The ASC Speck and NLRP3 Inflammasome Function Are Spatially and Temporally Distinct

Although considered the ternary inflammasome structure, whether the singular, perinuclear NLRP3:ASC speck is synonymous with the NLRP3 inflammasome is unclear. Herein, we report that the NLRP3:ASC speck is not required for nigericin-induced inflammasome activation but facilitates and maximizes IL-1β processing. Furthermore, the NLRP3 agonists H2O2 and MSU elicited IL-1β maturation without inducing specks. Notably, caspase-1 activity is spatially distinct from the speck, occurring at multiple cytoplasmic sites. Additionally, caspase-1 activity negatively regulates speck frequency and speck size, while speck numbers and IL-1β processing are negatively correlated, cyclical and can be uncoupled by NLRP3 mutations or inhibiting microtubule polymerization. Finally, when specks are present, caspase-1 is likely activated after leaving the speck structure. Thus, the speck is not the NLRP3 inflammasome itself, but is instead a dynamic structure which may amplify the NLRP3 response to weak stimuli by facilitating the formation and release of small NLRP3:ASC complexes which in turn activate caspase-1.

Since NLRP inflammasome activity and speck assembly are concomitant and both require interaction between the NLRP and ASC, NLRP3:ASC specks are generally equated with the inflammasome (13). Studies using small molecules including colchicine, nocodazole, and others which prevent both speck assembly and inflammasome function support this view (12,18). However, biochemical analysis suggests that active NLRP1:ASC: caspase-1 complexes may be less than~0.45 mm in diameter (~700 kDa) (19), values generally accepted for NLRP3 complexes as well (20) and are much smaller than the 1-μm speck structure. Likewise, in vitro-assembled NLRP1 (15 Å) and NLRP3 (100 nm), inflammasome structures have an approximately 100-1,000-fold smaller diameter than the speck, different arrangements, and distinct stoichiometries (20,21). Furthermore, while NLRP3:ASC speck formation requires a few minutes, IL-1b processing is detected within 1 h (22). Such size, stoichiometric, and temporal disparities between the speck and the inflammasome suggest that these structures may have distinct functions. Nevertheless, the relationship between the inflammasome and the speck is still unclear. Whether such distinctions reflect separate functions for the speck and inflammasome is unanswered.
This study establishes that NLRP3-mediated caspase-1 activity (the inflammasome) is distinct and independent of the NLRP3:ASC speck. While not required and insufficient for NLRP3 inflammasome function, the NLRP3:ASC speck is a dynamic, spatially, and temporally distinct structure regulated in part by caspase-1. Furthermore, the speck lowers the NLRP3 inflammasome activation threshold, potentially amplifying the formation and release of much smaller inflammasome complexes. The implications of our data and the relationship between ASC speck structures and the NLRP3 inflammasome are discussed.

Transfection Conditions
For assays of stimulus-dependent NLRP3-inflammasome IL-1b maturation and release (inflammasome reconstitution), HEK293T cells were seeded (2.5 × 10 4 cells/well/ml) in 24-well plates, cultured overnight, and transfected with 40 ng CASP1, 200 ng IL1B, 8 ng ASC, and 100 ng of either an NLR or control (pcDNA3) plasmid. Under these conditions, specks are not evident and NLRP3 and ASC cannot be detected by immunofluorescence staining due to low levels of expression.
Observing macrophage-like, agonist-dependent speck formation in HEK293T cells either alone [using time-of-flight inflammasome evaluation (TOFIE)] or in conjunction with caspase-1 activity [using inflammasome and caspase-1 activity characterization and evaluation (ICCE)] requires higher levels of expressed NLRP3 and ASC. HEK293T cells (2 × 10 5 ) were seeded per well/ml in 12-well plates and, after overnight incubation, transfected with 100 ng NLRP3, 50 ng GFP-ASC, and either alone or with caspase-1 (20 or 50 ng). The empty vector, pcDNA3, was used to adjust the final amount of DNA to 1 mg prior to transfection.
For the evaluation of speck formation capacity by microscopy, 5 × 10 5 HEK293Ts were seeded per well/ml in

FRET and Caspase-1 Activation Assay
The specificity of caspase-1 is significantly increased if peptide flanking the caspase-1-specific sequence YVAD is increased by only four amino acid sequences (27). Thus, a double reporter (GFP-YVAD-RFP) construct with the sequence "YVAD" separating GFP and RFP was generated, which is expected to have higher specificity than FLICA (Supplementary Figure S1). HEK293Ts (2 × 10 5 ) were seeded per well/ml in 12-well plates, after overnight incubation, and transfected with 100 ng NLRP3, 100 ng GFP-ASC, 400 ng CASP1 (WT or mutant), and GFP-YVAD-RFP. As a positive control for FRET, cells were transfected with all the above stated plasmids except caspase-1, and for negative control, cells were transfected with GFP and RFP expressed on different plasmids instead of GFP-YVAD-RFP expressing plasmid. Transfections were carried out at a constant DNA concentration of 1 mg/well. The empty vector, pcDNA3, was used to adjust the final amount of DNA to 1 mg. After overnight incubation, cells were either analyzed by microscopy to evaluate the distribution of GFP-YVAD-RFP reporter. To measure caspase-1 activation using FRET, cells were harvested after overnight incubation and fixed using 4% paraformaldehyde (PFA) as described above. The cells were acquired on LSRII flow cytometer equipped with 405, 488, and 642 nm lasers with longpass filter of 505 nm and band-pass filters of 450/50, 530/30, and 660/20 nm. Acquisition was done using BD FACSDiva software. Data were analyzed using FlowJo. Samples were gated to exclude debris and cell doublets. Singlet population was further gated for GFP staining. A stop gate of 10 4 cells was set on the GFP-positive gate. GFP+ cells were plotted for RFP versus GFP staining. Since the flow cytometer lacked lasers to excite RFP, only the source for RFP signal was through GFP emission leading to FRET. Thus, events in FRET channel were normalized to 100 for the GFP-YVAD-RFP-transfected sample and normalized to 0 for the GFP+RFP-transfected samples. FRET was calculated using the following formula: For caspase-1 activity measurements, the caspase-1transfected sample was normalized to 100 and the GFP-YVAD-RFP-alone-transfected sample was normalized to 0. Caspase-1 activity was measured using the following formula:

