LPS does not stimulate HMGB1 release from macrophages
The literature contains conflicting reports regarding the secretion of HMGB1 in response to LPS alone, with some papers demonstrating this effect4, and others that do not26. To evaluate whether LPS alone leads to HMGB1 secretion, we first assessed HMGB1 secretion in bone marrow-derived macrophages (BMDM) by ELISA. LPS-stimulated BMDM did not release HMGB1 into the extracellular medium (Fig. 1a). In addition, LPS stimulation did not change HMGB1 expression levels (Fig. 1b). Moreover, immunostaining for HMGB1 in LPS-stimulated cells did not reveal any translocation from the nucleus to an intracellular vesicular compartment (Fig. 1c, d), such as the endolysosome compartment. Specifically, HMGB1 did not colocalize with either Rab7 or LAMP1, respective markers of late endosomes and lysosomes (Fig. 1c, d). While some reports have documented the bulk translocation of HMGB1 from nucleus to cytosol following LPS stimulation, we do not observe this phenomenon. The lack of observable HMGB1 puncta in LPS-stimulated BMDM similarly argues against any other secretory vesicle-mediated HMGB1 secretion pathway.
Pyroptosis activators lead to HMGB1 release from macrophages
Having failed to elicit HMGB1 release using LPS alone, we next sought to test for inflammasome-mediated HMGB1 release by stimulating mouse BMDM with known activators of pyroptosis. In LPS-primed BMDM, nigericin or potassium depletion activated the inflammasome as confirmed by the presence of mature IL-1β in the extracellular culture medium, detected by immunoblotting (Fig. 2a). No IL-1β secretion was observed in untreated cells or in cells treated with LPS alone. HMGB1 was found in the culture medium under the same inflammasome-activating conditions (Fig. 2a). GAPDH, present in the cytosol as a tetramer of 37 kDa subunits27, was also found in the extracellular medium in all conditions of inflammasome activation (Fig. 2a). Since this protein complex is larger than the GSDMD pore, we hypothesized that this was indicative of cellular lysis during inflammasome activation. To further investigate this finding, we performed a colorimetric assay to test for LDH release from the BMDM under the same set of conditions. Release of LDH, a large cytosolic protein, is a common indicator of cell rupture and cytotoxicity24,28. Consistent with prior publications, LPS + nigericin or potassium depletion caused pyroptosis, as indicated by LDH release (Fig. 2b). In comparison, LPS treatment alone caused minimal release of LDH, comparable to that of control untreated cells (Fig. 2b). Thus, the appearance of LDH in the supernatant mimics the release of GAPDH detected by western blotting.
To confirm the link between inflammasome activation and HMGB1 release, we performed immunofluorescence on BMDM treated with LPS alone or in combination with nigericin. LPS + nigericin stimulated inflammasome activation in a large proportion of BMDM, as shown by the presence of ASC specks (Fig. 2c). In cells with ASC specks, HMGB1 was lost almost entirely from the cell, indicating a strong relationship between inflammasome activation and HMGB1 release. From these initial data, two possibilities exist: HMGB1 is either released by viable BMDM or escapes from BMDM following lytic cell death.
Pyroptosis-induced HMGB1 release is gasdermin D-dependent
The data presented above indicate that inflammasome activation results in concomitant release of IL-1β and HMGB1 into the extracellular space, and that cytotoxicity occurs simultaneously with this release. Since other cytosolic proteins, including IL-1β and IL-18, are secreted from cells through GSDMD, we hypothesized that HMGB1 may traverse a similar route. To determine if the GSDMD pore might be required for HMGB1 release in response to LPS + nigericin we first inhibited GSDMD with necrosulfonamide21. Extracellular release of both IL-1β and HMGB1 was completely inhibited by necrosulfonamide (Fig. 3a), as was cytotoxicity (LDH release) (Fig. 3b). As a more specific approach to inhibit the GSDMD pore, BMDM were isolated from Gsdmd−/− mice or littermate controls as described in Materials and Methods. The lack of GSDMD protein in these cells was confirmed by western blot (Fig. 3d). Following treatment with LPS + nigericin, Gsdmd−/− BMDM did not undergo pyroptosis, as indicated by the lack of LDH release as compared to littermate control BMDM (Fig. 3c). As expected, LPS + nigericin treatment of wild-type cells stimulated the release of HMGB1 and processed IL-1β into the extracellular medium (Fig. 3e). However, in the absence of GSDMD, none of LDH, IL-1β, nor HMGB1 are released from the BMDM following inflammasome activation, despite the fact that the inflammasome was activated even in Gsdmd−/− BMDM (ASC specks in Fig. 3f, caspase 1 cleavage Supplementary Fig. S1). Therefore, as with IL-1β, inflammasome-mediated release of HMGB1 is GSDMD dependent.
HMGB1 is not secreted through the gasdermin D pore
Two scenarios still remain consistent with these data. In one, HMGB1 directly traverses the GSDMD pore in a manner similar to IL-1β. In a second model, GSDMD inserts in the membrane, and under in vitro conditions, water then enters the cell, resulting in cell lysis. HMGB1 then exits the cell in a nonspecific fashion, thereby relying on GSDMD as a stimulus for cell rupture, not as a direct conduit for HMGB1. This distinction is critical since it has recently been demonstrated that inflammasome activation need not result in pyroptosis18,24,29. In an effort to separate inflammasome activation from pyroptosis and cellular rupture, we repeated several experiments in the presence of glycine. While its mechanism of action is poorly characterized, glycine is known to prevent pyroptotic cell lysis, while leaving the upstream steps of inflammasome activation and insertion of GSDMD into the plasma membrane unperturbed30. In line with this prior observation, the addition of glycine effectively blocks LDH release in response to LPS + nigericin or potassium depletion (Fig. 4a) without affecting the secretion of IL-1β (Fig. 4b, c). However, glycine inhibits the release of HMGB1 from BMDM (Fig. 4b, d), consistent with the notion that that HMGB1 release in response to inflammasome activation occurs only in the presence of lytic cell death.
