Quantification of increased biologically active CXCL12α plasma concentrations after ACKR3 antagonist treatment in humans

Abstract

CXCL12 acts as a chemoattractant by binding to the receptor CXCR4. The (atypical) chemokine receptor ACKR3 (CXCR7) scavenges CXCL12. Antagonism of ACKR3 thus leads to an increase in CXCL12 concentrations that has been used as a pharmacodynamic biomarker in healthy adults. Increased CXCL12 concentrations have also been linked to repair mechanisms in human diseases and mouse models. To date, CXCL12 concentrations have typically been quantified using antibody‐based assays with overlapping or unclear specificity for the various CXCL12 isoforms (α, β, and γ) and proteoforms. Only the N‐terminal full‐length CXCL12 proteoform is biologically active and can engage CXCR4 and ACKR3, but this proteoform could so far not be quantified in healthy adults. Here, we describe a new and fit‐for‐purpose validated immunoaffinity mass spectrometry (IA‐MS) assay for specific measurement of five CXCL12α proteoforms in human plasma, including the biologically active CXCL12α proteoform. This biomarker assay was used in a phase I clinical study with the ACKR3 antagonist ACT‐1004‐1239. In placebo‐treated healthy adults, 1.0 nM total CXCL12α and 0.1 nM biologically active CXCL12α was quantified. The concentrations of both proteoforms increased up to two‐fold in healthy adults compared to placebo following drug administration. At all dose levels, 10% of the CXCL12α was the biologically active proteoform and the simultaneous increase of all proteoforms suggests that a new steady state has been reached 24 h following dosing. Hence, this IA‐MS biomarker assay can be used to specifically measure active CXCL12 proteoform concentrations in clinical trials to demonstrate target engagement and correlate with clinical outcomes.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Clinical trials usually use antibody‐based assays to measure CXCL12 for simplicity, high‐throughput and low costs. CXCL12 exists in various isoforms and proteoforms with different biological activity. However, antibody‐based assays lack specificity to quantify specific forms of CXCL12 and assays to measure endogenous levels of active CXCL12 were lacking.

• WHAT QUESTION DID THIS STUDY ADDRESS?

At which concentrations is biologically active CXCL12 present in human plasma? Is active CXCL12, a pharmacodynamic biomarker of ACKR3 antagonism, increased upon treatment with the selective ACKR3 antagonist ACT‐1004‐1239, in humans?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

A new assay was developed and used in a phase I clinical trial. This trial shows that only 10% of total CXCL12 in human plasma is biologically active and that ACKR3 antagonism increases dose‐dependently biologically active CXCL12.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

The developed assay enables measuring endogenous CXCL12 proteoform concentrations in clinical trials. To understand the pharmacology of ACKR3 antagonism, it was important to ensure that the biologically active CXCL12 is increased upon drug treatment.

INTRODUCTION

Chemokines are small, secreted proteins that act as a chemoattractant for cells expressing the corresponding chemokine receptors. Currently around 50 chemokines and 24 receptors are known and play a crucial role in human development, homeostasis, inflammation, and tumorigenesis. ¹ Target engagement and pharmacodynamic biomarkers support the development of drugs and are important to understand the biological response following pharmacologic intervention, in this case, following perturbation of the chemokine system. The presence of various forms of post‐translationally modified chemokines with different biological activity renders the measurement and interpretation of such biomarkers challenging.

CXCL12 (also known as stromal cell‐derived factor 1) is a widely expressed homeostatic chemokine attracting leukocytes, stem, and cancer cells expressing the receptor CXCR4. ² The G‐protein coupled receptor CXCR4 was initially discovered as a cofactor facilitating entry of human immunodeficiency virus (HIV) into T‐helper cells and is widely expressed on hematopoietic stem cells, leukocytes, endothelial cells, neurons, microglia, and astrocytes. ³ CXCR4 is a “typical” chemokine receptor that mediates chemotaxis along CXCL12 gradients via GPCR signaling and thus, CXCL12 is a potent chemoattractant for multiple immune cells including primary human T cells. ⁴ CXCL12 is the only endogenous ligand for CXCR4 and it also binds to the atypical chemokine receptor ACKR3 (also called CXCR7) with reported higher affinity than to CXCR4 ⁵ (Figure 1a). ACKR3 is an atypical receptor mainly expressed on endothelial cells that does not signal via G‐proteins but can recruit β‐arrestin after agonist binding. ⁶ Additionally, internalization without β‐arrestin recruitment has been reported. ⁷ Consequently, ACKR3 is believed to act as a scavenger receptor, internalizing and degrading CXCL12 that has been secreted by astrocytes or stromal cells, in particular fibroblasts. ACKR3 thus shapes the concentration gradient of CXCL12 that guides the migration of CXCR4‐expressing cells. ⁸ The CXCR4/ACKR3‐CXCL12 axis has been implicated in various human pathologies. ⁹ Furthermore, pharmacologically antagonizing ACKR3 in mouse models using ACT‐1004‐1239 showed an increase in CXCL12 levels, reduced migration of CXCR4‐positive immune cells and increased the number of oligodendrocytes and myelination. ⁸ Hence, ACKR3 is a promising drug target to modulate immune cell trafficking in immunology indications. ⁸ As CXCL12 levels increase following ACKR3 antagonism, this was used as a pharmacodynamic biomarker both in preclinical animal models ⁸ , ¹⁰ and first‐in‐human studies of the ACKR3 antagonist ACT‐1004‐1239 using an antibody‐based enzyme‐linked lectin assay (ELLA). ¹¹ , ¹² In humans, CXCL12 plasma concentrations increased up to twofold following ACKR3 antagonism.

