PD-1 binding to PD-L1 in the competition ELISA in contrast to other major candidate variants MLE031, MLE249, and MLE309, a competition LigandTracer assay was performed. In this experiment, increasing concentrations of PD-1 (30, 50, and 150 nM) were applied every 30 min into the cell culture medium. An assay was performed on four selected MLE variants (Fig. 6). Out of the tested MLE variants, the MLE105 protein only presented a decrease in the fluorescence signal after the addition of the soluble PD-1 protein, thus confirming the inhibitory potential of MLE105 (Fig. 6).

Fig. 6.

Competition binding assay for MLE variants measured by LigandTracer. The binding of four MLE proteins to cells was monitored in the presence of soluble recombinant PD-1 (depicted as colored lines) or in the absence of PD-1 (presented as black lines). Green arrows show when the association phase measurements were started by the addition of 50 nM of MLE variants. Red arrows indicate time points when increasing concentrations of recombinant hPD-1 protein were added into the culture medium containing constant MLE protein concentration. The dissociation phase was started by the exchange of the culture medium for a fresh one without MLE proteins (pointed by the black arrow). Fluorescence signal rates were normalized to percent values (the highest fluorescence value collected by the end of the association phase was assigned as 100%)

Binding of MLE variants to PD-L1 in paraffin tissue samples

We verified the binding ability of MLE binders to PD-L1 present on cells in tonsil tissue by comparing with anti-PD-L1 antibodies used in the clinics as diagnostic tools, namely anti-PD-L1 antibody [28-8] and PD-L1 IHC 22C3 pharmDx antibody (Fig. 7, Fig. S9). The formalin-fixed paraffin-embedded tonsillitis tissue section was processed according to the protocol recommended for the used antibody, and then co-staining with antibody and MLE binder was performed. In addition to the merged picture by ImageJ software, we also performed an overlap analysis and calculated the Pearson´s coefficient.

Fig. 7.

Immunofluorescence analysis of PD-L1 on FFPE tonsillitis tissue. PD-L1 detection was performed with binders visualized by anti-V5-tag monoclonal antibody Alexa Fluor 647 (red) and (A) Alexa Fluor 488 Anti-PD-L1 antibody [28-8] (abbreviated as 28-8) (green), or (B) PD-L1 IHC 22C3 pharmDx antibody (abbreviated as 22C3) (green). Cell nuclei were stained with DAPI (blue). Image J figures display composite figures showing an overlap between binder and Alexa anti-PD-L1 antibody, where an overlap of both channels is shown as a yellow color. Pearson’s coefficient was calculated by ImageJ software. Values over 0.7 mean excellent, 0.5–0.7 is strong, and 0.3–0.5 is moderate correlation. Magnification is 400x

In Fig. 7A and Fig. S9A, we compared MLE binders with an anti-PD-L1 antibody [28-8], which is approved by the FDA and EMA as a diagnostic tool for OPDIVO treatment (anti-PD-1 therapy) (PMA number P150025, EMA product number EMA/4952/2025). Two binders, MLE309 and MLE249, exhibit excellent (Pearson’s coefficient 0.731 and 0.836) the other three, MLE031, MLE105, and MLE270, have moderate correlation with the anti-PD-L1 antibody [28-8] (Pearson’s coefficient 0.392, 0.326, and 0.457). Next, we compared MLE binders with the PD-L1 IHC 22C3 pharmDx antibody, which is approved by the FDA and EMA for Keytruda treatment (anti-PD-1 therapy) (PMA number P150013, EMA product number EMA/457510/2024) (Fig. 7B, Fig. S9B). Our results show that MLE309 gives excellent (Pearson’s coefficient 0.876), MLE270 and MLE031 moderate (Pearson’s coefficient 0.377 and 0.421), and MLE249 and MLE105 weak correlation (Pearson’s coefficient 0.183 and 0.228). The difference between the overlap rate of MLE249 and anti-PD-L1 antibody [28-8] and MLE249 and PD-L1 IHC 22C3 pharmDx antibody is very likely caused by different antigen retrieval procedures necessary for these antibodies (antigen retrieval at pH 6.0 for PD-L1 IHC 22C3 pharmDx antibody and antigen retrieval at pH 9.0 for anti-PD-L1 antibody [28-8] antibody). Different antigen retrieval has only limited effect on detection ability of other binders, since their Pearson´s coefficients are similar in these treatments. To make our fluorescent staining more credible, we calculated Pearson’s coefficients for three independent regions of tonsil tissue presented in Fig. 7, these Pearson’s coefficients are summarized in Table S5.

