Cell surface IL‐1α trafficking is specifically inhibited by interferon‐γ, and associates with the membrane via IL‐1R2 and GPI anchors

IL‐1 is a powerful cytokine that drives inflammation and modulates adaptive immunity. Both IL‐1α and IL‐1β are translated as proforms that require cleavage for full cytokine activity and release, while IL‐1α is reported to occur as an alternative plasma membrane‐associated form on many cell types. However, the existence of cell surface IL‐1α (csIL‐1α) is contested, how IL‐1α tethers to the membrane is unknown, and signaling pathways controlling trafficking are not specified. Using a robust and fully validated system, we show that macrophages present bona fide csIL‐1α after ligation of TLRs. Pro‐IL‐1α tethers to the plasma membrane in part through IL‐1R2 or via association with a glycosylphosphatidylinositol‐anchored protein, and can be cleaved, activated, and released by proteases. csIL‐1α requires de novo protein synthesis and its trafficking to the plasma membrane is exquisitely sensitive to inhibition by IFN‐γ, independent of expression level. We also reveal how prior csIL‐1α detection could occur through inadvertent cell permeabilisation, and that senescent cells do not drive the senescent‐associated secretory phenotype via csIL‐1α, but rather via soluble IL‐1α. We believe these data are important for determining the local or systemic context in which IL‐1α can contribute to disease and/or physiological processes.


Introduction
Inflammation defends against acute insults, such as infection or injury, but also drives numerous pathologies such as atherosclero-vascular leakage [2], T H 17 cell differentiation, and T-cell expansion and survival [3]. In addition, senescent cells utilize IL-1α to drive the senescence-associated secretory phenotype (SASP) [4][5][6] that is vital for removal of premalignant cells, but mediates deleterious effects when senescent cells accumulate during aging.
Shortly after the cloning of IL-1 [13], a plasma membraneassociated form was reported on mouse macrophages [14] and human monocytes [15]. This cell surface IL-1 (csIL-1) was induced by heat-killed bacteria or LPS, suggesting an important role in immune responses, and was later shown to be exclusively IL-1α [16]. However, because IL-1α does not contain a signal peptide for translocation into the ER, or any hydrophobic regions for membrane insertion, the existence of csIL-1α was refuted [17,18] and remains controversial. Most studies localizing IL-1α to the cell surface utilized IL-1-dependent bioassays (e.g. thymocyte proliferation) that responded to formaldehyde-fixed cells bearing putative csIL-1α [14-16, 19, 20]. However, others demonstrated that fixation causes leakage of intracellular IL-1α, and thus, bioassays may have reported this rather than csIL-1α [17,18]. In addition, although subsequent studies utilized flow cytometry to directly identify csIL-1α [4,16,21], without a viability dye for a dead cell gate, permeabilised cells could allow antibody binding to intracellular IL-1α and/or nonspecific antigens.
We show that macrophages genuinely present csIL-1α after TLR ligation, and this requires de novo synthesis. csIL-1α tethering occurs through IL-1R2 and a glycosylphosphatidylinositol (GPI)anchored protein, and csIL-1α can be activated and released after protease cleavage. Importantly, trafficking of IL-1α to the plasma membrane is exquisitely sensitive to inhibition by IFN-γ. Our data suggest that prior csIL-1α detection may have been due to inadvertent cell permeabilisation and leakage of intracellular IL-1α. Indeed, we show that senescent cells do not drive the SASP via csIL-1α [4], but rather via soluble IL-1α. We believe this novel data are important to understand the local or systemic situations in which IL-1α can participate to disease and/or physiological processes.