Time-of-Flight Inflammasome Evaluation
A portion of the samples prepared for ImageStream ® X Mark II (ISXII) analysis (below) was diluted to 200 ml in ISXII cell suspension buffer and acquired on an LSRII flow cytometer equipped with 405, 488, and 642 nm lasers with 505 nm longpass filter and band-pass filters of 450/50, 530/30, and 660/20 nm. BD FACSDiva software was used for acquisition and data were analyzed using FlowJo. Samples were gated to exclude debris and cell doublets. The single-cell population was further gated for GFP staining. A stop gate of 10 4 cells was set on the GFP-positive gate. The percentage of cells containing an ASC speck was determined by analyzing the height (H), width (W), and area (A) of the GFP pulse area (high H:A and low W:A indicate speck-positive cells) as described previously by Sester and group (28,29).

Colocalization
Colocalization of proteins with a resolution of 0.5 mm per pixel in the X-and Y-axis can be determined with the AMNIS ISXII using Bright Detail Similarity (BDS index scores, a logtransformed Pearson's correlation coefficient) (30). A BDS index score ≥1.5 is considered significant for colocalization (31). The BDS index was calculated using the ASC speck mask (mask 14) and FLICA spot mask (mask V) on a pixel-by-pixel and cell-by-cell basis with the Amnis IDEAS Colocalization Wizard tool. Using events from population S as an internal negative control, a threshold BDS index of 1.5 was set to determine colocalization of speck and active caspase-1 (FLICA) aggregates.

Microscopy
For microscopy, THP-1 cells were fixed and stained as described for ICCE. NLRP3 was stained using Cryopyrin (H-66) rabbit polyclonal IgG (1:1,000) (Santa Cruz; Cat. # sc-66846) or mouse monoclonal ANTI-FLAG ® M2 antibody (1:1,000) (Sigma Aldrich; Cat. # F1804) while performing experiments in HEK293Ts. HEK293T cells were fixed with 4% PFA and washed three times with 1× PBS, dipped in distilled water, and mounted on a slide using 10 ml of Fluoro-Gel II mounting medium with DAPI (EMS; Cat. # 17985-51) and visualized using an Axio Observer Z1 fluorescence microscope (Zeiss). Images were acquired using ZEN-Blue at an optimal setting to avoid saturation. Image acquisition was run at identical setting. Image analysis was performed using Fiji-ImageJ software. ASCpositive cells containing specks were counted manually from randomly selected fields acquired at ×20 or ×60 magnifications. The percentage of speck-containing cells was calculated as the fraction of ASC-positive cells containing specks. The percentage of cells containing GFP-RFP aggregates was calculated as the fraction of GFP-RFP-positive cells containing none, single, or multiple aggregates.

Quantification and Statistical Analysis
At least three independent experiments were performed with two/three technical repeats unless otherwise indicated. The statistical tests used for data analysis are stated in the figure legends. A p-value of ≤0.05 was considered statistically significant. All statistical analyses were performed using Prism 7.