FIGURE 1.

Functional characterization of CXCL12α proteoforms (proteoforms are labeled using the first and last amino acid). (a) Scheme of the chemokine CXCL12 interacting with the chemokine receptors ACKR3 and CXCR4 as well as the protease DPP4. Inhibition of ACKR3 by our selective antagonist ACT‐1004‐1239 is depicted. (b) β‐arrestin recruitment with different CXCL12α proteoforms by ACKR3 (n = 2 and error bars depict standard deviation). (c) Internalization of various CXCL12α proteoforms by ACKR3 (n = 5 and error bars depict standard error of the mean). (d) Migration of human T cells in response to various CXCL12α proteoforms (n = 1). (e) CXCR4 calcium mobilization in response to various CXCL12α proteoforms (n = 2 and error bars depict standard error of the mean). N‐terminally intact forms of CXCL12α are depicted in green, N‐terminally truncated proteoforms of CXCL12α in blue.

CXCL12 can be expressed as at least three different major isoforms (CXCL12α, β, and γ) that differ at their C‐terminus. ² The physiological roles of the different isoforms of CXCL12 are largely unclear, but the differing positive charge in the C‐termini are believed to determine differential binding to negatively charged glycosaminoglycans and thus influence formation of CXCL12 gradients. CXCL12α is the most abundantly expressed and studied isoform. Moreover, CXCL12 can be processed post‐translationally into different proteoforms by protease‐mediated cleavage or other modifications. ² , ¹³ The protease DPP4 is known to rapidly cleave the first two N‐terminal amino acids leaving CXCL12 inactive. ¹⁴ , ¹⁵ Furthermore, CXCL12α is known to be cleaved at the C‐terminus by carboxypeptidases and studies have shown that this reduced its activity. ¹⁶ , ¹⁷ Other proteases have been reported to cleave CXCL12α at the same or additional sites. ² However, measuring the concentration of individual CXCL12 proteoforms is challenging and requires highly specific analytical methods. Coupling immunoaffinity enrichment with top‐down (i.e., direct quantification of the whole protein) mass spectrometry was used to quantify spiked CXCL12α^1‐67 and CXCL12α^3‐67 in human, rhesus monkey, and murine plasma. ¹⁸ , ¹⁹ , ²⁰ It was shown that the DPP4 inhibitor sitagliptin increases concentration of CXCL12α^1‐67 in human plasma within 3 h and leads to a concomitant reduction in the concentration of CXCL12α^3‐67. ¹⁸ Notably, the concentration of active CXCL12α^1‐67 was much lower than of the inactive CXCL12α^3‐67 and it could not be detected at baseline, highlighting the challenging analytical quantification of CXCL12α^1‐67. To our knowledge, no assay has been yet described with sufficient sensitivity to specifically measure active N‐terminal full‐length CXCL12α^1‐67 at baseline levels in human plasma.

To understand whether the active proteoform of CXCL12α that is required for the pharmacological effect increases in CXCL12 upon ACKR3 antagonism, ¹¹ , ¹² we established a fit‐for‐purpose biomarker assay with sufficient sensitivity to quantify CXCL12α proteoforms in human plasma at baseline levels. This assay was then used to measure CXCL12α proteoform concentrations in plasma samples from a study assessing ACKR3 antagonism in healthy adults administered ACT‐1004‐1239. ¹²

METHODS

β‐arrestin and receptor internalization assay for ACKR3

Tango CXCR7‐bla U2OS cells were used for the β‐arrestin assay as recommended by the manufacturer (Thermo Fisher Scientific, K1832). ACKR3 internalization was tested in an assay based on diffusion‐enhanced resonance energy transfer between cell surface GPCRs labeled with a luminescent terbium cryptate donor and a fluorescein acceptor present in the assay buffer as described by Levoye et al. ²¹

Human primary T cell migration assay

Blood from healthy donors was collected in EDTA‐containing blood collection tubes, and T lymphocytes were isolated using MACSxpress Cell Isolation Kits (Miltenyi, human pan T Cell, #130‐098‐193). The choice of isolation method is crucial for studying the CXCR4 receptor in vitro as in‐house data have shown that lymphocytes isolated with MACSxpress preserve CXCR4 receptor surface expression in contrast to density gradient isolation (unpublished observation). The migration assay was performed using 101‐3 plates (Neuroprobe) as per the manufacturer's instruction. Briefly, 150,000 Calcein‐AM labeled T cells in RPMI supplemented with 10% FBS were seeded on top per well, the final concentration of CXCL12 proteoforms in the lower chamber was 12 nM (also in RPMI with 10% FBS). Cell number in the bottom chamber were counted after 1 h incubation at 37°C, 5% CO₂.