Results
Fixation or absence of a viability dye leads to misidentification of csIL-1α. As most studies identifying csIL-1α on macrophages used bioassays requiring fixation [14-16, 19, 20], we investigated if fixation permeabilised cells. Flow cytometry with an anti-IL-1α monoclonal antibody revealed low levels of IL-1α staining with unfixed BM-derived macrophages (BMDMs), which was increased to suspiciously high levels after fixation ( Fig 1A). Furthermore, significant levels of IL-1α leaked from LPS-treated fixed BMDMs into the conditioned media ( Fig 1B). A viability dye also showed high levels of macrophage permeabilisation after fixation (Fig 1C), which occurred with low concentrations of formaldehyde and increased at higher levels ( Fig 1D). Finally, using flow cytometry to examine unstimulated BMDMs with anti-IL-1α and viability dye costaining revealed minimal IL-1α stained cells in the "live" gate, but higher levels of putative IL-1α in the permeabilised "dead" cell gate ( Fig 1E). Importantly, these permeabilised cells inadvertently increased during csIL-1α staining, as freshly harvested unstained BMDMs were approximately 90% viable ( Fig 1F). Together this shows that fixation leads to permeabilisation and IL-1α leakage from cells, suggesting studies using fixation may misidentify csIL-1α, and that flow cytometry without a viability gate can report intracellular IL-1α staining.
A proportion of csIL-1α associates with the membrane via a GPI-anchor. As csIL-1α did not drop to baseline after excess IL-1RA or on Il1r2 −/− BMDMs (Fig 3), we investigated how else IL-1α could associate with the plasma membrane. Previous work shows pro-IL-1α to be the form associated with the cell surface [20,22]. Thrombin specifically cleaves pro-IL-1α N-terminal to the calpain cleavage site [11]. Incubation of LPS-treated BMDMs   on WT BMDMs treated with LPS, followed by incubation with interleukin-1 receptor antagonist (IL-1RA) or IL-1α before staining for csIL-1α (n = 5-10 combined from five to ten independent experiments). (B, C) Flow cytometry for csIL-1α (n = 7-9 combined from seven to nine independent experiments) (B) and western blot for total IL-1α with tubulin loading control (representative of n = 2 independent experiments with one sample per experiment) (C) on WT and Il1r2 −/− BMDMs treated ±LPS. (D) Flow cytometry for csIL-1α on Il1r2 −/− BMDMs treated with LPS, followed by incubation with IL-1RA or IL-1α before staining for csIL-1α (n = 4 combined from four independent experiments). Data represent mean ± SEM; n value corresponds to independent experiments performed with one sample of BMDMs from different mice per experiment; p = *ࣘ0.05, n.s. = not significant, using unpaired t-test (B) or 1-way ANOVA (D) with Dunnett's (A).
with thrombin reduced csIL-1α ( Fig 4A) and transferred active IL-1α into the media (Fig 4B and C), suggesting pro-IL-1α constitutes some of the csIL-1α. Because all the IL-1α antibodies used are raised against the mature IL-1α C-terminus, pro-IL-1α must tether via its N-terminus-otherwise thrombin cleavage would leave the antigenic C-terminus on the cell surface and the released fragment would be undetectable.
Approximately 20% of membrane proteins are GPI-anchored and can be specifically cleaved off with phosphoinositide phospholipase C (PI-PLC) [26]. Thus, incubation of EL4 cells with PI-PLC removed the prototypic GPI-anchored protein Thy-1 from the surface (Fig 4D), while the non-GPI-anchored CD45 or CD115 were not (Fig 4E and F), demonstrating specificity. Incubation of LPS-treated BMDMs with PI-PLC reduced csIL-1α (Fig 4G), with denatured enzyme and vehicle controls without effect, and transferred cleaved IL-1α to the media (Fig 4H). Furthermore, incubation of LPS-treated Il1r2 −/− BMDMs with PI-PLC reduced csIL-1α to baseline (Fig 4I), suggesting loss of all csIL-1α. However, as pro-IL-1α does not contain signal sequences for translocation into the ER where GPI-modifications occur, this implies that csIL-1α can interact with another GPI-anchored protein. Interestingly, csIL-1α is reportedly tethered via a glycoprotein, which are usually GPI anchored [20], while GPI anchors require mannose residues and excess mannose disrupts csIL-1α [20]. However, BMDMs treated with mannose (90 mg/mL as previously used [20]) led to cell permeabilisation that was potentiated by LPS (Supporting information Fig S3), making determination of csIL-1α impossible.
www.eji-journal.eu senescence, would make senescent cells appear to express more csIL-1α than growing cells. Because IL-1α clearly drives the SASP [4,5,29], we tested conditioned media for soluble IL-1α, which showed significant release (ß120 pg/mL) from senescent cells, but none from growing cells (Fig 6G). Similarly, utilizing an anti-IL-1α ELISpot to capture IL-1α as it is released from producing cells revealed multiple spots with senescent cells, and background levels in no-capture antibody controls (Fig 6H). Intriguingly, freshly harvested growing and senescent cells both showed similar low levels of permeabilisation (Fig 6I), suggesting that IL-1α is unlikely to be released from senescent cells through gasdermin D pores or permeabilised membranes. Finally, we also found no csIL-1α on neutrophils, B-cells, T-cells, and DCs (Supporting information Fig  S5-S8), which have been varyingly described to express csIL-1α [11,21,[30][31][32][33].