Colchicine Prevents Speck Formation But Not Inflammasome Activation
Treatment of macrophages with nigericin concentrations above 3 μM stimulates maturation and release of IL-1b even in the presence of colchicine, a microtubule inhibitor that blocks speck assembly (32). Therefore, we considered that higher concentrations of nigericin might activate the inflammasome in the absence of the speck, but whether colchicine blockade of NLRP3:ASC speck assembly is overcome under these conditions is untested. PMA-treated THP-1 macrophages were stimulated with increasing concentrations of nigericin in the presence or absence of colchicine. In the absence of colchicine, nigericin stimulation led to maximal IL-1b processing (~30-fold over untreated controls) irrespective of the dose of nigericin ( Figure 1A). At lower concentrations of nigericin (1 and 2.5 μM), IL-1b maturation and release was nearly blocked by colchicine treatment, but at 5 mM, nigericin overcame this blockade ( Figure 1A). Although somewhat reduced, the IL-1b response to 5 μM nigericin in the presence of colchicine was approximately 20-fold higher than the untreated control and 65% of that was seen without colchicine. While speck formation was expected to accompany inflammasome activity, less than 1% of colchicine-treated cells had ASC specks despite stimulation with 5 μM nigericin ( Figures 1B, C). Thus, IL-1b processing can occur in the absence of specks, and speck formation may not be essential for the NLRP3 inflammasome response.
Prevailing dogma equates speck assembly and inflammasome activation of caspase-1. However, increased speck frequency does not correlate well with the levels of released IL-1b. For example, (G) THP-1 cells were stimulated with PMA as in (E) and lysed with 0.09% Triton X-100. IL-1b lysates were diluted 1:5 in assay buffer and measured by ELISA. Data represented as mean ± SEM for a minimum of three independent experiments, unless otherwise mentioned. **p < 0.01, ***p < 0.001, ****p < 0.0001 for comparison with respective untreated controls; two-way ANOVA followed by Dunnett's (A) multiple comparison tests. # p < 0.05, ### p < 0.001 Student's t-test for comparison between colchicine-treated cells stimulated with or without 5 mM nigericin. PMA-treated THP-1 macrophages stimulated with 5 mM nigericin yielded approximately three-fold more speckcontaining cells than those stimulated with 1 mM despite comparable IL-1b release (data not shown). This lack of correlation between specks and the magnitude of the IL-1b response together with the near complete blockade of IL-1b when colchicine is used with lower doses of nigericin suggests the possibility that ASC specks might improve, or even maximize, caspase-1 activation by suboptimal stimuli. If correct, the IL-1b response to nigericin is expected to be dose-dependent in the absence of specks. PMA-treated THP-1 macrophages were stimulated with colchicine and stimulated with increasing concentrations of nigericin (1.0 to 7.5 μM). In the presence of colchicine, specks were not observed in cells treated with any concentration of nigericin ( Figure 1D) and little IL-1b was produced after stimulation with up to 2.5 mM nigericin ( Figure 1E). Interestingly, in the absence of specks (i.e., with colchicine), supernatant IL-1b levels were dependent upon the dose of nigericin for concentrations between 2.5 and 7.5 mM. A similar dose dependency was observed with LPS-primed THP-1 ( Figure 1F). In the presence of colchicine, higher concentrations of ATP and MSU were not evaluated because they induced cell death. We confirmed that the IL-1b ELISA kit used does not detect uncleaved pro-IL-1b ( Figure 1G) released upon cell death. Thus, only IL-1b actively cleaved following stimulation is detected in these assays. While these data demonstrate that the speck is not required for caspase-1 cleavage of IL-1b, they do not rule out the ASC speck as a site of caspase-1 activity. Furthermore, these data also suggest that ASC specks facilitate and maximize caspase-1 cleavage of IL-1b, perhaps through lowering the agonist threshold required for NLRP3 inflammasome activation.

Caspase-1 Activity Is Distal to the ASC Speck Organization Site
As IL-1b processing can occur in the absence of ASC specks, the site of caspase-1 activation may be distinct from the speck structure when one is present. We previously demonstrated that ASC specks and caspase-1 activity can be visualized simultaneously in cells using imaging flow cytometry (24). Although our focus was on cells with specks and organized caspase-1 activity, we observed that active caspase-1 was distributed throughout the cytoplasm. We manually reevaluated micrographs from these experiments to consider cells with individual specks and active caspase-1 regardless of organization. Inflammasome-reconstituted cells receiving either 20 or 50 ng of caspase-1 can be divided into four subpopulations based on their distribution of specks and active caspase-1 ( Figure 2A) Figures 2B, C).
Importantly, the amount of transfected caspase-1 did not significantly alter these distribution patterns or the frequency of cells exhibiting each pattern ( Figures 2B, C). Approximately 15%-20% of the cells had features of more than one distribution pattern for speck and active caspase-1 and were quantified as overlapping sets ( Figures 2B, C). Populations with ASC specks and organized caspase-1 (S and C) account for 22%-30% of the cells (Figures 2B, C). In population S (2% to 4%), caspase-1 activity was organized and located distally to the ASC speck, suggesting an unexpected spatial disconnect between active caspase-1 and the speck in this subpopulation. Viewing a 3D event in 2D may cause two separate events in distinct planes to appear colocalized. To evaluate colocalization of ASC and active caspase-1 in cells with concomitant staining (population C) using the ImageStream, we used the BDS index function and IDEAS Colocalization Wizard Tool (see Methods). A BDS index of 3.0 indicates a high degree of colocalization (both objects are in the same focal plane), where an index of 1.0 is not colocalized (not in the same focal plane). Using events from population S as an internal negative control, the BDS index threshold was set to 1.5. The speck and active caspase-1 (FLICA) are considered colocalized in cells with scores ≥1.5 ( Figure 2D). For nigericinstimulated cells, the mean BDS index for the ASC speck and caspase-1 activity was approximately 0.25, irrespective of the amount of caspase-1 transfected ( Figure 2E), indicating a lack of colocalization. Surprisingly, colocalization of the ASC speck and caspase-1 activity (BDS index ≥ 1.5) occurred in fewer than 2% of cells receiving 50 ng of caspase-1 and none of those receiving 20 ng ( Figure 2F). Thus, even in population C, where specks and caspase-1 activity appear concomitant, ASC specks and the site of active caspase-1 are spatially separated. Indeed, approximately 90% of cells with specks had diffuse caspase-1 activity (populations D and N), and of these, roughly a third were devoid of caspase-1 activity at the speck itself ( Figures 2B, C).
Collectively, these data unexpectedly suggest that in cells with a discrete NLRP3:ASC speck, caspase-1 activity is almost exclusively separate from the speck structure.