Pulldown of CXCL12 from plasma

To prepare quality control (QC) samples, plasma was either diluted or spiked with synthetic CXCL12^1–68 and CXCL12^3–67 (Bachem, Bubendorf, Switzerland) containing the correct formation of disulfide bonds. The study or QC samples (250 μL) were then diluted with 250 μL PBS and incubated on a rotating wheel for 2 h at room temperature with biotinylated anti‐CXCL12 polyclonal goat antibody (#BAF310, R&D Systems) at a concentration of 4.6 μg/mL. Streptavidin magnetic beads (25 μL of 10 mg/mL; 25 μL #650.01, Invitrogen) were added to the samples and incubated on a rotating wheel for a further 30 min at room temperature. The beads were pulled down with a magnet for 3 min and the supernatant was removed and kept at −80°C to assess pulldown efficiency. The beads were then washed three times with 1 mL PBS on the magnet. After the last wash, the beads were re‐suspended in 100 μL IgG Elution buffer, pH 2.8 (#21004, Thermo Scientific) and incubated for 30 min on a shaker. The beads were then pulled down with a magnet for 3 min, before the eluate was transferred into new Protein LoBind tubes 0.5 mL (022431064, Eppendorf, Hamburg, Germany) and neutralized with 25 μL 1 M Tris‐HCl, pH 8.0 (Gibco, #15568‐025, Thermo Scientific). The eluates were stored at −80°C until further processing. Pulldown efficiency was assessed using the Mesoscale U‐plex assay for mCXCL12 (MSD K152VBK‐1 with the calibrator from human DuoSet, DY350, R&DSystems). The accuracy of using a human CXCL12 for the mouse assay has been tested (human and mouse CXCL12 only differ by one amino acid).

Processing of pulled down CXCL12 to peptides

The eluates were thawed for 20–30 min at room temperature and 55 μL samples, calibration standards (CALs), QC samples or blank were transferred to a Protein LoBind 96 well plate (0030 504.100, Eppendorf, Hamburg, Germany). To denature the samples, 79 μL denaturing buffer was added, resulting in a concentration of approximately 3.4 M urea, 5 mM TCEP, and 1 μg carrier protein (E. coli lysate), and a heavy isotopologue of the CXCL12^1–67 proteoform (ThermoFisher Scientific) was used as an internal standard. Samples were incubated for 30 min at 37°C before alkylating the cysteines with 10 mM iodoacetamide for 45 min at 25°C in the dark. Samples were split in two and digested with 5 ng/μL trypsin or 2.5 ng/μL GluC overnight at 37°C. Digestion was stopped by acidifying with TFA to a pH of 2–3. Then labeled peptides (ThermoFisher Scientific) were added as additional internal standards, and a solid‐phase extraction was performed on HLB μElution plates (18600182BA, Waters). Desalted peptides were resuspended in 2% ACN, 0.05% TFA in water containing iRT peptides (Ki‐3002‐2, Biognosys, Schlieren, Switzerland) before liquid chromatography mass spectrometry measurements.

Acquisition of peptide samples on the mass spectrometer

Samples were injected on an Orbitrap Exactive HF‐X (Thermo Scientific) mass spectrometer connected to a Dionex Ultimate HPLC (Thermo Scientific) system. The injected sample (10 μL) was first loaded on a trap column (Acclaim PepMap 100; 75 μm × 2 cm; C18; 3 μm; 100 Å (P/N: 164535) at 32°C for 5 min with 0.1% FA in water at 6 μL/min before switching the trap column in the nano‐flow path. Peptides were then separated on a PepMap RSLC C18; 75 μm x 15 cm; 2 μm; 100 Å (P/N: ES804A) at 40°C using a gradient from 5% Buffer B (ACN/water/FA 80:20:0.1, v/v/v) in Buffer A (water/FA 100:0.1, v/v) to 35% Buffer B for 10 min at 300 nL/min before washing and equilibrating the column again. The mass spectrometer was operated in positive ion mode cycling over a Full MS scan for monitoring purposes, followed by the PRM scans. The Full MS scans were performed at 60 k resolution, using an AGC target of 3e6, maximum injection time of 120 ms, and a scan range of 300–1000 m/z. The scheduled PRM scans were performed at 30 k resolution at 200 m/z, with an AGC target of 2e5, maximum IT of 50 ms and 1.4 m/z isolation window. In total, 20 precursors were selected for fragmentation in the trypsin method and 17 in the GluC method.

