Evaluation of human platelet granules by structured illumination laser fluorescence microscopy

Abstract Many roles of human platelets in health and disease are linked to their ability to transport and secrete a variety of small molecules and proteins carried in dense (δ-) and α-granules. Determination of granule number and content is important for diagnosis of platelet disorders and for studies of platelet structure, function, and development. We have optimized methods for detection and localization of platelet proteins via antibody and lectin staining, imaging via structured illumination laser fluorescence microscopy (SIM), and three-dimension (3D) image analysis. The methods were validated via comparison with published studies based on electron microscopy and high-resolution fluorescence microscopy. The α-granule cargo proteins thrombospondin-1 (TSP1), osteonectin (SPARC), fibrinogen (FGN), and Von Willebrand factor (VWF) were localized within the granule lumen, as was the proteoglycan serglycin (SRGN). Colocalization analysis indicates that staining with fluorescently labeled wheat germ agglutinin (WGA) allows detection of α-granules as effectively as immunostaining for cargo proteins, with the advantage of not requiring antibodies. RAB27B was observed to be concentrated at dense granules, allowing them to be counted via visual scoring and object analysis. We present a workflow for counting dense and α-granules via object analysis of 3D SIM images of platelets stained for RAB27B and with WGA. Abbreviation: SIM: structured illumination microscopy; WGA: wheat germ agglutinin; FGN: fibrinogen; TSP1: thrombospondin 1; ER: endoplasmic reticulum Plain Language Summary Platelets support blood clotting, wound healing, and other essential processes. These functions rely on the ability of platelets to transport and release small molecules like serotonin carried in dense granules and a wide range of proteins carried in alpha granules. Several conditions have been linked to abnormalities in one or more of platelet granule number, content, structure, and function. These conditions can be difficult to diagnose because platelet granules are so small they cannot be consistently resolved by conventional light microscopy, while higher power electron microscopy is not widely accessible. The goal of this study was to develop a method for counting and examining platelet dense and alpha granules without the need of electron microscopy. Key to this was the discovery that alpha granules can be reliably stained with the plant lectin wheat germ agglutinin, which has the advantages of being a smaller and less expensive molecule than the antibodies commonly used to detect alpha granule cargo proteins. We also establish that dense granules can be detected with high specificity via antibody staining of the membrane-associated protein RAB27B. We used structured illumination laser fluorescence microscopy to obtain high-resolution images of stained platelets. These were assembled into 3D renders using image analysis software, which was used to validate a protocol for rapidly counting granules within individual platelets. Our method supports the relatively rapid, accurate, and cost-effective assessment of platelet granules. We have already shown that it can confirm dense granule deficiency, and we anticipate that this approach will also prove useful in diagnosing and studying alpha granule abnormalities.

In addition to supporting the diagnosis of deficiencies in δ- [18] and/or α-granules [4,19], EM has long been used to image platelet structure and contents with high precision.Immunoelectron microscopy has proven useful in determining the intracellular localization of proteins (limited by the availability of suitable antibodies) including α-granule borne Von Willebrand Factor (VWF) [20], thrombospondin-1 (TSP1/THBS1), fibrinogen (FGN) [21], and the membrane protein P-selectin (SELP) [22].Fluorescence microscopy imaging has also been extensively used to examine platelet proteins [23], and while it provides lower resolution than EM, it supports more rapid acquisition of 3-dimensional (3D) image information for entire platelets.Fluorescence microscopy has been considerably advanced via high-resolution modalities including stimulated emission depletion microscopy [24] and single-molecule localization microscopy [25,26], used for detailed imaging of platelet morphology and behavior [27].3D structured illumination fluorescence microscopy (SIM) has proven especially useful in the acquisition of images suitable for visualization of proteins in platelet granules and other structures [10,[28][29][30], supporting diagnosis of deficiencies in granules [30] and their cargo [31].
Here, we evaluate platelet proteins localized to δ-and αgranules using SIM and 3D image analysis.We employed our established methods [28] for high-resolution imaging of nonactivated and intact resting human platelets to: 1) confirm their effectiveness for accurately localizing proteins associated with platelet secretory granules and other structures; 2) establish the utility of wheat germ agglutinin (WGA) as a stain for α-granules; 3) establish RAB27B as a δ-granule marker; 4) develop a workflow for rapid evaluation of δ-and α-granules via analysis of 3D images of platelets stained for RAB27B and with WGA.

Materials and methods
SIM imaging requires signals of high intensity above the background that remain consistent during repeated exposure to high levels of laser illumination.This makes SIM useful for examining relatively abundant proteins, as were those examined here (see Supplemental Table S1 for concentrations in human platelets) [32].Immunofluorescence microscopy also requires validated antibodies, and those used here (Supplemental Table S2) have been validated by one or more of published studies, the human proteome atlas (HPA) project [33], and direct comparisons [28].Individual images of non-activated platelets were used for voxel colocalization analysis based on the wellestablished thresholded Pearson correlation (PCC) and Manders (MCC) coefficients, and also for object-based image analysis, both performed with Imaris 9.9 software.Further details are provided in Supplemental Methods.