As MLE249 is the binder with the best match to the anti-PD-L1 antibody [28-8] and MLE309 is the binder with the best match to the PD-L1 IHC 22C3 pharmDx antibody on tonsillar tissue, we selected them to verify their specificity on tumor tissue (Fig. 8). We chose three NSCLC FFPE tissues, where PD-L1 positivity was proved by IHC with the PD-L1 IHC 22C3 pharmDx antibody. MLE249 exhibited an excellent correlation (Pearson´s coefficient 0.883) with anti-PD-L1 [28-8] antibody for tumor 1, a strong correlation for tumour 2 (Pearson´s coefficient 0.629), and a moderate correlation for tumor 3 (Pearson´s coefficient 0.455). Binder MLE309 exhibited a strong correlation for all three NSCLC tissues (Pearson’s coefficient for tumor 1 is 0.619, for tumor 2 is 0.599, and for tumor 3 is 0.534).

Fig. 8.

Immunofluorescence analysis of PD-L1 on NSCLC tissue. PD-L1 detection was performed with binders visualized by anti-V5-tag monoclonal antibody Alexa Fluor 647 (red) and (A) Alexa Fluor 488 Anti-PD-L1 antibody [28-8] (abbreviated as 28-8) (green), or (B) PD-L1 IHC 22C3 pharmDx (abbreviated as 22C3) (green). Cell nuclei were stained with DAPI (blue). ImageJ figures display composite figures showing an overlap between binder and anti-PD-L1 antibody, where the overlap of both channels is shown as a yellow color. Pearson´s coefficient was calculated by ImageJ software. Values over 0.7 mean excellent, 0.5–0.7 is strong, 0.3–0.5 is moderate, and lower than 0.3 is a weak correlation. Magnification is 400x. Scale bar is 25 μm

Our goal is to develop a more sensitive tool for PD-L1 recognition, which can be used for the refinement of tumor diagnostics. Further, we aimed to modify the best performing MLE binders to adopt their detection to a routine laboratory diagnostics system such as Ventana SP142, Dako 22C3 or 28-8 which rely on secondary antibody usage. Therefore, we fused MLE binders with Fc fragments of rabbit IgG heavy chain, which allowed us to use a secondary anti-rabbit IgG Alexa Fluor 555 antibody conjugate. MLE249-rFc, MLE309-rFc, and MLE031-rFc exhibited strong correlation (0.691, 0.61, and 0.667, respectively), and MLE105-rFc exhibited a low correlation (0.276) with the anti-PD-L1 antibody [28-8] antibody on palatine tonsil tissue (Fig. 9, Fig. S10).

Fig. 9.

Comparison of immunofluorescent staining of PD-L1 on palatine tonsil tissue using MLE249-rFc, MLE031-rFc, MLE105-rFc and MLE309-rFc binders and Alexa Fluor 488 anti-PD-L1 antibody [28-8] (green) (400x magnification). MLE binders were detected using a secondary anti-rabbit IgG antibody, Alexa Fluor 555 (red). Cell nuclei were stained with DAPI (blue). ImageJ software processes a composite figure showing the overlap between the binder and Alexa Fluor 488 anti-PD-L1 antibody [28-8] antibody, where the overlap of both channels is shown as a yellow color. Pearson’s coefficient was calculated by ImageJ. Values over 0.7 mean excellent, 0.5–0.7 is a strong correlation, and lower than 0.3 is a weak correlation. The scale bar is 25 μm

Finally, we proved the three best MLE chimera binders—MLE249-rFc, MLE031-rFc, and MLE309-rFc, for immunohistochemistry staining (Fig. S11). All these binders show clear membrane positivity with light nuclear positivity and a significantly higher number of positive cells in comparison with the PD-L1 IHC 22C3 pharmDx antibody.

Finally, we stained a set of tissues from ten diagnosed NSCLC patients with either adenocarcinoma (ADC) and squamous cell carcinoma (SCC) with varied expression of PD-L1, to compare the reactivity of one of the best candidates MLE249-rFc with diagnostic antibody Anti-PD-L1 antibody [28-8] (Fig. 10). The correlation analysis proved the suitability of MLE249-rFc for further testing as a promising candidate for routine NSCLC PD-L1 diagnostics. The Pearson´s coefficient was over 0.7 (excellent correlation) in four tissues, the other 5 was over 0.5 (very good correlation). Lower correlation coefficient is due to unspecific background staining, which is individual for individual tissues. In the tissue “Tumor 10R” which is negative for PD-L1, we observed only background staining, thus Pearson´s coefficient is almost 0. To deeper characterize MLE149-rFc staining, we counted cells stained positive for antibody [28-8], for MLE249-rFc and double positive cells ([28-8] and MLE249-rFc). We found 6 out of 9 positive NSCLC tissues had more positive cells for MLE249-rFc binder (Table S6). The increased positivity varied between 115 and 149% of positive cells. The other two stained NSCLC tissues had a decrease in positivity match to 59 and 70%. The last positive tissue had almost 96% match, which can be assumed as a full match. No dependence of the positivity on the type of NSCLC, either adenocarcinoma or squamous cell carcinoma, was detected.