Data analysis

The raw files were imported into the software Skyline (MacCoss Lab, University of Washington) for data extraction from the acquired spectra. Up to five product ions of each precursor ion were selected for quantification and the data was subsequently exported and analyzed using an R Markdown script. The integrated signal for each endogenous isotopologue was normalized to the internal standard (signal of the heavy isotopologue) and the concentration was determined based on the calibration curve produced with standards consisting of 10 CAL samples with different amounts (0.059–8.230 nM) of an equimolar mix of synthetic CXCL12^1–68 and CXCL12^3–67 in a surrogate matrix of 0.001% BSA in water. The regression was performed using a quadratic fit (y ~ ax ² + bx + c) and a 1/x ² weighting. On average, the back‐calculated concentrations of the calibration standards had an accuracy of 5.5%. For all peptides, it was possible to quantify the resulting peptides (working range was 0.0295–4.14 nM for the terminal peptides and twice the concentrations for central peptides) across all CAL samples, except for the peptide YLEKALNK that needed to be quantified above a concentration of 0.059 nM.

The analysis of the N‐terminal cleavage (Figure 2c) was performed before the assay was set up completely and the intensity of the light isotopologue was normalized by the mean intensity of the three central peptides.

FIGURE 2.

Immunoaffinity‐mass spectrometry assay to measure CXCL12α proteoforms. (a) Scheme of sample processing and acquisition. (b) Depiction of which surrogate peptides are quantified for which proteoform (labeled in different color). (c) Relative ratio of the peptide signal from blood sampled with either BD P800 tubes (EDTA‐P800) or conventional EDTA tubes (EDTA‐PPT) from five independent experiments and at least eight different donors.

Fit‐for‐purpose validation

To assess accuracy and precision of the immunoaffinity mass spectrometry (IA‐MS) assay, four QC sample levels were used based on diluted (QC1: 1:1 dilution of QC2), neat (QC2: pool of plasma from 5 healthy donors), and spiked human plasma (QC3 and QC4: QC2 spiked with 10 ng/mL and 20 ng/mL respectively of both synthetic CXCL12α^1–68 and CXCL12α^3–67).

Furthermore, carry‐over was assessed by analyzing the intensity of the signal in blank injections (0.001% BSA without internal standards but processed analogously) after the highest calibration standard and QC sample. Mean carry‐over was 0.36% for endogenous and spiked heavy isotopologues as well as 0.12% for the endogenous isotopologue. Thus, samples with large concentration differences can be measured sequentially without the need to inject blank samples. Additionally, various stability assessments were performed. Keeping plasma for 1 h at the room temperature, frozen for 6 months, or performing freeze–thawing did not result in degradation of a proteoform (Figure S1A–C). Whereas the immunocapture eluate was stable when frozen for additional 2 months (Figure S1D), additional freeze–thawing of the immunocapture resulted in approximately 30% loss of signal both after one and two cycles (Figure S1E,F). Re‐injectability of processed samples (i.e., peptides) both after three freeze–thaw cycles and storage in the autosampler at 10°C for 4 days showed no sign of degradation (Figure S1G,H). Moreover, no substantial donor‐specific matrix effect could be observed when assessing 1:1 dilutions of plasma from four individual donors (Figure S1I).

Clinical samples

Blood samples for assay development and validation were collected from healthy adults under the approval of the Ethics Committee of Northwestern and Central Switzerland (EK179/11 and EK249/02). The study samples were collected in the first‐in‐class ACKR3 antagonist ACT‐1004‐1239 multiple ascending dose phase I clinical study where adults were dosed at 30 mg, 100 mg, and 200 mg once daily (ClinicalTrials.gov: NCT04286750). ¹² Blood was collected in EDTA BD P800 tubes to prevent cleavage of the N‐terminus of CXCL12 and plasma was generated.

Measurement of clinical samples

The samples were analyzed in four batches using a block‐randomized approach and bracketed by CAL samples. As QC samples, three QC2 samples were included in each batch to control for assay performance. The mean intra‐batch precision of the QC samples had a coefficient of variation (CV) of 6.2% and all proteoforms had less than 10% CV, confirming the high precision of the assay (Figure S2A). The mean inter‐batch precision was 7.7% and all proteoforms and QCs showed less than 15% CV (Figure S2B). The accuracy was assessed by comparing the concentrations of the QC samples to the empirical concentration obtained during the validation of the assay. No sample in any batch had more than 20% deviation (Figure S2C). Hence, the results were precise and accurate within the predefined limits. Results on peptide level for QC and study samples are shown in Figure S3.

