A New Serum Macrophage Checkpoint Biomarker for Innate Immunotherapy: Soluble Signal-Regulatory Protein Alpha (sSIRPα)

Background and Aims: The macrophage “don’t eat me” pathway CD47/SIRPα is a target for promising new immunotherapy. We hypothesized that a soluble variant of SIRPα is present in the blood and may function as a biomarker. Methods: Monocyte derived macrophages (MDMs) from human buffy-coats were stimulated into macrophage subtypes by LPS and IFN-γ (M1), IL-4 and IL-13 (M2a), IL-10 (M2c) and investigated using flow cytometry. Soluble SIRPα (sSIRPα) was measured in cell cultures and serum by Western blotting and an optimized ELISA. Serum samples were obtained from 120 healthy individuals and from 8 individuals challenged by an LPS injection. Results: All macrophage phenotypes expressed SIRPα by flowcytometry, and sSIRPα was present in all culture supernatants including unstimulated cells. M1 macrophages expressed the lowest level of SIRPαand released the highest level of sSIRPα (p < 0.05). In vivo, the serum level of sSIRPα increased significantly (p < 0.0001) after an LPS challenge in humans. The median concentration in healthy individuals was 28.7 µg/L (19.8–41.1, 95% reference interval), and 20.5 µg/L in an IFCC certified serum reference material. The protein was stable in serum for prolonged storage and repeated freeze/thawing. Conclusions: We demonstrate that sSIRPα is produced constitutively and the concentration increases upon macrophage activation both in vitro and in vivo. It is present in human serum where it may function as a biomarker for the activity of tumor-associated macrophages (TAMs), and for monitoring the effect of immunotherapy.


Introduction
T-cell based immunotherapy has revolutionized the treatment of certain cancers; however, in most cancers significant response rates are still not reached. Research is therefore focused now on ways to involve the innate immune system in cancer therapy [1]. Tumor-associated macrophages (TAMs) can phagocytize cancer cells and stimulate adaptive immunity; however, TAMs predominantly suppress antitumor immunity. This is in part due to cancer cells expressing so-called "don't eat me" signals that inhibit phagocytosis through receptor-binding to TAMs [1]. An important signal system is the CD47/signal regulatory protein alpha (SIRPα) pathway, which is the target for several inhibitory drugs that are currently being investigated in clinical trials including soluble variants of SIRPα to outcompete macrophage binding to Figure 1. CD47 on tumor cells protects against phagocytosis. It interacts with SIRPα on the surface of TAMs resulting in decreased macrophage phagocytic capacity. Activation of macrophages leads to shedding by metalloproteases of soluble receptors (sCD163, sCD206) that function as biomarkers of macrophage activation in the blood. Soluble SIRPα (sSIRPα) is thought to be the product of extracellular domain shedding by the action of ADAM10 that, upon activation, cleaves SIRPα at a juxta-membrane position (15). Created with BioRender.com.
SIRPα is an approximately 90 kDa glycoprotein with four possible N-linked glycosylation sites (https://www.uniprot.org/uniprot/P78324, accessed on 20 June 2022). In humans, the SIRP gene family consists of SIRPα, SIRPβ1, SIRPβ2, SIRPγ, and SIRPδ and is clustered on chromosome 20p133 [3][4][5][6]. SIRPα consists of three extracellular immunoglobulin superfamily (IgSF) domains, a single-span transmembrane region and a tyrosinerich cytoplasmic tail. The IgSF domains including one IgV-like and two IgC-like domains are proximal to the N terminal. The IgV-like domain is responsible for binding antigens such as the IgV-like domain of CD47. The cytoplasmic tail contains four tyrosine residues, which upon phosphorylation recruit the tyrosine phosphatases SHP-1 and SHP-2 resulting in downstream signaling. SIRPα is normally expressed on neural tissue and on myelomonocytic cells, including monocytes, macrophages, granulocytes, dendritic cells and their precursors [4, [7][8][9][10]. SIRPα is also expressed on TAMs, and recent evidence links high SIRPα expression to poor survival in patients with esophageal squamous cell carcinoma [11]. Thus, SIRPα is both a potential therapeutic target, and a potential biomarker in cancer.
A soluble form of SIRPα has been detected in vitro [12][13][14]. This protein is thought to be the product of extracellular domain shedding by the action of a disintegrin and metalloproteinase domain-containing protein 10 (ADAM10) that, upon activation, cleaves SIRPα at a juxta-membrane position to produce soluble SIRPα (sSIRPα) [12] (Figure 1). This soluble form may be present in human blood, and its concentration could reflect the level of intra-tumoral immune activation.
We hypothesized that sSIRPα is present as a stable protein in human blood and may be a useful biomarker with the potential for monitoring immunotherapy. Furthermore, we wanted to establish a robust method for its determination. It interacts with SIRPα on the surface of TAMs resulting in decreased macrophage phagocytic capacity. Activation of macrophages leads to shedding by metalloproteases of soluble receptors (sCD163, sCD206) that function as biomarkers of macrophage activation in the blood. Soluble SIRPα (sSIRPα) is thought to be the product of extracellular domain shedding by the action of ADAM10 that, upon activation, cleaves SIRPα at a juxta-membrane position (15). Created with BioRender.com.
SIRPα is an approximately 90 kDa glycoprotein with four possible N-linked glycosylation sites (https://www.uniprot.org/uniprot/P78324, accessed on 20 June 2022). In humans, the SIRP gene family consists of SIRPα, SIRPβ1, SIRPβ2, SIRPγ, and SIRPδ and is clustered on chromosome 20p133 [3][4][5][6]. SIRPα consists of three extracellular immunoglobulin superfamily (IgSF) domains, a single-span transmembrane region and a tyrosine-rich cytoplasmic tail. The IgSF domains including one IgV-like and two IgC-like domains are proximal to the N terminal. The IgV-like domain is responsible for binding antigens such as the IgV-like domain of CD47. The cytoplasmic tail contains four tyrosine residues, which upon phosphorylation recruit the tyrosine phosphatases SHP-1 and SHP-2 resulting in downstream signaling. SIRPα is normally expressed on neural tissue and on myelomonocytic cells, including monocytes, macrophages, granulocytes, dendritic cells and their precursors [4, [7][8][9][10]. SIRPα is also expressed on TAMs, and recent evidence links high SIRPα expression to poor survival in patients with esophageal squamous cell carcinoma [11]. Thus, SIRPα is both a potential therapeutic target, and a potential biomarker in cancer.
A soluble form of SIRPα has been detected in vitro [12][13][14]. This protein is thought to be the product of extracellular domain shedding by the action of a disintegrin and metalloproteinase domain-containing protein 10 (ADAM10) that, upon activation, cleaves SIRPα at a juxta-membrane position to produce soluble SIRPα (sSIRPα) [12] (Figure 1). This soluble form may be present in human blood, and its concentration could reflect the level of intra-tumoral immune activation.
We hypothesized that sSIRPα is present as a stable protein in human blood and may be a useful biomarker with the potential for monitoring immunotherapy. Furthermore, we wanted to establish a robust method for its determination.