Speck Formation Is Not Equivalent to IL-1b Processing
Our analysis suggests that NLRP3:ASC speck and the inflammasome may be distinct entities. Alternatively, if the speck is the site of caspase-1 activity, speck formation should be proportional to IL-1b production. While studies demonstrate that small molecule NLRP3 inhibitors or mutations in inflammasome components block both specks and IL-1b maturation (18,20,33), we found no study reporting proportionality between speck frequency and IL-1b maturation. NLRP3 contains a C8:C108 disulfide bond suggested to be important for NLRP3-ASC interaction (34). Individually, inflammasome activation does not strictly require cysteine 8 or 108, but IL-1b production is increased by C108A and C108S single-point mutants and decreased by C8/108 double mutants (25,35). Since ASC speck formation by these mutants is expected to reflect their inflammasome activity, we evaluated IL-1b release and speck formation in inflammasome-reconstituted  Figure 3A). However, with nigericin stimulation, IL-1b production increased approximately 1.5× for single-point C108 mutants and decreased by 2× for C8/108 double mutants relative to wtNLRP3, while speck frequency was diminished by approximately 25% for all C8 and C108 mutants ( Figure 3B). Thus, IL-1b production by C8/108 mutants correlated closely with ASC speck formation (IL-1b:speck ratio of about 1) in response to Fn U112 infection ( Figure 3C), but was not correlated  when nigericin was used. Indeed, despite all C8/108 mutants having uniformly diminished speck frequency, single-point C8 mutants decreased nigericin-induced speck formation but had little impact on IL-1b, where C108 mutants increased IL-1b production despite diminished speck-forming capacity. C8/108 double mutants markedly decreased IL-1b but had little further impact on nigericin-induced ASC speck formation. Thus, speck frequencies do not simply parallel or equate to inflammasome activity (IL-1b production). That IL-1b processing can increase or decrease irrespective of ASC speck formation is consistent with our hypothesis that the site of inflammasome IL-1b processing is distinct from the ASC speck structure.
Our results with NLRP3 mutants provide compelling evidence that speck formation and inflammasome function are likely disconnected. Although speck formation follows stimulation with inflammasome agonists and precedes IL-1b elaboration, they may not be the site of IL-1b processing. If correct, other indications may be evident during agonist stimulation of wild-type NLRP3. Reconstitution of agonist-responsive NLRP3 inflammasomes is accomplished by transfecting limited quantities of inflammasome components encoding plasmids (36). In such experiments, ASC specks are rarely observed beyond background levels despite robust IL-1b production comparable to similarly stimulated macrophages (data not shown). Indeed, to evaluate agonist-dependent induction of specks similar to that seen in macrophages, higher expression of NLRP3 and ASC is required (24,28,36). Therefore, we compared ASC speck formation capacity (TOFIE) and IL-1b processing (inflammasome reconstitution assay) in HEK293Ts in response to agonists thought to use different activation pathways including pore formation/K + efflux (nigericin), phagosome rupture (Fn infection and MSU stimulation), and ROS generation (H 2 O 2 ). In cells expressing NLRP3, nigericin treatment induced a 15-fold increase in IL-1b processing and an approximately three-fold increase in speck formation ( Figure 4A). As expected, no significant speck formation or IL-1b production occurred in unstimulated cells or those lacking NLRP3. Thus, speck formation and IL-1b processing are positively correlated for nigericin stimulation. However, Fn infection induced speck formation in the absence of NLRP3, but no IL-1b processing above background was observed ( Figure 4B). This result is consistent with a previous work reporting that specks form in the absence of NLRP3 but lack caspase-1 activity (37). In cells expressing NLRP3, Fn infection induced a robust IL-1b release without a significant further increase in speck frequency ( Figure 4B). Thus, for Fn infection, IL-1b production (inflammasome function) appears disconnected from speck formation. Fn-induced inflammasome function, but not speck formation, is dependent on NLRP3. Surprisingly, neither H 2 O 2 nor MSU induced speck formation, but significant levels of IL-1b were produced (Figures 4C, D). In these experiments, approximately 25% of the cells contained a speck, but no IL-1b was produced above controls without stimulation ( Figures 4C, D). For H 2 O 2 and MSU, speck formation does not equate directly with NLRP3 inflammasome activation. Thus, in inflammasomereconstituted cells, agonists like H 2 O 2 and MSU may activate the NLRP3 inflammasome without inducing specks.
As our H 2 O 2 and MSU results were surprising, we considered that specks induced by these agonists may have been formed but rapidly degraded prior to enumeration. To evaluate this alternate hypothesis, speck formation was tracked for 60 min using increasing concentrations of H 2 O 2 or MSU in cells expressing NLRP3 and ASC. H 2 O 2 induces IL-1b approximately 60 min after stimulation ( Figure 4E). As expected, speck formation increased over time and with escalating nigericin stimulation ( Figure 4F). Interestingly, no concentration of H 2 O 2 tested induced speck formation above basal levels within 60 min, and no decrease in basal speck frequency was evident ( Figure 4F). Results with MSU stimulation were essentially identical ( Figure 4G). Consistent with the transfected cell data, neither H 2 O 2 nor MSU induced specks in THP-1 cells (Supplementary Figure S4). Thus, rapid speck formation followed by degradation does not account for the absence of specks following H 2 O 2 or MSU stimulation. Collectively, these data suggest that even when wild-type NLRP3 is considered, speck formation is not equivalent to IL-1b processing, and thus, the inflammasome may be distinct from the speck.