To further investigate this finding, we employed two other reagents known to activate inflammasomes while resulting in minimal cell death. One reagent tested was 1-palmitoyl-2-glutaryl-sn-glycero-3-phosphocholine (PGPC), which is a component of oxidized phospholipids31. These lipids, released by dead cells, are known to induce inflammasome activation and IL-1β release in the absence of cell death18. The other reagent which induces IL-1β release from living BMDM is a mutant strain of S. aureus lacking O-acetyltransferase A (ΔOatA)32. We first confirmed that both reagents cause minimal cell lysis and LDH release in BMDM pre-treated with LPS (Fig. 4e). Consistent with our other findings, both PGPC and ΔOatA stimulated IL-1β processing and release in LPS-treated BMDM (Fig. 4f). As seen when cell lysis was prevented with glycine, HMGB1 was not significantly released in response to either of these reagents as measured by ELISA (Fig. 4g) or HMGB1 localization, which remained nuclear (Fig. 4h, i). Taken together, our data show that inflammasome activation is not sufficient for HMGB1 release, and that unlike prior descriptions26,33, HMGB1 release in this context of inflammasome activation requires lytic cell death or membrane rupture.
HMGB1 is released through other large membrane pores
Our data suggest that HMGB1, under conditions of inflammasome activation, is not released through GSDMD pores, but is released passively during cell lysis. To further investigate these findings, we assessed the effect of another membrane pore on HMGB1 release from BMDM. S. pneumoniae produces a pore-forming toxin called pneumolysin34. Pneumolysin inserts in the plasma membrane of host cells, and forms a large pore, with an internal diameter of ~30 nm35, which is 2–3 times larger than the size of the GSDMD pore. Upon treatment of BMDM with pneumolysin, even in the absence of LPS priming, there is significant release of both HMGB1 and LDH into the culture medium (Fig. 5a, b). In contrast, pneumolysin has no effect on IL-1β, release, which requires LPS priming and a pyroptosis activator (e.g., nigericin or potassium depletion). As expected, pneumolysin alone causes a large reduction in nuclear HMGB1 levels (Fig. 5c).
These data, as well as our data shown above regarding HMGB1 release in response to LPS + nigericin, are surprising given the overwhelming localization of HMGB1 to the nucleus in resting cells. To our knowledge, neither nigericin nor pneumolysin directly affect the nuclear envelope, and therefore should not increase nuclear permeability for HMGB1. Together, these data demonstrate that a large plasma membrane pore is sufficient for the release of HMGB1 from cells. Additionally, these data are consistent with a model whereby HMGB1 is in rapid dynamic equilibrium between the nucleus and cytosol, with the vast majority of HMGB1 inside the nucleus at steady state in intact cells.
LPS increases serum HMGB1 independent of gasdermin D
The release of HMGB1 into the blood during endotoxemia has been well-documented9,10,36. Endotoxemia has similarly been shown to stimulate inflammasome activation26 and, GSDMD deletion reduces mortality in this model20,21. Given that (a) endotoxemia stimulates inflammasomes, (b) GSDMD blockade is protective, and (c) HMGB1 is released during endotoxemia, several groups have drawn the conclusion that HMGB1, therefore, is released in vivo by an inflammasome- and GSDMD-dependent mechanism25. However, the precise relationship of these states has not been thoroughly probed, and an alternative hypothesis is conceivable, whereby HMGB1 release and inflammasome activation are, in fact, parallel processes. Our in vitro data demonstrate that inflammasome-mediated HMGB1 release requires cell lysis, and there is a growing body of evidence suggesting that lytic cell death may not be a common endpoint after inflammasome activation in vivo. Indeed, macrophage can activate inflammasomes without cell lysis18, and cells can repair membrane pores using the ESCRT III system and survive following inflammasome activation24.
To test the hypothesis that HMGB1 release and pyroptotic cell death are spatially unrelated in endotoxemia, we used an acute intraperitoneal LPS injection model in wild-type and Gsdmd−/− mice. In this model, serum HMGB1 levels increase likely from a hepatocyte source37 and remain elevated along with other pro-inflammatory cytokines including IL-1β within hours of LPS administration36. Mice were injected with 20 mg kg-1 LPS or vehicle, and were sacrificed 6 h later. Blood was collected by cardiac puncture. Splenic whole cell lysates (Fig. 6a, b) demonstrate LPS-dependent upregulation of NLRP3 and proteolytic cleavage of caspase-1, indicative of inflammasome activation. As has been shown previously, wild-type animals secreted both HMGB1 and IL-1β into the blood following LPS treatment (Fig. 6c, d). As expected, Gsdmd−/− animals did not secrete IL-1β in response to LPS, demonstrating the requirement for GSDMD in IL-1β secretion. In contrast, HMGB1 release into the blood in response to LPS was not affected by GSDMD loss. Both wild-type and Gsdmd−/− animals secreted similar amounts of HMGB1, in spite of the fact that Gsdmd−/− mice did not secrete IL-1β, and had similar degrees of inflammasome activation as wild-type mice.
Taken together, these data demonstrate several novel phenomena. First, HMGB1 release in vivo does not require GSDMD. Second, IL-1β and HMGB1 do not exit cells by the same pathway in this model. Third, given that inflammasome activation in vitro leads to IL-1β release in all cases, yet only stimulates HMGB1 release if cells rupture, our in vivo data are consistent with a model in which cell lysis (pyroptosis) is not a significant feature of inflammasome activation in vivo.