Cross‐reactivity experiments

Cross‐reactivity was investigated for the antibody‐based assays from ELLA (Biotechne SPCKB‐PS‐000299 CXCL12) and Mesoscale (see above). Concentration curves of four different proteoforms containing the correct formation of disulfide bonds (CXCL12α^1–68, CXCL12α^1–67, CXCL12α^3–67, CXCL12α^5–67; Bachem, Bubendorf, Switzerland) were prepared in the corresponding assay diluent and measured with the antibody‐based assay. Concentrations investigated ranged from 0.3 to 20 ng/mL for ELLA and 5 pg – 20 ng/mL for Mesoscale. The assay cross‐reactivity was calculated for the concentrations 5 ng/mL and 20 ng/mL using CXCL12α^1–67 as 100%.

RESULTS

Variable biological activity of the CXCL12α proteoforms

The proteoforms CXCL12α^1–67, CXCL12α^3–67, and CXCL12α^5–67 (proteoforms are labeled using the first and last amino acid) were tested for their capability to engage the CXCR4 and ACKR3 receptors. Only the N‐terminal full‐length CXCL12α^1–67 was able to induce ACKR3‐dependent β‐arrestin recruitment in recombinant cell lines (Figure 1b) and become internalized by ACKR3 (Figure 1c). No β‐arrestin recruitment or internalization were observed for the N‐terminally truncated proteoforms. Furthermore, the ability of the different CXCL12α proteoforms to attract freshly isolated human peripheral T cells was tested in a modified transwell migration assay. Only the N‐terminal full length CXCL12α^1–67 was able to induce chemotaxis above background levels (Figure 1d). Finally, the different CXCL12α proteoforms were tested in a calcium mobilization assay in a CXCR4 expressing cell line, which tests for G‐protein mediated signaling. Again, only the N‐terminally intact form of CXCL12α showed activity (Figure 1e). In conclusion, cleavage of the first two or four N‐terminal amino acids of CXCL12α results in loss of its agonistic properties and hence biological activity on both ACKR3 and CXCR4.

Immunoaffinity‐mass spectrometry biomarker assay to quantify CXCL12α proteoforms

To quantify CXCL12 proteoforms at baseline levels in human plasma, an IA‐MS assay was developed (Figure 2a). The average immunoaffinity depletion for the samples was 99% (range: 97.6% to 99.4%). To quantify both the N‐ and C‐termini in such an assay, immunoprecipitated CXCL12 was digested separately with the proteases trypsin and GluC. Digestion with trypsin generated N‐terminal and central peptides, whereas GluC generated N‐terminal and C‐terminal peptides (Figure 2b). From the generated peptides, nine were quantified as surrogate peptides for five different proteoforms (CXCL12^total, CXCL12^1‐xx, CXCL12^3‐xx, CXCL12^x‐67, and CXCL12^x‐68). Each proteoform concentration was thus inferred by one to three independent peptide concentrations (Figure 2b). Calculating the difference in abundance between the quantified total CXCL12 concentration and the sum of the terminal proteoforms (N‐terminal or C‐terminal respectively), the remaining concentration of “unaccounted” N‐ or C‐terminal proteoforms could be calculated (CXCL12^unk‐xx for CXCL12 with an unaccounted N‐terminus and CXCL12^x‐unk for CXCL12 with an unaccounted C‐terminus). Thus, concentrations of five directly measured (CXCL12^total, CXCL12^1‐xx, CXCL12^3‐xx, CXCL12^x‐67, and CXCL12^x‐68) and two calculated (CXCL12^unk‐xx and CXCL12^x‐unk) proteoforms were quantified (Figure 2b).

Rapid N‐terminal cleavage of CXCL12 in blood drawn into conventional PPT tubes

DPP4 is the main protease known to cleave the first two amino acids from CXCL12 at the N‐terminus. To assess if CXCL12 might be cleaved after blood draw by proteases, we sampled in parallel blood from healthy adults into EDTA BD P800 tubes or conventional EDTA PPT tubes and saw similar signals for all proteoforms except for the N‐terminal full‐length CXCL12^1‐xx (Figure 2c). The rapid N‐terminal truncation of CXCL12 in blood has important consequences for pre‐analytical handling of human blood for measurement of CXCL12α^1–67. To ensure that CXCL12 is not further cleaved after the blood draw, BD P800 tubes containing a mix of protease inhibitors were used.