Specks Are Dynamic and Negatively Correlated With IL-1b Release
Our data above strongly suggest that the speck is not necessary for inflammasome activity and that specks may function to instead facilitate inflammasome formation. As active caspase-1 is not localized to the speck, it follows that inflammasome complexes might form outside the speck or alternately be shed from the speck upon caspase-1 activation. Since both the speck and the inflammasome require NLRP3:ASC interaction, envisioning how to demonstrate the formation of inflammasome structures outside the speck or shedding of such complexes is problematic. If inflammasomes form exclusively outside the speck, the speck is expected be a static structure. In contrast, if the speck is shedding inflammasomes, its structure should be dynamic (e.g., size may be . Data represented as mean ± SEM for two independent experiments (F, G). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 for comparison with respective untreated; two-way ANOVA followed by Sidak's multiple comparison tests. altered with time and/or activation). To evaluate changes in ASC speck size during NLRP3 activation, we measured speck area in nigericin-stimulated cells expressing GFP-ASC, FLAG-NLRP3, and caspase-1 using our imaging flow cytometry-based ICCE assay (24). Both nigericin treatment and increasing the amount of available caspase-1 significantly decreased speck area ( Figure 5A). The frequency of ASC speck-containing cells also decreased with increased caspase-1 expression in the absence of stimulation ( Figure 5B). These inverse relationships demonstrate that caspase-1 activity negatively regulates both speck area and speck formation and strongly suggest that ASC specks are dynamic in nature. Since caspase-1 activity was negatively correlated with speck frequency in transfected cells, we also examined this relationship in macrophages. Nigericin was selected because neither H 2 O 2 nor MSU induced specks in THP-1 cells (Supplementary Figure S4), and nigericin-induced speck formation appears to be directly correlated with IL-1b processing ( Figure 4A). No significant LDH release was detected over a 6-h time course ( Figure 5C). PMAprimed THP-1 cells were activated with nigericin, and speck (A) Cells were analyzed for the presence of ASC speck and active caspase-1 (positive for FLICA staining) from the double-positive gates as analyzed by ICCE. The area of the speck mask was calculated by ICCE. Data represented as mean ± SEM for a minimum of three independent experiments. **p < 0.01, ****p < 0.0001 for comparison with respective untreated, two-way ANOVA followed by Sidak's multiple comparison tests. #### p < 0.0001 for comparison with respective untreated, two-way ANOVA followed by Dunnett's multiple comparison tests. (B) Cells positive for GFP-ASC were analyzed for the presence of ASC specks by ICCE. Data represented as mean ± SEM for a minimum of three independent experiments. **p < 0.01, ****p < 0.0001 for comparison with 0 ng transfected casp-1; # p < 0.05, ### p < 0.001 for comparison with 50 ng transfected casp-1; +++ p < 0.001 for comparison between 50 and 100 ng transfected casp-1; one-way ANOVA followed by Tukey's multiple comparison tests. (C) Percentage LDH release compared with 0.09% Triton X-100-lysed cells. (D) Line graph showing the frequency of ASC speck+ cells (red) and the release of IL-1b (black). Data are normalized to maximal response of speck frequency and released IL-1b over the time course of the experiment. Data represented as mean ± SEM for a minimum of three independent experiments. *p < 0.05, ***p < 0.001, for comparison of IL-1b release with untreated control (0 h); # p < 0.05, ## p < 0.01, #### p < 0.0001 for comparison of speck frequency with untreated control (0 h); one-way ANOVA followed by Sidak's multiple comparison tests. formation and IL-1b release were followed for 6 h; beyond 6 h, cells begin to detach from the plate complicating the assessment of speck formation. Specks were induced by nigericin, but their frequency was cyclical over time ( Figure 5D) and reached maximum around 1 h. Speck frequency declined to near basal levels over the next 2 h, increased to about 50% of the maximum by 5 h post-stimulation, and then again decreased. IL-1b release was also cyclical. IL-1b levels increased over time but did not reach a maximum until around 3 h. By 5 h, IL-1b levels had declined to about 50% of the maximum before increasing to near maximum levels around 6 h. Importantly, ASC specks were not detected outside the cells (data not shown). While both speck formation (frequency) and IL-1b were cyclic in nature, they were out of phase, with IL-1b increasing after peaks of speck formation. Although specks precede IL-1b production, consistent with specks facilitating the production of IL-1b, the temporal lag between speck frequency and IL-1b levels (2 h for the first cycle) reveals that peak caspase-1 activity likely follows well after peak speck formation (see Discussion). Under in vitro conditions, the Vmax of caspase-1 in an NLR complex is around 2 nM/min (21). At this rate, the quantity of IL-1b present at 3 h (5 ng/ ml;~0.3 nM) should have been converted in less than 1 min, i.e., with no temporal lag. Thus, it is likely that IL-1b processing does not occur simultaneously with the formation of the speck, but instead develops gradually over time following speck formation.