Fig. 10.

PD-L1 immunofluorescence staining of adenocarcinoma (ADC) and squamous cell carcinoma (SCC) NSCLC tissues. PD-L1 detection in FFPE NSCLC tissues with MLE249-rFc binder detected using rabbit monoclonal antibody conjugated with Alexa Fluor 647 (red) and anti-PD-L1 antibody [28-8] (abbreviated as 28-8) (green). Cell nuclei were stained with DAPI (blue). Tumor 8R 8431/19B has a significant number of erythrocytes, which are cross-reacting with both 28-8 antibody and MLE249-rFc binder. The Tumor 10R is an example of a PD-L1 negative tumor tissue. ImageJ software processes a composite figure showing the overlap between the binder and Alexa Fluor 488 anti-PD-L1 antibody [28-8] antibody, where the overlap of both channels is shown as a yellow color. Pearson’s coefficient was calculated by ImageJ. Values over 0.7 mean excellent, 0.5–0.7 is a strong correlation, and lower than 0.3 is a weak correlation. Scale bar is 25 μm

Discussion

One of the ways to improve patient response to anti-cancer treatment is the application of therapy that targets immune checkpoint proteins PD-L1 and PD-1. Before doing so, however, it is crucial to evaluate whether immune checkpoint blockade therapy will be of benefit to the patient [30]. PD-L1 IHC tests routinely used for this purpose, especially in the case of NSCLC, may be challenging in the interpretation. Therefore, small proteins targeting PD-L1 may be applied as an alternative to diagnostically used monoclonal antibodies and also may be used for imaging [31]. However, antibodies vary in their detection platforms, interpretation criteria, and corresponding positivity thresholds, leading to heterogeneity in the detection of PD-L1 expression [32–34]. Such results may arise due to variations in antibody affinity, restricted specificity, or differences in target epitopes [35]. Therefore, alternative approaches being applied to identify high-affinity small binding proteins are screening strategies such as a yeast two-hybrid screening assay for affibody binders [36], phage display for obtaining affibody binders [37] or nanobodies [38], yeast display for identification of fibronectin III domain binders [39, 40] or high-affinity PD-1 variants [41], mRNA display for identification of SPAM peptide (18 amino acids linear peptide binding to PD-L1) [42], and adnectins [43–46]. There are also peptides [¹⁸F]AlF-NOTA-PCP2 [47] and [⁶⁸Ga]Ga-AUNP-12 [48] used as radiotracers in PET imaging. Although, in many studies, the main aim of developing small binding proteins is for therapeutic purposes, due to the features of such binders, similar screening techniques may be utilized for developing proteins for diagnostics.

Immunohistochemistry has proven to be an effective method in evaluating the level of expression of PD-L1 in lung cancer. Several immunohistochemistry companion diagnostic assays have been released, and they play a crucial role as predictive markers. PD-L1 checkpoint blockade therapy (ICB) utilizing antibodies pembrolizumab, atezolizumab, and nivolumab has been approved for indications with requirements for IHC companion testing: PD-L1 IHC 22C3 pharmDx by Agilent/Dako, PD-L1 (SP142) IHC by Roche Tissue Diagnostics, and PD-L1 IHC 28-8 pharmDx by Agilent/Dako for each ICB, respectively. Furthermore, durvalumab was also approved with an IHC complementary test PD-L1 IHC SP263 by Roche Diagnostics [18]. All these tests exhibit different PD-L1 sensitivities, which are described as TPS scores. This calculation provides critical information that can be applied to guide the use of immunotherapy, which can greatly impact patients with NSCLC [49]. Interestingly, we measured kinetic parameters for 28-8, 22C3 diagnostic as well as therapeutic durvalumab antibodies using PD-L1 transfected HEK293T cells. For 28-8 antibody, our estimated KD is 25.20 nM, that is comparable to our MLE variants, i.e. KD value of 18.80 nM for MLE249. However, Lawson et al. [50] published a KD value of ≤ 6 pM for this antibody measured to recombinant PD-L1 by SPR technique. Regarding the 22C3 antibody, our estimated KD value of 40 pM by LigandTracer which is in a good correlation to the published value ≤ 40 pM by Lawson et al. [50]. For durvalumab, our estimated KD value is 50 pM while lower affinity (KD = 667 nM) for this antibody was reported by SPR [51, 52]. Other studies also demonstrated a difference in KD values for durvalumab binding to glycosylated (0.2 nM) or non-glycosylated (0.4 nM) PD-L1 [53, 54].