Human Monocyte-Derived Macrophages
Buffy coats (one sample, ∼50 mL) from five healthy donors were obtained from the blood bank at the Department of Clinical Immunology, Aarhus University Hospital, Aarhus, Denmark (project no. 0094). According to Danish law, the use of anonymized buffy coats does not require specific ethical approval.
Monocytes were isolated from the buffy coats as described previously [15]. In short, the buffy coats were diluted 1:2 in 0.9% NaCl and the peripheral blood mononuclear cells (PBMCs) were isolated using density gradient centrifugation on a Histopaque-1077 gradient (Sigma-Aldrich, Munich, Germany). Monocytes were subsequently isolated from the PBMCs using EasySep™ Human Monocyte Isolation Kit (Stemcell Technologies, Vancouver, Canada) according to the manufacturer's protocol.
The cells were analyzed on a Navios Flow Cytometer (Beckman Coulter, Brea, CA, USA). Spectral overlap compensation was performed using single-stained antibody capture beads, BD™ CompBeads Plus (BD Biosciences), MACS ® Comp Bead Kit anti-REA (Miltenyi Biotech), and ArC™ Amine Reactive Compensation Bead Kit for Live/dead (ThermoFisher). Data processing was performed using FlowJo 10.7 for Windows (FlowJo, LLC, Ashland, OR, USA). The gating strategy for both panels is presented in Supplemental Figure S1.
Plates were washed and subsequently 100 µL of the sample, diluted in PBS-albumin, was added and incubated for 1 h. The wells were then washed, and 100 µL of in-house biotinylated monoclonal anti-SIRPα antibody (Abcam, Cambridge, UK, EPR22930-163) was diluted to 0.4 mg/L in PBS-albumin, added, and incubated for 1 h.
A standard curve spanning a range from 0.125-8 µg/L was prepared from a recombinant human SIRPα (R&D Systems, 9378-SA-050). A human serum quality control sample (HK12, DEKS, Glostrup, Denmark)) was included in all runs, diluted 1:100. One hundred and twenty human serum samples were obtained from the blood bank at the Department of Clinical Immunology, Aarhus University Hospital, Aarhus, Denmark (project no. 304) and diluted 1:100. The concentration of sSIRPα was also determined in a certified reference material (ERM ® -DA470k/IFCC). Recombinant human SIRPβ1 (R&D Systems, 9978-SB) and recombinant human SIRPγ (R&D Systems, 9999-SB) were used to test for specificity.