Small Inflammasome Complexes Are Shed From the Speck
The physical and temporal separation of caspase-1 activity from specks suggests that inactive NLRP3:ASC:caspase-1 complexes might be shed from the speck with activation of caspase-1 and processing of IL-1b occurring sometime later. If correct, cells with multiple sites of caspase-1 activity (with or without some diffuse activity) are expected to be abundant. Therefore, we evaluated the frequency of such cells in our HEK293T data from Figure 2.
While not direct evidence, the substantial number of cells with the expected phenotype is consistent with the shedding of inflammasome complexes from the speck. Speck-independent activation of small inflammasome complexes could also explain the observed pattern of caspase-1 activity. Indeed, small "death complexes" have been implicated in pyroptosis (caspase-1dependent cell death) and IL-1b processing after Aim2 and NLRC4 inflammasome activation in macrophages (33,38).
To better distinguish between these possibilities, we attempted to address whether caspase-1 is activated in the vicinity of the speck or at some distance using various caspase-1 mutants (Supplementary Figure S2A). Cleaved caspase-1 p35/ p10 is the active form and remains complexed with NLRP3:ASC until further cleavage generates inactive p10/p20 tetramers which leave the complex (39). Mutation of caspase-1 aspartic acids (D) 103 and 119 to asparagine (N) in caspase-1 UKN12 prevents inactivating cleavage events and the dissociation of active caspase-1 (p35/p10) from the NLRP3:ASC complex. UKN1234 has mutations (D!N) in all four aspartic acid cleavage sites preventing its activation. Mutation of caspase-1 cysteine 285 to alanine alters the active site reducing activity (DEAD). Although FLICA staining appears to be caspase-1 specific ( Figure 6C), others have suggested that FLICA reagents are promiscuous (40). Thus, for these experiments, the FRET-reporter GFP-YVAD-RFP was used to detect active caspase-1 (see Methods; Supplementary Figure S1). Wild-type and mutant caspase-1 activity measured using the GFP-YVAD-RFP reporter was comparable to that reflected by secreted IL-1b levels, suggesting similar kinetics and specificity for both methods (Supplementary Figures S2B, C). Localization of GFP-RFP signal to a single perinuclear complex is expected if caspase-1 activation is occurring at the ASC speck. In contrast, multiple GFP-RFP aggregates would indicate that caspase-1 is activated outside the speck ( Figure 6D). Active inflammasomes were reconstituted in HEK293T cells expressing wild-type or mutant caspase-1 in the presence of the GFP-YVAD-RFP reporter and evaluated for their GFP-RFP staining pattern by wide-field microscopy (single, multiple, or diffuse) ( Figure 6D). No aggregates were observed in cells transfected with GFP-alone, RFP-alone, GFP with RFP, or the GFP-YVAD-RFP reporter ( Figures 6E, F), indicating that these constructs do not spontaneously aggregate. With wild-type caspase-1, multiple GFP-YVAD-RFP complexes were present in 25% of the cells, while 50% had a single complex, and 25% had diffuse staining ( Figures 6D-G). Using the activatable, non-released caspase-1 mutant UKN12 did not alter these frequencies. Both types of complexes were absent when caspase-1 was inhibited with a caspase-1-specific inhibitor (YVAD) or a pan-caspase inhibitor (ZVAD) and when the enzymatically inactive caspase-1 DEAD or uncleavable caspase-1 mutant UKN1234 was used. Some caspase-1-transfected cells had a single speck-like GFP-RFP aggregate (Supplementary Figure S5A). The absence of active caspase-1 colocalization with specks ( Figures 2E, F) suggests this activity is near, but not coincident with, the ASC speck. In cells expressing the activatable, non-released UKN12 mutant, a zone of no GFP-RFP staining was observed at the likely position of the speck (Supplementary Figure S5B). This result is similar to that seen in the non-FLICA speck (N) cell population (Figure 2A). That UKN12 appears inactive at the likely position of the speck suggests that caspase-1 is activated after leaving the speck structure, not at the speck itself. Taken together, our data suggest that caspase-1 is most likely activated within small inflammasome complexes sometime after leaving the speck.

Caspase-1 Is Activated Outside the Speck
Our data provide the first clear demonstration that the speck is not the site of NLRP3-mediated caspase-1 activity. Instead, in both transfected cells and primary macrophages, NLRP3activated caspase-1 is almost exclusively present in multiple  cytoplasmic aggregates. This observation seemingly contradicts two studies suggesting that caspase-1 is activated at the NLRP3-ASC speck complex. In one, active caspase-1 was co-isolated with chemically crosslinked ASC (39). However, whether the caspase-1 activity observed in such studies was isolated from the speck itself or some other smaller complexes is uncertain as ASC crosslinking does not distinguish between these complexes. In the other study, active caspase-1 was colocalized to the center of the speck structure formed during NLRC4:NLRP3 inflammasome activation (41). Although we observed some cells with a similar coincident pattern, further analysis indicates vertical separation of the NLRP3:ASC speck and active caspase-1. In contrast to these studies, we observed that NLRP3 directed caspase-1 activity localized within multiple small cytoplasmic complexes. Consistent with our results, Akira et al. also observed multiple NLRP3-ASC complexes in IL-1b-producing cells stimulated with monosodium urate crystals using proximity ligation assay (12), although the absence of a speck in these assays was not mentioned or discussed. Furthermore, we also observed this localization pattern with the caspase-1 mutant UKN12. As UKN12 cannot be released from the inflammasome complex upon activation (39), these data confirm the presence of multiple cytoplasmic inflammasome complexes distinct from the speck.