Context‐of‐use and fit‐for‐purpose validation of IA‐MS biomarker assay

Before quantifying CXCL12 proteoforms in samples from a clinical trial, a fit‐for‐purpose assay validation using predefined acceptance criteria adapted to the complexities of the assay was performed. ²³ , ²⁴ , ²⁵ The assay had high intra‐ and inter‐batch precision across four validation batches with a mean CV of 6.8% and 17% across all QCs, respectively (Figure 3a,b). No proteoform had a higher CV than 15% or 30%, respectively. Absolute accuracy could not be determined as human plasma contains CXCL12. However, a nominal concentration based on the mean back‐calculated concentration of the QC2 samples (neat plasma) was calculated: [QC1] = [QC2]/2; [QC3,4] = [QC2] + [spike]. Comparing the measured concentrations against the calculated nominal concentration showed some batch‐to‐batch variation but an acceptable accuracy (<±30% in 2 out of 3 QC samples) was reached for all proteoforms except for CXCL12^x‐67 and CXCL12^x‐68 (Figure 3c,d). For the C‐terminal proteoforms CXCL12^x‐67 and CXCL12^x‐68 substantially higher or lower measured values were obtained, respectively, suggesting that the spiked full‐length CXCL12α^1–68 was rapidly cleaved at the last amino acid. Interestingly, no cleavage was observed of the endogenous C‐terminal full‐length CXCL12, showing that the endogenous CXCL12 behaved differently than the synthetic CXCL12α. Furthermore, the repeatability was assessed by comparing the measured proteoform concentration against the empirical value from all batches (Figure 3e). This indicated some batch‐to‐batch variation but two out of three values were within 30% of the empirical concentration for all proteoforms, except for the CXCL12^x‐68 proteoform measurements in the last batch.

FIGURE 3.

Fit‐for‐purpose validation of MS‐based assay. (a) Intra‐batch precision of proteoforms. (b) Inter‐batch precision of proteoforms. (c) Relative accuracy to nominal concentration across batches. (d) Average relative accuracy to nominal concentration. (e) Relative accuracy to empirical concentration. CV, coefficient of variation; MS, mass spectrometry.

In summary, the predefined criteria were met and the fit‐for‐purpose validation was successful (see also Methods). The proteoforms were stable in various stress tests, and only when performing freeze–thaw cycles of the immunocapture eluate, a loss of signal of about 30% was observed at both one and two freeze–thaw cycles.

Measurement of CXCL12α proteoforms in a clinical trial

In the multiple ascending dose study with the ACKR3 antagonist ACT‐1004‐1239 (NCT04286750), ¹² plasma was collected in BD P800 EDTA tubes on day 2 and day 8 at trough from 30 adults. In the 59 samples, CXCL12 proteoforms were quantified. The results indicated a clear dose‐dependent increase of all CXCL12 proteoforms from placebo to the highest dose (200 mg) both at day 2 and day 8. An exception was CXCL12^x‐unk that was not detected at significant levels showing that the CXCL12 proteoforms measured were predominantly of the CXCL12α isoform (i.e., ending at residue 67 or 68). The total CXCL12α concentration in placebo‐treated adults was, on average, 1.0 nM, whereas the concentration of CXCL12α^1‐xx, CXCL12α^3‐xx, CXCL12α^unk‐x, CXCL12α^x‐67, CXCL12α^x‐68, and CXCL12α^x‐unk was 0.10 nM, 0.21 nM, 0.68 nM, 0.88 nM, 0.08 nM, and 0.04 nM, respectively (Figure 4a). Expressed as fold‐change to placebo at the same timepoint, the total CXCL12 increased at the highest dose by 2.36‐fold at steady‐state condition (day 8); the increases in CXCL12α^1‐xx, CXCL12α^3‐xx, CXCL12α^unk‐x, CXCL12α^x‐67, and CXCL12α^x‐68 were 2.58, 2.33, 2.34, 2.40, and 1.74‐fold, respectively, indicating a parallel increase of the concentrations of all proteoforms (Figure 4b). As the CXCL12^x‐unk concentrations were not significantly different from 0, no fold‐change was calculated. The relative amount of biologically active CXCL12α^1‐xx to total CXCL12 concentration was stable at around 10% across the different doses (Figure 4c).

FIGURE 4.

Quantification of CXCL12 proteoforms in clinical study. (a) Quantified proteoform concentrations in plasma samples across placebo and three doses of ACT‐1004‐1239 after the first dose (day 2) or seven doses (day 8). (b) Relative fold‐change in proteoform concentrations compared to placebo. (c) Relative amount of biologically active CXCL12 (N‐terminal full‐length) to total CXCL12 across dose groups. All plots depict placebo (n = 6), ACT‐1004‐1239 doses (n = 8, except n = 7 for 200 mg at d8) and error bar displays 95% confidence interval.

In summary, an up to approximately twofold increase of biologically active CXCL12 following ACKR3 antagonist treatment was quantified. The ratio of biologically active to total CXCL12 remained constant across the dose groups and timepoints demonstrating an establishment of this proteoform equilibrium within 24 h.