Human LPS Exposure
We measured sSIRPα in plasma samples from a previously conducted human randomized crossover trial (clinicaltrials.gov registration no. NCT01705782) investigating eight young, lean, healthy male volunteers for six consecutive hours following LPS injection [16]. A single dose of 1 ng/kg bodyweight of Escherichia coli endotoxin was given at time = 0 min (The United States Pharmacopeial Convention, Inc., Rockville, Maryland, lot H0K354;). Blood samples were collected at baseline, 180 min, and 360 min following LPS exposure. Plasma was stored at −20 • C until analysis.

Statistical Methods
Multiple comparisons were performed by ANOVA followed by Tukey's multiple comparison test. Non-parametric tests were used to investigate the effects of freeze-thaw cycles, storage temperature, storage time, and time to centrifugation. For sample stability before centrifugation and freeze-thaw stability, the Friedman test was used. The stability of pipetted samples, at different storage temperatures for up to one month, was tested with Wilcoxon matched-pairs signed rank test. A parametric 95% reference interval was established by calculating the mean ±1.96 σ from log-transformed data. Statistical analysis was performed using GraphPad Prism 8 and Analyse-it for Microsoft Excel.
All MDM subtypes expressed SIRPα at levels measurable by flow cytometry (Table 1), and there were significantly different expression levels (one-way ANOVA p = 0.0087) between subtypes, with the lowest expression on M1 MDMs (Table 1, Figure 2a).  Soluble SIRPα was present in supernatants from all MDM subtypes, indicating a constitutive production (Figure 2b). Statistically significantly higher levels were measured in supernatants from M1 MDMs, in accordance with proteolytic shedding by metalloproteases upon LPS stimulation (Table 2, Figure 2b).   Soluble SIRPα was present in supernatants from all MDM subtypes, indicating a constitutive production (Figure 2b). Statistically significantly higher levels were measured in supernatants from M1 MDMs, in accordance with proteolytic shedding by metalloproteases upon LPS stimulation (Table 2, Figure 2b). Soluble SIRPα was detectable by Western blotting in human serum showing an apparent molecular weight similar to recombinant extracellular SIRPα (Figure 2c).

Shedding of sSIRPα In Vivo
To investigate if LPS mediated SIRPα shedding could be demonstrated in vivo, we measured the concentration of sSIRPα in consecutive blood samples from eight healthy individuals injected with a bolus of LPS in a randomized crossover design (Figure 3a). We compared the kinetic to two endocytic receptors sCD206 (Figure 3b) and sCD163 (Figure 3c), also known to be shed from macrophages upon protease activity. The plasma level of sSIRPα increased by 16 and 20% after 3 and 6 h (repeated measures one-way ANOVA, p < 0.0001) which was comparable to sCD206 (21 and 23%, respectively, p < 0.0001), but lower than for sCD163 (109 and 41%, respectively, p = 0.0002). The level of sCD163 reached maximum already after 3 h, whereas sSIRPα and sCD206 showed a weaker and more prolonged response. 3c), also known to be shed from macrophages upon protease activity. The plasma level of sSIRPα increased by 16 and 20 % after 3 and 6 h (repeated measures one-way ANOVA, p < 0.0001) which was comparable to sCD206 (21 and 23%, respectively, p < 0.0001), but lower than for sCD163 (109 and 41 %, respectively, p = 0.0002). The level of sCD163 reached maximum already after 3 h, whereas sSIRPα and sCD206 showed a weaker and more prolonged response.

Preanalytical Factors
There was no difference between serum-and plasma-samples (p = 0.95). In pools of serum, sSIRPα was stable for more than 24 h at RT and 4 • C before centrifugation of blood-samples (Figure 4a). In pipetted serum, sSIRPα was stable for at least one week at 4 • C (Figure 4b), and for at least 29 days at −20 • C (Figure 4c). The concentration of sSIRPα was found to be stable through at least five rounds of freezing and thawing (Figure 4d). olecules 2022, 12, x FOR PEER REVIEW

Preanalytical Factors
There was no difference between serum-and plasma-samples (p = serum, sSIRPα was stable for more than 24 h at RT and 4 °C before centri samples (Figure 4a). In pipetted serum, sSIRPα was stable for at least (Figure 4b), and for at least 29 days at −20 °C (Figure 4c). The concentrat found to be stable through at least five rounds of freezing and thawing

Serum Concentration of sSIRPα in Healthy Individuals
Soluble SIRPα was measured in serum from 120 healthy blood d women, age 18-69) and was clearly detectable in all samples. The sam tween 17.4 and 54.6 µg/L (median 28.7) and showed a log-Gaussian dis metric 95% reference interval was established from log-data to 19.8 µg 20.8) to 41.1 µg/L (90 % CI 39.2-43.1). There was no positive correlatio and age (R 2 = 0.01, p = 0.22). Men had marginally higher levels than wom 28.5 µg/L, p = 0.017).