Discussion
Inflammation is a near-universal response that rapidly acts to reinstate homeostasis and remove the insult that perturbed the system. This innate immune response relies on speed to accomplish its objectives, with the specificity of adaptive immunity complementing the process and providing long-term host protection. However, if not carefully controlled, inflammation can result in greater damage to host tissues than the original insult. Thus, the immune system contains multiple levels of control and checkpoints to ensure responses are appropriately graded. Inflammasome activation is extensively controlled, given that activation leads to all-or-nothing release of IL-1β and IL-18. However, as IL-1α signals identically to IL-1β and can be exteriorized independently of inflammasomes, IL-1α may be a key mediator of initial low-level innate responses.
Using a robust validated system, we provide evidence that macrophages present genuine csIL-1α after activation. IL-1α associates with the plasma membrane in part through its cognate receptor IL-1R2, and via association with a GPI-anchored protein. Importantly, trafficking of IL-1α to the cell surface is exquisitely sensitive to IFN-γ, which also induces MHC II on macrophages, and thus may prevent csIL-1α expression during antigen presentation. In addition, these findings suggest that previous studies utilizing IL-1 bioassays with fixed cells, or flow cytometry without a dead cell gate, likely detected intracellular IL-1α, and thus, multiple cell-types described to express csIL-1α should be reappraised.
csIL-1 was reported before the discovery that IL-1 is the product of two distinct genes, and due to its induction by bacterial agents was proposed to be important in immune responses [14]. Most cytokines are secreted proteins, and thus can have far-reaching effects. In contrast, cell-associated cytokines can only induce autocrine or juxtacrine signaling, limiting effects to the immediate vicinity. This is clearly advantageous for powerful cytokines that induce cell death (e.g. TNF-α) or activate both innate and adaptive immunity (e.g. IL-1), and thus, potentially drive tissue damage and loss of tolerance. Many cell surface cytokines also exist as a soluble form, with shedding of membrane TNF-α by ADAM17, the canonical route for secretion. However, both IL-1α and IL-1β can be released as soluble forms at high concentrations and induce identical signaling after ligating IL-1R1, begging the question, what does csIL-1α do?
Macrophage IL-1α/β release typically occurs after inflammasome engagement, which requires initial priming (e.g. LPS) to upregulate pro-IL-1α/β and inflammasome components, followed closely by secondary signals (e.g. ATP) that activate the inflammasome. Thus, if a patrolling macrophage encounters a low-level infection, LPS would induce csIL-1α and priming, but without significant cell death there would be no ATP to activate the inflammasome. This could enable a localized low-scale inflammatory response driven by csIL-1α to try and eliminate infection with minimal tissue damage (e.g. "DEFCON 3"). If the insult generated host-or pathogen-derived proteases, csIL-1α would be cleaved from the macrophage surface, allowing fully active IL-1α to diffuse and signal more widely (e.g. "DEFCON 2"). If the infection grew, or the macrophage initially encountered an advanced infection causing host cell death, this would provide ATP (i.e. a danger signal) to activate the inflammasome, leading to high level IL-1 and IL-18 release and licensing of a full-scale immune response with accompanying collateral tissue damage (e.g. "DEFCON 1").
csIL-1α is reported to be the pro-IL-1α form ( [20,22] and Fig 4A to C), and pro-IL-1α is significantly less active than the mature form [6,9,10,34]. Thus, the biological activity of csIL-1α, per se, is unclear. Previous assays identified putative csIL-1α via its activity, but we and others show that fixed macrophages used in these bioassays leak intracellular IL-1α (Fig 1B to D and [17,18]), and thus apportioning IL-1α activity specifically to the cell surface form is technically challenging. Furthermore, any csIL-1α associated via IL-1R2 (Fig 3) will be devoid of activity unless it dissociates and binds IL-1R1, which because of a higher affinity of IL-1α for IL-1R1 (10 −10 M) versus IL-1R2 (10 −8 M) could make this plausible [35]. Alternatively, calpain [10], thrombin [11], caspase-5 [6], granzyme B, elastase, and chymase [9], all cleave pro-IL-1α, resulting in increased activity and potential shedding of soluble mature IL-1α (as shown for thrombin in Fig 4A to C and [11]). Interestingly, most of these proteases could be considered to be activated under situations of "danger" [36] (e.g. calpain upon necrosis, thrombin during hemostasis, granzyme B during cytotoxic T-cell degranulation), and thus, csIL-1α on macrophages could be cleaved, resulting in activation and release to instigate signaling. Alternatively, given that IL-1 enhances T-cell expansion [37], if macrophage csIL-1α has sufficient activity it could perhaps act as a form of costimulation during antigen presentation. However, as IFN-γ induces macrophage MHC II expression [38], but profoundly inhibits macrophage csIL-1α (Fig 5), this response could act to limit IL-1-driven Th17 cell differentiation and reinforce Th1 cells.
