Ancient pyroptotic machinery via GSDMA/B cleavage by LPS-activated caspase-1 in cartilaginous fish

Reviewer #1 (Public review):

Summary

The authors present a valuable study of the gasdermins and caspases encoded by Callorhinchus milii, a shark that is one of the most basal members of the cartilaginous fishes. C. milii encodes GSDME and PJVK as well as another gene here called GSDMA/B (which has also been termed GSDMEc in other work). This latter gene is the ancestral gene for bird/reptile/amphibians GSDMA that, in turn, is the ancestral gene to mammal GSDMA, B, C, and D. Prior work had shown that more ancient animals have only GSDME and PJVK, and in these animals caspase-3 and caspase-1 can both independently cleave GSDME. Prior work had also shown that in birds/reptiles/amphibians, GSDMA is cleaved by caspase-1, and GSDME is only cleaved by caspase-3. Here, the authors demonstrate that the more ancient C. milii gene is similar to the bird/reptile/amphibian GSDMA in that it is cleaved by caspase-1. They further demonstrate that C. milii does not encode inflammasomes that would activate CmiCASP1, and instead this caspase is an LPS sensor through its CARD domain analogous to mammal caspase-4/5/11. The data supporting these conclusions are convincing, and could be strengthened by primary cell studies from C. milii in future studies. They further provide evidence that this gasdermin can kill bacteria directly, but the data supporting this conclusion are incomplete.

Strengths:

The data demonstrating that CmiCASP1 is an LPS sensor via its CARD domain is thorough and convincing.

The data demonstrating that CmiCASP1 cleaves and activates GSDMA/B and that this causes pyroptosis is also thorough and convincing.

Weaknesses:

I think that the gene/protein referred to in this paper as GSDMA/B was in prior publications called GSDMEc (doi 10.3389/fcell.2022.952015). Is this correct? If not, the relationship or lack thereof to GSDMEc needs to be described. If the authors wish to rename the gene, this needs to be justified and discussed clearly. Also, a gene name with a slash is not typical and was initially confusing to me as it made me think the authors were referring to two different genes.

The authors do not have data from primary cells from C. milii to demonstrate that the LPS sensing by CmiCASP1 and the pyroptosis induction by GSDMA/B is relevant in the native cell types. This is a common limitation in publications that seek to study diverse animals where tools may not be available. This issue can be studied in future publications.

The ability of gasdermins to kill bacteria is controversial.

This bactericidal effect was first shown by the cited article Liu et al. 2016 from Judy Lieberman's lab. I reviewed that manuscript at Nature, and I implored the authors to remove that data from the paper because the experimental design had a high risk of not being physiologically relevant. Indeed, my lab had previously published that bacteria survive the process of pyroptosis and they must be killed by secondary efferocytic phagocytes attracted to the pyroptotic corpse (doi 10.1084/jem.20151613). I have continued to consider whether gasdermins could kill bacteria over the decade since that 2016 paper, and wrote a detailed argument describing how this is unlikely to be physiologically relevant in a recent review article (see Box 3 in doi 10.1038/s41564-026-02272-z).

In the author's current manuscript, the experiments performed show a very mild effect in the linear range of a reduction of perhaps 20% of the control bacteria in Figure 5A. This is a minimal effect compared to antimicrobial peptides, which will reduce colony-forming units by 99.9%. Take a look at the magnitude of effect in Figure 1 of an example paper looking at polymyxin or colistin killing of Acinetobacter (doi: 10.1128/AAC.00756-12), where CFUs are reduced by about 3 logs (1000-fold) in 30 minutes. The magnitude of effect in Figure 5A is not even 2-fold. Further, it would be very challenging to determine whether the concentration of gasdermin protein used in the assay is equivalent to the concentration that exists in cells. The methods section needs to be clearer to explain how many effective cell lysates of 293T cells were exposed to how many bacteria, because these concentrated lysates are of unspecified concentration.