Meta-analysis of clinical trials and laboratory-developed assays comparing results between the 28-8, 22C3, SP142, and SP263 antibodies demonstrated high concordance, except for SP142 [19]. The differences between antibodies are likely secondary to laboratory protocol designs and their implementation rather than to the innate performance of the antibody/antigen interactions [19, 55–58]. Our work aimed to develop a useful tool that will make the diagnosis of PD-L1 positivity more accurate. The diagnosis of PD-L1 positivity is expressed as a TPS, which is for PD-L1 IHC 22C3 pharmDx more than 1% and for anti-PD-L1 [28-8] antibody more than 50% (PMA number P150013, EMA product number EMA/457510/2024).

We confirmed that selected MLE variants can identify PD-L1 molecules both in ELISA setup and in tissue sections form human tonsils and human tumor sections particularly in NSCLC using immunofluorescence setup. We observed variation in overlap of MLE variants reactivity and diagnostic antibody reactivity which let us select the best candidates. Four MLE candidates were further modified by fusion with rabbit immunoglobulin heavy chain constant region to allow their detection by secondary antibodies used in routine immunohistochemistry staining systems. Although the use of our identified MLE binders in immunohistochemistry diagnostics and prediction of PD-L1-targeting immunotherapy is currently limited, particularly due to established binding of companion or complementary diagnostics with individual therapeutic antibodies approved by EMA and FDA, the binders’ price and ease of production in E. coli seems to be optimistic for future applicability. From the presented results MLE seems very useful both for immunohistochemistry and immunofluorescence. We tested one of the best variants - MLE249-rFc for immunofluorescence staining on 10 NSCLC tissue samples. MLE249-rFc exhibited great overlap with diagnostic anti PD-L1 antibody [28-8]. We characterized the overlap by Pearson´s coefficient and by manual counting of positive cells, which is more exact, since the background is not included into comparison. Our data shows higher sensitivity of MLE249-rFc in 6 out of 9 PD-L1-positive NSCLC tissues, the same sensitivity of binder and antibody in one tissue and lower sensitivity of binder compared to antibody in 2 tissues.

Based on immunofluorescence analysis, we can assume that our MLE binders can identify PD-L1 molecules that are recognized by the antibodies, plus additional PD-L1 molecules that have escaped the antibody recognition. We hypothesize that the lower recognition sensitivity of antibodies observed in the majority of tissue section analyzed may be due to the different expressions and conformation of PD-L1 on different cells which could be caused by heterogenous post-translational modifications of PD-L1, including N-glycosylation, phosphorylation, or acetylation, demonstrated to regulate PD-L1 stability, expression level and subcellular localization [59]. As the antibodies recognize the entire extracellular domain, its conformation is critical for binding. In addition, antibody specificity/sensitivity, and we could expect similar effect on MLE, is closely related to the presence of glycans on proteins, as they could partially or completely shield recognized epitopes [60]. Our MLE proteins are smaller molecules and thus can eventually penetrate the glycan network and bind directly to the peptide independent of its conformation.

Although convincingly promising, the use of MLE binders in diagnostics must be further verified in large cohorts of patients followed for a period allowing for the assessment of the clinical response to PD-L1 blocking therapy. Since the production of binders is significantly cheaper compared to monoclonal antibodies, we assume that our binders have the potential to become a cheap and accessible diagnostic tool, and thanks to its higher sensitivity, it can help to recognize more PD-L1 positive patients suitable for anti-PD-L1 targeted therapy.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

H.P. assembled Myomedin combinatorial library, performed ribosome display, large scale screening of Myomedin variants and binding analysis by ELISA, processed data and wrote the paper. J.M.M. analyzed MLE proteins by fluorescent microscopy, confocal microscopy, performed kinetic assessments by LigandTracer and wrote the paper. M.K. constructed MLE chimeric vectors, produced MLE chimeras and analyzed them by ELISA. P.C. performed immunofluorescence analysis of tonsillar and cancer tissue. L.R.K performed immunohistochemical evaluation on tissue sections, analyzed data and wrote the paper. J.V. and J.S. performed the immunohistochemistry of tonsil tissue. P.K. characterized MLE proteins and chimeric variants and analyzed data. O.F. provided and characterized tissue specimens from diagnosed NSCLC patients. P.M. and M.R. conceptualized the project, directed research, designed research, analyzed data, and wrote the paper.

Funding

The authors are thankful for the received support by Czech Health Research Council, Ministry of Health of the Czech Republic by the grant No. NU21-03-00372, the Institute of Biotechnology of the Czech Academy of Sciences v.v.i. (Institutional Research Concept, RVO: 86652036) and University Hospital Olomouc (Conceptual development of research organization grant MHCZ-DRO, FNOL, 00098892.

Data availability

All data generated or analyzed during this study are included in this article and its Supplementary Information/Source Data file. Source data are (will be) provided with this paper.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

All authors have read the journal’s policy on disclosure of potential conflicts of interest and declared no conflict of interest.