Speck Formation Is Not Essential or Sufficient for Inflammasome Response
Formation of the singular ASC speck is very rapid, occurring within 3 min of NLRP3 activation (7,42). In addition, inhibiting speck assembly using small-molecule drugs (e.g., MCC950) or pyrin-related peptides (e.g., POP2) also impairs NLRP3 inflammasome activity (18,43). Such observations have been interpreted to indicate that the speck is equivalent to the inflammasome or, at least, an initial step in inflammasome activation. However, despite this strong correlative evidence, no direct experimentation has confirmed that the speck is actually the inflammasome. We previously noted that nigericin activates caspase-1 without discernable ASC speck formation in NLRP3 inflammasomereconstituted HEK293T cells (24), leading us to question the requirement of specks for inflammasome function. In the present study, the sterile agonists MSU and H 2 O 2 also induce IL-1b processing without speck formation in HEK293T cells. Compan et al. observed that stimulation with 200 mg/ml of MSU for 16 h induced speck formation in BMDMs (37). In our hands, THP-1 cells only required 2 h of stimulation with 150 mg/ ml MSU to produce IL-1b, and speck formation was absent (Supplementary Figure S4). Furthermore, the MSU used by Compan et al. was manufactured by a different company. Thus, these differences may account for the contrasting observations between the studies. Nevertheless, if MSU-induced specks form at 16 h in THP-1, this appears to be well after the initiation of IL-1b processing. Of note, and as mentioned above, Akira et al. observed multiple NLRP3-ASC complexes, as opposed to single specks, upon stimulation with MSU (12). H 2 O 2 induces inflammasome processing of IL-1b in THP-1 and other cells (44), but no studies report that H 2 O 2 induced speck formation. In THP-1 cells, H 2 O 2 did not stimulate speck formation (Supplementary Figure S4). Indeed, although inducing inflammasome activity, H 2 O 2 may inhibit speck formation as suggested during Streptococcus pneumoniae infection (45).
Moreover, speck assembly does not necessitate activation of caspase-1. We observed that Fn U112 elicits ASC speck in cells lacking NLRP3, but IL-1b processing is not detected. Others report that osmotic changes induce speck formation in the absence of NLRP3, but these cells also do not process IL-1b (37). Moreover, chimeric NLRP3 constructs containing the pyrin domain of either NLRP1 or 2 form specks but not an active inflammasome (Supplementary Figure S6) (25). Of note, reconstitution of speck formation and inflammasome function requires different amounts of transfected ASC. When agonistdependent inflammasome activation is reconstituted in HEK293T cells, specks are not evident; higher expression of ASC is required to observe agonist-dependent or independent speck formation (24), again supporting the conclusion that ASC speck formation and inflammasome function are distinct. Although surprising given the seemingly exclusive linkage between ASC specks and active inflammasomes, our data demonstrate that formation of an ASC speck is not required for NLRP3-dependent caspase-1 activation, nor is the ASC speck sufficient for caspase-1 activation.