Regarding cardiolipin binding, this is a lipid that has a small head group attached to 4 lipid chains, resulting in a cone-like shape with the polar groups at the cone tip and the lipids forming the wide cone base. As such, it creates a larger lipid area than polar area, thus naturally creating a curved membrane shape such that cardiolipin is in the leaflet of the interior of a curvature. Thus, in the mitochondria, it exists in the inner membrane in the mitochondria and allows for the bends that form the cristae, where it faces the surface that is concave (nicely diagrammed in Figure 1 of doi 10.3390/biom16010071). Similarly, cardiolipin enriches in the inner leaflets of membranes at the poles of rod-shaped bacteria to allow for the curvature of the membrane at the poles. Therefore, cardiolipin is not exposed in the outer leaflet of the outer membrane of Gram-negative bacteria; instead, the primary lipid in the outer leaflet is LPS.

A competing mechanism that could explain the results is that the opening of gasdermin pores in eukaryotic cell plasma membranes occurs concomitant with the generation of ROS from mitochondria. This could occur by gasdermins inserting into mitochondria, as supported by DOI: 10.1038/s41419-025-07760-4. The resulting ROS production due to mitochondrial dysfunction could cause the bactericidal toxicity seen in the cell extracts.

Reviewer #2 (Public review):

The authors investigate the mechanism by which a gasdermin pore-forming effector of the cartilaginous fish Callorhinchus milii, GSDMA/B (CmiGSDMA/B), is activated. This potentially provides information on the ancestral function of gasdermins, a class of proteins broadly important in human health and disease. By reconstituting components of this system in vitro using transfection models, they show that GSDMA/B is activated by cleavage by the caspase-1 homolog CmiCASP1, which directly senses lipopolysaccharide. This mechanism is broadly similar to the non-canonical pathway in mammals, wherein caspase-4/5/11 cleaves GSDMD upon cytosolic LPS sensing. The conclusions of these interactions are mostly well supported by data, but some aspects need clarification, and based on the experimental approaches, some of the broader interpretations have limitations that should be considered and further discussed.

A more detailed analysis and discussion on the differences between caspases with regard to their LPS-binding capacity would be valuable for comparison. The analysis of Figure 4A and 4B effectively shows that there are similarities between CmiCASP1 and some of the studied mammalian caspases. However, part of this analysis is to make the point that some caspases do not bind LPS, and it would benefit from the inclusion of additional relevant LPS-insensitive caspases to show the connection between the chondrichthyan caspase residues highlighted and LPS-binding dependence. Modeling the LPS binding site (such as in Figure 1F) would further help clarify whether these are appropriately positioned for coordination, or for non-conserved residues, if there are alternate binding modes thought to have biological relevance.

The authors note that two different cleavage products are formed, with variable function, which is of interest. The results of Figure 2b suggest that the 241A mutation (blocking the 30 kDa product) increases processing to the larger 35 kDa product, while the 288A mutation decreases processing of the 30 kDa product (also blocking the 35 kDa form). Paired with the lysis data (Figures 2C-2E), its not clear that the 35 kDa product is anything but inactive, but this is quite different from the observations in the experiments with each form (Figure 3N-3Q). A more detailed kinetic and stoichiometric analysis between full-length, N241, and N288 would be important for clarifying the potentially interesting observation of N288 inhibition of N241.

The mechanism of bacteriocidal activity proposed in the final model and by the experiments of Figure 5 would benefit from further development to support the claim. The experiments do not adequately address whether, during pyroptosis, there is release of N241-like fragments that can kill bacteria. Figure 3G would indicate that it stays in the cell, either in the membrane or mitochondria, and it's not clear there would be circumstances where it could be extracted from it to then target bacteria. Figure 5B might require additional explanation and analysis, but the appearance of similar colonies between conditions would appear to support that there is not measurable antibacterial activity. More rigorous support would come from differences in bacterial killing by knockout Callorhinchus cells, but a minimal step to demonstrating the relevance would be MIC assays, and connecting the effective concentration with one that could naturally occur in Callorhinchus.