In addition, pure DCs are incapable of releasing IL-1, with previous examples due to contaminating macrophages [40]. Using our validated system, we show neither DCs, T-cells, B-cells nor neutrophils express true csIL-1α before or after stimulation (Supporting information Fig S5-S8), suggesting previous findings may have been artefact, or the result of examining transformed lymphocyte clones [30,31]. So why does csIL-1α appear so unique to macrophages when multiple other cell types constitutively or inducibly express IL-1α? One possibility is that all macrophages we have tested do not express IL-1R1 (JNeC, MC unpublished), and thus macrophage csIL-1α cannot bind its signaling receptor to induce autocrine stimulation, which could lead to sustained IL-1 signaling and potentially chronic inflammation. We were unable to disassociate csIL-1α upon mannose treatment [20]. Instead, mannose lead to cell permeabilisation, and thus, inability to distinguish csIL-1α from intracellular IL-1α (Supporting Information Fig S3). Thus, apparent csIL-1α displacement by mannose [20], could be due to intracellular IL-1α leakage from mannose permeabilised cells before fixation. We were also unable to detect csIL-1α on senescent fibroblasts [4], but instead provide clear evidence that soluble IL-1α is readily released (Fig 6).
Due to the limited knowledge on csIL-1α, any role it may play in disease is not understood. csIL-1α inhibits hepatocellular carcinoma development in a model using hepa1-6 cells stably expressing exogenous Il1a that had the nuclear localization signal and calpain site mutated [41]. However, the outcome of these mutations was not sufficiently validated, and Il1a was not controlled by its endogenous promoter. Similarly, csIL-1α worsens collagen destruction and arthritis in transgenic mice expressing truncated human IL-1α from the chicken β-actin promoter [42]. However, transgenic cells released more soluble IL-1α, making interpretation problematic. Thus, until we identify a unique mechanism for csIL-1α tethering that we can modulate, the precise role of csIL-1α remains unknown. We report IFN-γ as the first negative regulator of IL-1α trafficking to the cell membrane ( Fig 5). However, we currently do not know the pathway downstream of IFN-γ signaling that mediates this, or if in the future we could leverage this information to specifically inhibit csIL-1α independent of other IFN-γ effects (e.g. MHC II expression). However, IFN-γ does not block csIL-1α via reduced IL-1R2 expression (Supporting information Fig S4). Similarly, we also do not understand whether the csIL-1α associated with IL-1R2 is a binding "artefact" or represents an important biological pool of IL-1α. Finally, we also show that IL-1α release from senescent cells occurs independently of membrane pores or permeabilisation (Fig 6I), suggesting IL-1α exits cells by an unknown mechanism, or is perhaps immediately cleaved and released from csIL-1α as the substrate.
In conclusion, we show that bone fide csIL-1α is expressed on the surface of macrophages after TLR ligation. csIL-1α tethers to the plasma membrane through IL-1R2 or by association with a GPIanchored protein, and can be cleaved, activated, and released by proteases. Importantly, trafficking of pro-IL-1α to the macrophage surface is inhibited by IFN-γ. However, we find no evidence of csIL-1α on other cell types. We believe these data are important for determining whether IL-1α acts in a local or systemic context during immune responses, and thus, how it contributes to disease and/or physiological processes.

Materials and methods
All materials are from Sigma-Aldrich unless otherwise stated.

Gene expression analysis
RNA was isolated (RNeasy, Qiagen) and converted to cDNA using AMV reverse transcriptase (Promega). qPCR used Taqman probes with Amplitaq Gold (Life Tehcnologies) in a RotorGene thermocycler (Corbett). Gene expression was evaluated using the 2ˆ− CT method using Gusb and Tbp as reference genes.

Statistical analysis
Statistical testing used Prism 8 (GraphPad). Unpaired t-tests compared two conditions. One-way ANOVA compared three or more conditions. If ANOVA was significant, Tukey's post-hoc compared all conditions to each other, and Dunnet's compared all conditions to control. Key assays were performed in duplicate.