The Speck Lowers the Agonist Threshold for NLRP3 Inflammasome Activation
If neither required for NLRP3 inflammasome activation nor sufficient, what is the function of the speck? Given the prion-like polymerization of ASC and caspase-1 along with the defined perinuclear localization of singular specks in cells, the ASC speck has been suggested to be a supramolecular organizing center (SMOC) (8)(9)(10). SMOCs are cell location-specific higher-order signaling complexes that concentrate signaling components to facilitate binary, all-or-none, responses through a cooperative assembly mechanism (8)(9)(10). An ASC SMOC is expected to recruit most if not all of the available ASC and caspase-1 to the structure, maximize proximity-induced activation of caspase-1, and reduce the threshold of activation. Thus, activation of caspase-1 in this fashion should also occur in an essentially binary, "on-off" fashion. Why caspase-1 is not obligately activated upon speck formation is puzzling, but our data offer some insight. NLRP1 or NLRP2 pyrin domain chimeras of NLRP3 form specks (Supplementary Figure S6A) and Fn U112 elicits ASC specks in cells lacking NLRP3. Yet, IL-1b processing is either absent or significantly reduced versus controls in both cases (Supplementary Figure S6B). The simplest interpretation is that a specific NLR is required to either recruit or activate recruited caspase-1. However, this conflicts with current paradigms. Oligomerization of ASC recruits caspase-1 (20), and once recruited, caspase-1 is expected to undergo proximity-induced activation (46). Paradoxically, our data indicate that NLRP3dependent caspase-1 activity is found in small cytosolic complexes distinct from the speck. Furthermore, activity of the caspase-1 UKN12 mutant (activatable, non-dissociating) was also not localized to the speck. As UKN12 cannot dissociate from NLRP3:ASC and UKN12, complexes with active UKN12 must have been activated either independently or after release from the speck (39). A potential resolution to this dilemma is that while small, non-speck NLRP3 inflammasome complexes can form without the speck itself, the speck facilitates the formation and release of small inflammasome complexes, thereby amplifying or lowering the threshold for inflammasome activation. Consistent with this idea, Dick et al. suggest that the NLRC4-induced ASC speck serves as a signal amplification mechanism through the formation of multiple caspase-1 activation sites (33). While exploration of this hypothesis is needed, the seminal studies of extracellular inflammasomes/specks do not appear to be contradictory (42,47).
Such a mechanism might also help explain the threshold effect of the speck for nigericin-induced NLRP3 inflammasome activation. When speck formation is inhibited with colchicine, low-dose nigericin fails to activate the NLRP3 inflammasome, but higher doses of nigericin still elicit IL-1b (32,48). In our hands, higher doses of nigericin elicited IL-1b maturation without speck assembly. Notably, when speck formation was inhibited by colchicine, nigericin induction of IL-1b was largely dose dependent. In contrast, without colchicine, IL-1b was elaborated at or near maximal levels irrespective of the nigericin concentration used. Thus, while not essential for NLRP3 inflammasome activation and unlikely the site of caspase-1 activity, the speck reduces the threshold required for agonist induction of full caspase-1 activity. Our findings are consistent with the NLRP3:ASC speck having SMOC-like behavior lowering the threshold of inflammasome activation. However, the physical separation of caspase-1 activity from the ASC speck suggests that the NLRP3:ASC SMOC facilitates the formation and release of small, cytosolically active inflammasomes. Although consistent with our data, this hypothetical mechanism requires further investigation.
The Speck Is a Dynamic Structure Regulated by Caspase-1 Whether the ASC speck is a SMOC that facilitates the formation and release of small inflammasome complexes is unclear, and demonstrating this presents numerous challenges. ASC specks are larger in cells with enzymatically inactive caspase-1 (caspase-1 DEAD) compared with cells with wild-type caspase-1 (49). Moreover, ASC specks formed by ASC alone are larger than NLRP3:ASC specks, and the size of NLRP3:ASC specks is further reduced after nigericin stimulation (24). In this study, NLRP3: ASC speck size was inversely proportional to caspase-1 activity and nigericin stimulation further reduced the size of these specks. Thus, speck size is dynamic and is regulated by caspase-1 activity likely resulting from triggering of NLRP3. This change in size suggests that NLRP3 activation of caspase-1 may allow shedding of component proteins from the speck and could account for the presence of multiple cytosolic sites of inflammasome caspase-1 activity. While a change in protein conformation and/or spatial organization could account for the reduction in speck size, such changes do not help explain the multiplicity of active caspase-1 sites.
Interestingly, specks and inflammasome function are not only spatially separated but temporally distinct as well. We found that FIGURE 7 | Schematic of the relationship between the NLRP3 inflammasome and speck. NLRP3-ASC specks form by rapid relocation of these proteins into a perinuclear toroidal structure. Specks allow inflammasome complexes to be activated at a lower stimulus threshold and thus facilitate an optimal inflammasome response to weaker signals. However, inflammasome complexes also form independently of the speck and stronger signals may not require or elicit speck formation. The speck likely sheds small inflammasome complexes where caspase-1 becomes active. As caspase-1 activity regulates the size of the speck, active caspase-1 may facilitate the release of small inflammasome complexes, acting as an intracellular feedforward loop maximizing IL-1b processing. speck frequency and IL-1b processing were negatively correlated over time. Furthermore, the relationship between specks and IL-1b processing was also cyclical with declines from peak speck frequency preceding IL-1b processing, suggesting an inverse relationship between the speck and inflammasome function. Furthermore, in inflammasome-reconstituted HEK293T cells, abundant IL-1b processing occurred in the relative absence of ASC specks, but prior to their peak (~7%) at 14 h ( Supplementary Figures S6C-F). Curiously, osmotic stressinduced ASC specks in bone marrow-derived macrophages also suggest a similar inverse relationship between the speck and inflammasome activity (37). The temporal disconnect between speck formation and inflammasome function provides additional evidence that speck function is dynamic. Moreover, ASC speck formation might not be irreversible as believed.
In summary, we conclude that ASC specks and NLRP3 inflammasomes are distinct entities and that not all NLRP3 inflammasome agonists trigger speck formation. While small active NLRP3 inflammasome complexes are dispersed throughout the cell following activation and caspase-1 activity is not present at the NLRP3:ASC speck, the speck appears to enhance this process by reducing the threshold of stimulus needed for maximal caspase-1 activation. Our data suggest that when specks are formed, NLRP3 inflammasome IL-1b processing and speck size are related in a dynamic and cyclical process regulated by caspase-1. We propose a refined hypothetical model for the relationship between the NLRP3 speck and inflammasome where the formation of an NLRP3:ASC: caspase-1 speck is energetically favored and releases small active inflammasome complexes but is not required to generate them ( Figure 7).

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
Conceptualization, methodology, validation, and writing: AN and JH. Investigation, formal analysis, and visualization: AN and TR. Resources, supervision, and funding acquisition: JH. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported in part by NIH Grants R01AI072259, P01AI056320 (subproject 2), and R03AI146596 to JH. Supplementary Figure S6 | Uncoupling of speck formation and IL-1b processing. (A) HEK293T cells were transfected with 1 µg myc-ASC and 1µg Flag-NLRP3 or a chimeric Flag-NLRP133 or Flag-NLRP233. After 18 hours cells, fixed and permeabilized cells were stained with anti-Flag (M2) and anti-ASC (Santa Cruz) and then analyzed for speck formation. Percentage of cells containing specks. Minimum of 140 cells were counted from 4 individual field of view; for NLRP233, 30 cells were counted. Data is presented as Mean ± SEM and analyzed using one-way ANOVA followed by Tukey's multiple comparison test. (B) IL-1b processing in inflammasome reconstituted HEK293Ts expressing NLRP3 or the indicated chimeras and treated/infected with stated NLRP3 agonists (See Methods). (C-F) HEK293T cells were transfected with 100 ng NLRP3, 100 ng GFP-ASC, 20 ng CASP1 and 200 ng IL1B allowing agonist-independent maturation and release of active IL-1b. At stipulated time-points, cells and culture supernatants were harvested. Cells were fixed and analyzed for GFP expression (C), Culture supernatant was analyzed for released IL-1b (D), ASC speck (E) and speck diameter (F).