Broadly, the methods of reconstitution of components of this system demonstrate the sufficiency of LPS for activating Casp1, and Casp1 for activating GSMDA/B. However, in more established models, it is clear that there are inhibitors, feedback mechanisms, alternative pathways, and regulation that could render these interactions irrelevant in Callorhinchus. For example, it's not clear where Casp1 and GSMDA/B are ever expressed in the same cell, at quantities sufficient for this mechanism, or that Casp1 doesn't induce more rapid death by acting on something other than GSDMA/B, or that Casp1 is irrelevant because GSDMA/B can be activated more readily by another mechanism. Therefore, while the insights into the evolution of the individual factors of GSDMA/B and Casp1 are interesting and of potential value to the field, reconstituting choice components by transfection of human HeLa and HEK293 cells introduces limitations to how far these experiments can be interpreted as a system. The abstract, for example, states this is a "pyroptosis pathway in cartilaginous fish". However, for all the interest of these data in the evolution of these proteins, the evidence falls short of this. It establishes a biological potential, but it's not clear this is an active pathway in fish.

Reviewer #3 (Public review):

In this manuscript, the authors focused on Callorhinchus milii GSDMA/B (CmiGSDMA/B) and its upstream inflammatory caspase, CmiCASP1, and revealed that LPS directly engages the CARD domain of CmiCASP1, triggering its activation, which subsequently promotes the proteolytic cleavage of CmiGSDMA/B, yielding two N-terminal fragments with opposite functions. Moreover, consistent with GSDMD, the functional N241 of CmiGSDMA/B can mediate pyroptosis and exhibit bactericidal activity against Gram-negative bacteria in vitro. Based on these observations, the authors clarified that they uncovered an ancestral LPS-sensing CASP1-GSDMA/B axis in cartilaginous fish; however, several issues should be addressed.

(1) The evidence for direct and functional LPS sensing by CmiCASP1 remains insufficient. Although the authors propose that LPS directly binds the CARD domain of CmiCASP1 to trigger a non-canonical inflammasome-like pathway, the current support mainly comes from pull-down, competition, and in vitro cleavage/activity assays. These results are suggestive but do not yet establish a direct, specific, and physiologically relevant interaction. Additional quantitative binding and specificity analyses are needed to exclude indirect association, aggregation, or other assay artifacts. Therefore, the claim that CmiCASP1 functions as a bona fide direct LPS sensor appears overstated at this stage.

(2) The proposed antagonistic role of N288 is not yet convincingly supported. While the "dual-fragment antagonistic regulation" model is interesting, it currently relies mainly on overexpression/co-expression, co-IP, and localization analyses, which do not clearly distinguish a physiological inhibitory mechanism from a non-specific dosage or sequestration effect. Stronger support would require evidence for the relative generation and timing of N241 and N288, as well as quantitative data showing that N288 interferes with N241 membrane targeting, oligomerization, or pore formation. Testing the effect of selectively blocking D288 cleavage in the full-length protein would also strengthen this conclusion. At present, the antagonistic model remains premature.

(3) The physiological and evolutionary claims are stronger than the available evidence. Although the study shows that CmiCASP1 can cleave CmiGSDMA/B and that this module can be reconstituted in heterologous mammalian systems, these data do not demonstrate that such a pathway operates in elephant shark cells or tissues under physiological conditions. A similar concern applies to the antibacterial assays, which use HEK293T lysates rather than purified N241, making it difficult to exclude contributions from host-derived factors. The authors should either provide more direct evidence in a relevant chondrichthyan context or substantially tone down the evolutionary and physiological interpretations.

(4) The inhibitor data do not convincingly demonstrate suppression of CmiCASP1 activation. Although the authors state that Z-VAD-FMK blocks CmiGSDMA/B cleavage and pyroptotic phenotypes, Figure 1L and Figure 2H do not clearly show that CmiCASP1 activation or processing itself is inhibited. If CmiCASP1 remains processed in the presence of the inhibitor, it becomes unclear whether Z-VAD-FMK blocks CmiCASP1 activation, catalytic activity, or only downstream substrate cleavage. This point should be clarified with more direct biochemical evidence.

(5) The dosage control for GSDM-derived proteins in the antibacterial assays is unclear. In Figure 5, antibacterial activity is tested using HEK293T lysates or concentrated supernatants containing full-length CmiGSDMA/B, N241, or N288, but it is not clear how protein amounts were normalized across conditions. Differences in expression, stability, or recovery could substantially affect the apparent antibacterial activity. The authors should clarify how input was controlled and ideally provide quantitative normalization or matched-concentration assays to support the comparison.