Cross‐reactivity in antibody‐based analysis of CXCL12

In the multiple ascending dose trial, CXCL12 was originally quantified using an antibody‐based assay (ELLA). ¹¹ In a cross‐reactivity assessment, the ELLA assay demonstrated higher reactivity towards CXCL12α^1–68 and similar signals were obtained for CXCL12α^1–67, CXCL12α^3–67, and CXCL12α^5–67 (Figure 5a). At CXCL12α concentrations between ~500 pM (5 ng/mL) and 2000 pM, (20 ng/mL) CXCL12α^1–68 had, on average, 229% cross‐reactivity compared to the same concentration of the CXCL12α^1–67 proteoform. Another assay, the Mesoscale antibody‐based assay, showed higher cross‐reactivity toward the truncated proteoforms, in particular CXCL12α^5–67 (211%) when compared to the same concentration of the CXCL12α^1–67 proteoform (Figure 5b). Thus, quantifying CXCL12α using either antibody‐based assay cannot report results specific to the biologically active CXCL12α proteoform. Using the mass spectrometry‐based assay, CXCL12α^1‐xx proteoform concentrations of 100 pM, CXCL12α^3‐xx proteoform concentrations of about 210 pM, and CXCL12α^x‐67 proteoform concentrations of about 680 pM were quantified in the placebo‐treated adults. Hence, with the determined cross‐reactivity, a significant signal from inactive CXCL12α^3–67 was expected, masking the signal for the active CXCL12α^1–67 to a large part. Using the same clinical samples with the total CXCL12α proteoform concentrations, the Mesoscale assay reported similar concentrations as the developed mass spectrometry‐based assay (Figure 5d) and the ELLA antibody‐based assay underestimated the total CXCL12 concentration by 80% (Figure 5c).

FIGURE 5.

Comparison of results from antibody‐based assays. (a) Cross reactivity of ELLA assay across different CXCL12 proteoforms normalized to CXCL12^1–67. (b) Cross reactivity of Mesoscale assay across different CXCL12 proteoforms normalized to CXCL12^1–67. (c) Correlation of concentrations of total CXCL12 from the IA‐MS biomarker assay to the CXCL12 concentrations measured using the ELLA assay with function from Deming regression. (d) Correlation of concentrations of total CXCL12 from the IA‐MS biomarker assay to the CXCL12 concentrations measured using the Mesoscale assay with function from Deming regression. ELISA, enzyme‐linked immunosorbent assay; ELLA, enzyme‐linked lectin assay; IA‐MS, immunoaffinity mass spectrometry.

DISCUSSION

A new IA‐MS biomarker assay to measure CXCL12α proteoform concentrations in human plasma is described. This fit‐for‐purpose validated assay was used to measure the increased concentrations of biologically active and various other CXCL12 proteoforms in a phase I multiple ascending dose clinical study with the ACKR3 antagonist ACT‐1004‐1239. ¹² The assay consisted of antibody‐based enrichment of CXCL12 followed by proteolytic digestion and detection of the generated peptides using mass spectrometry. Thus, the specificity of the antibody to enrich CXCL12 proteoforms was combined with the high specificity and selectivity of high‐resolution mass spectrometry to differentiate the proteoforms. Such IA‐MS assays have also been published for other low‐concentration biomarkers or biomarkers existing as different proteoforms. ²⁶ , ²⁷ With the reported assay, the proportion of active CXCL12 after drug treatment and in various pathologies can now be quantified to understand regulation of the CXCL12‐CXCR4 chemokine axis. In general, a similar approach can be used for other biomarkers present in various proteoforms.

Like CXCL12, other chemokines have been reported to be proteolytically modified and exist in different proteoforms. ¹ , ²⁶ Using synthesized proteoforms of CXCL12α, our work confirmed that CXCL12α loses its biological activity after cleavage of the first two N‐terminal amino acids generating CXCL12α^3‐67. ²⁸ , ²⁹ Whether cleavage of the last C‐terminal amino acid affects activity to a lesser extent could not be shown with our experimental setup. To our knowledge, this is the first study to quantify active CXCL12α in healthy adults and found that only 10% of CXCL12α is present as the biologically active form in plasma. Our results are in line with a previous study that quantified CXCL12 proteoforms. ¹⁸ That study could, however, not quantify the active CXCL12α^1‐67 proteoform at baseline levels. With our lower limit of quantification of 30 pM, active CXCL12α^1‐xx could be quantified and the determined concentration of CXCL12^3‐67 was very similar to the previously published concentrations. ¹⁸ It has been previously described that CXCL12 is processed rapidly by DPP4 and this is probably the main reason for the low amount of biologically active CXCL12α. ²⁹ Without an antibody specific for biologically active CXCL12α, publications reporting CXCL12 levels overestimate or miss the active CXCL12α proteoform, potentially leading to biased findings.