Discussion
In this work, we show that SIRPα is present as a stable soluble blood which we denote sSIRPα. A soluble form of SIRPα has previousl vitro [11,13,14], but to our knowledge, this is the first time sSIRPα ha the human circulation.
It is possible that this soluble form of SIRPα may function as a bio vation of TAMs during immunotherapy targeting the innate immune the use of sPD1 during T-cell targeted therapy [19].
SIRPα is expressed on neurons and myelomonocytic cells [8,9]. Es anti-inflammatory M2-like TAMs is of interest. Here, it transmits a "do upon binding to the phagocytosis checkpoint CD47 on tumor cells. In this we observed a significantly lower SIRPα expression on inflamma parallel with an increase in soluble SIRPα in the adjacent cell medium partly, due to an inflammatory activation of sheddases, of which ADA involved in SIRPα shedding in humans [12]. The remaining intrace

Discussion
In this work, we show that SIRPα is present as a stable soluble protein in human blood which we denote sSIRPα. A soluble form of SIRPα has previously been detected in vitro [11,13,14], but to our knowledge, this is the first time sSIRPα has been detected in the human circulation.
It is possible that this soluble form of SIRPα may function as a biomarker of the activation of TAMs during immunotherapy targeting the innate immune system, similar to the use of sPD1 during T-cell targeted therapy [19].
SIRPα is expressed on neurons and myelomonocytic cells [8,9]. Especially its role in anti-inflammatory M2-like TAMs is of interest. Here, it transmits a "don't eat me signal" upon binding to the phagocytosis checkpoint CD47 on tumor cells. In accordance with this we observed a significantly lower SIRPα expression on inflammatory M1 MDMs in parallel with an increase in soluble SIRPα in the adjacent cell medium. This is, at least partly, due to an inflammatory activation of sheddases, of which ADAM10 seems to be involved in SIRPα shedding in humans [12]. The remaining intracellular fragment is cleaved further by γ-secretase. The accumulated intracellular product triggers phosphorylation of IKKα/β, which in turn activates NFκB and induces inflammation [14].
Soluble SIRPα was detectable in all investigated serum samples, and the concentration in serum increased after in vivo LPS challenge in healthy individuals. Although the increase was significant, it was less pronounced and acute than was seen for the endocytic macrophage receptor CD163. This may be due to a more direct LPS activation of the ADAM17 enzyme responsible for sCD163 shedding [20]. Nevertheless, the sSIRPα increase in response to LPS indicates that macrophage activation is reflected in circulating levels of sSIRPα.
As TAMs expressing SIRPα are abundant in cancer, sSIRPα may function as a biomarker, and we envision two potential uses of sSIRPα. Firstly, high pre-treatment levels of sSIRPα could potentially signal a poor prognosis reflecting sSIRPα constitutively released by high numbers of TAMs. This is supported by SIRPα being constitutively released by MDMs and by recent evidence linking high SIRPα expression to poor survival in sev-eral tumors [11,[21][22][23][24]. Secondly, a rise in sSIRPα levels after initiation of immunotherapy may reflect the level of immune activation achieved, and thus predict the successfulness of the treatment. This is supported by the observed increase in sSIRPα after pro-inflammatory activation both in vitro and in vivo. Importantly, the activation may also shift the balance between soluble and cell-bound SIRPα, resulting in a decreased CD47 binding of cellular SIRPα due to blockage by increasing sSIRPα [25,26].
A limitation of our study is that the precise identity of sSIRPα in serum is not known. In Western blotting, the size of the immunologically measured protein corresponds to the extracellular part of SIRPα. This could be strengthened by a comparison to sSIRPα from macrophage cell cultures and characterization of purified serum sSIRPα by massspectrometry [18].
Our results pave the way for future studies exploring the clinical use of sSIRPα. Overall sSIRPα proved to be a very stable protein in the blood, facilitating blood-sampling and biobanking. Additionally, to help transferability of results across assays, we established a reference interval for the healthy population, using blood donor samples traceable to an international reference serum.
In conclusion, we have demonstrated that soluble SIRPα is present in human serum (denoted sSIRPα); it is stable, and it can be measured using the described validated assay. It is produced constitutively by MDMs, and the concentration of sSIRPα increases upon macrophage activation both in vitro and in vivo. sSIRPα may therefore function as a biomarker for the activity of TAMs and for monitoring immunotherapy, which is now being investigated in several clinical studies.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by The Central Denmark Region Ethics Commitee approved the (1-10-71-410-12) and the study was reported at www.clinicaltrials.gov, accessed on 20 June 2022 (identification number NCT01705782) [16].
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study [16].

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.