The difference between the obtained total CXCL12 concentrations and the summed N‐ and C‐terminal peptides was calculated to show whether all proteoforms were quantified. As for greater than 95% of the CXCL12 a C‐terminal form could be quantified, the captured CXCL12 in plasma is predominantly CXCL12α. However, a possible limitation of this assay is the unknown specificity of the capture antibody to other CXCL12 isoforms (i.e., CXCL12β and CXCL12γ). At the C‐terminus, 90% of CXCL12 is cleaved at the terminal amino acid, and only about 8% of CXCL12α is full‐length (ending with amino acid 68). C‐terminal cleavage is thus very prevalent and also occurred to spiked synthesized CXCL12 during the pulldown procedure. About 10% of the N‐termini were full‐length, 21% had the first two amino acids cleaved, and the remaining 69% were unaccounted for. These results were observed with two independent sets of peptides generated by trypsin and GluC protease respectively. DPP4 is known to cleave the first two N‐terminal amino acids in plasma. ¹⁸ Sampling blood in the absence of protease inhibitors led to degradation of the N‐terminus (Figure 2c). However, no N‐terminal cleavage was seen on the spiked synthesized CXCL12 after sampling blood in the presence of protease inhibitors indicating conservation of the N‐terminus during sample processing with our method. Consequently, the exact N‐terminal ending of 69% of CXCL12 in plasma is unclear. CXCL12 is likely cleaved further by another protease, as has been suggested, or another modification such as citrullination may be involved. ² , ³⁰ , ³¹

With top‐down mass spectrometry no inference of proteoforms from peptides would need to be performed. Various recent studies have used a top‐down approach to measure spiked, ¹⁷ increased levels, or endogenous inactive proteoforms of CXCL12α. ¹⁷ , ¹⁸ , ²⁰ However, bottom‐up proteomic approaches have higher sensitivity and enabled here the quantification of peptides specific for the active CXCL12α^1–67 proteoform at baseline levels. New and more sensitive top‐down methods may be developed in the future to quantify CXCL12α proteoforms directly without the need for digestion of the protein.

In the ACT‐1004‐1239 multiple ascending dose clinical study, plasma CXCL12 concentrations were previously reported using an ELLA assay. ¹² The previously obtained concentrations correlated well with the total CXCL12α (r ² = 0.69) concentrations obtained with this method but were 80% lower (Figure 5c). However, as the correlation between the results of the assays was good and a steady‐state equilibrium was established within 24 h between the CXCL12 proteoforms (Figure 4c), the antibody‐based assay can be used as a surrogate for active CXCL12 levels at 24 h after dosing in healthy adults. Hence, any interpretations based on the previous results were still valid and it was confirmed that biologically active CXCL12, increases to a similar extent. Whether the same relative ratio of active versus total CXCL12 concentrations exist at earlier timepoints or peak ACT‐1004‐1239 concentrations, however, cannot be answered with this study as only samples at trough ACT‐1004‐1239 concentration were collected in the BD P800 tubes in the clinical study. Future studies assessing the dynamic phase of the pharmacodynamic effect will be able to assess this question.

In summary, the CXCL12 proteoform biomarker assay supported the development of ACT‐1004‐1239 by demonstrating an increase of biologically active CXCL12 in plasma. With the described measurements, it was shown that around 10% of CXCL12α in human plasma is biologically active. The assay will allow further characterization of ACT‐1004‐1239 in phase II clinical trials and could also be used in other clinical trials assessing CXCL12 as a biomarker to confirm target engagement or explore stratifying patients based on baseline levels or pharmacodynamic effect. Importantly, the approach outlined here could also be applied to quantify proteoforms of other chemokines to support drug target identification, drug development, or discoveries of novel aspects of chemokine biology.

AUTHOR CONTRIBUTIONS

P.B., M.K., and D.S.S. wrote the manuscript. P.B., H.F., B.D., T.L., C.H., P.N.S., J.D., R.W.D.W., M.K., and D.S.S. designed research. P.B., H.F., B.D., and N.S. performed the research and analyzed the data. T.L. contributed new reagents/analytical tools.

FUNDING INFORMATION

This study was sponsored by Idorsia Pharmaceuticals Ltd, Allschwil, Switzerland.

CONFLICT OF INTEREST STATEMENT

All authors were full‐time employees of Idorsia Pharmaceuticals Ltd. during the conduct of the studies included in this manuscript.

Supporting information

Figure S1.

Figure S2.

Figure S3.

Blattmann P, Farine H, Dan B, et al. Quantification of increased biologically active CXCL12α plasma concentrations after ACKR3 antagonist treatment in humans. Clin Transl Sci. 2024;17:e13708. doi: 10.1111/cts.13708

DATA AVAILABILITY STATEMENT

The mass spectrometry proteomics raw and processed data of the analysis of the clinical samples have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository ²² with the dataset identifier PXD038814.