Localization of α-granule proteins
SIM imaging of a representative normal human platelet stained for tubulin, the α-granule membrane protein P-selectin (SELP), and cargo proteins TSP1 and VWF is shown in Figure 1a (see also Supplemental Figure S1 and Video 1).The images used in this study had an XY resolution of 100-125 nm (see Supplemental Methods).As discussed by Poulter et al. [29], this is a substantial improvement over the ~250 nm XY resolution available with the best diffraction-limited methods, allowing improved resolution of subcellular structures and increased accuracy of colocalization analysis.The intact circumferential tubulin cytoskeletal ring indicates that the cell was in a resting state at the time of fixation, as in all platelet images analyzed here [28].TSP1 is visible within puncta <300 nm diameter, and VWF is present in puncta and elongate structures resembling VWF tubules observed by EM [20].TSP1 and VWF puncta show clear proximity but little visible overlap, while SELP appears to form a matrix that winds around the cargo proteins, a pattern obtained with multiple antibodies (Supplemental Figure S2a).As done previously [28], co-staining with antibody pairs was used to establish baseline ranges for PCC and MCC indicative of strong overlap in voxel colocalization analysis (Supplemental Figure S2b,c).
The secretory granule proteoglycan serglycin/SRGN is a polyanionic carrier of platelet factor 4 (PF4, CXCL4) [34], recently shown to be essential for murine α-granule secretion [35].We observed a strong overlap of SRGN with TSP1 in human platelet α-granules and weaker overlap of both proteins with VWF (Figure 1b-d).SRGN, TSP1, and VWF showed similar overlap with SELP, and MCC data (Supplemental Figure S3a) indicate that SRGN and TSP1 are distributed throughout the granule matrix.The PCC and MCC data for several α-granule cargo proteins are summarized in Supplemental Figure S3b.They consistently indicate that VWF has a distinct localization relative to other cargo, as does osteonectin/SPARC (Supplemental Figure S4a), which is reported in immunoelectron microscopy studies to be concentrated near the granule's inner surface [36].FGN is reported to be present in human platelet α-granules [21,23,37] and in compartments associated with protein uptake [38].Consistent with this, we observed FGN both within and outside the SELP-defined matrix (Supplemental Figure S4b).

Wheat germ agglutinin (WGA) as a stain for α-granules
WGA binds several human platelet glycoproteins, including GP1BA [39], VWF, and FGN [40,41], and we have shown that fluorescently conjugated WGA is a useful rapid stain for human platelets and megakaryocytes [28].We observed that when permeabilized platelets were stained for >60 min with fluorescently conjugated WGA, multiple internal structures produced a signal Platelet granule counting with SIM 3 that was much stronger than that of the cell membrane.When the platelets were stained with WGA and for VWF, apparent overlap with α-granules was observed (Figure 2a), and colocalization analysis (Figure 2b) indicated relatively strong overlap of WGA with VWF and SRGN.Overlap with SELP was comparable to granule cargo proteins.The MCC values (Figure 2c) indicate that WGA staining is widely distributed within α-granules relative to staining for individual cargo proteins, consistent with WGA staining several of these proteins.WGA does not appear to strongly stain TSP1, which may be due to its unique glycosylation (C-mannosylation and O-fucosylation) in human platelets [42].These results indicate that WGA is a useful stain for human platelet α-granules, which offers advantages over antibodies (especially primary/secondary pairings) with regard to cost, size (comparable to nanobodies), and ease of use.

P-selectin (SELP) and non-granule proteins
The localization of several proteins with varying reported associations with α-granules was assessed relative to SELP (Supplemental Figure S5).The colocalization analysis (Supplemental Figure S6) showed overlap consistent with published reports for SELP/αgranule interaction with the membrane receptor CD61/ITGB3 [43] and the vSNARE VAMP8 [44].We have shown that NBEAL2 localizes near α-granules and interacts with SELP [10], and we also observed a strong overlap with the BEACH family protein LRBA, indicating that it may be associated with α-granules, which does not appear to have been previously reported.SELP showed relatively weak overlap with the endoplasmic reticulum (ER) protein VAPA and with the Ca 2+ ATPase SERCA3 (ATP2A3).SERCA3 has been previously localized by immunoelectron microscopy [45] and confocal fluorescence microscopy [46] to the platelet periphery, where with vacuolar H+-ATPase (detected here via the V1 subunit ATP6V1B2) it helps maintain an acidic calcium store [47] required for δ-granule secretion [48].These results indicate that our methods are useful for assessing granule-associated proteins and other aspects of platelet functional morphology.

RAB27B as a marker of dense (δ-) granules
The lipid anchoring Rab GTPase RAB27B is abundant in platelets (Supplemental Table S1), and it has been reported to be associated with both α- [49] and δ-granules [50].Evidence for the functional association of RAB27B with δ-granules includes: 1) in mice loss of expression causes a 50% reduction in δ-granules [50]; 2) absence of δ-granules and normal α-granules has been reported in human RAB27B deficiency [51].Attempts to count δgranules via SIM imaging of CD63-stained platelets have yielded variable results [30], consistent with CD63 not being exclusively localized to δ-granules [52].Our examination of cells co-stained with SELP indicates that RAB27B is concentrated at structures distinct from α-granules (Supplemental Figure S5d).The highintensity RAB27B positive spots are consistent with the appearance of δ-granules, which vary considerably in size and shape [53].Platelets co-stained for RAB27B and CD63 showed consistent visual overlap (Figure 3a) and moderate colocalization (Figure 3b; mean PCC = 0.62; s.d.0.08; n = 67 platelets from 2 normal donors).Both proteins showed low overlap with WGA and thus presumably α-granules.RAB27B staining in platelets from a normal donor and a Hermansky-Pudlak syndrome (HPS) patient with confirmed δ-granule deficiency (Figure 3c) showed multiple strong RAB27B spots in normal platelets, while the HPS cells had one or two spots.Visual spot counting (Figure 3d) gave a mean count for HPS platelets of 1.9 (range 1-6; s.d.1.23; n = 26), while the normal donor mean was 6.9 (range 2-12; s.d.2.28; n = 30), within the range for mean δ-granules/platelet reported for adult males (4.9-8.2) from a large cohort study using whole mount EM [15].The normal mean is also comparable to the 6.8 CD63positive structures per platelet reported for automated scoring of SIM images [30].It is interesting to note that the platelets lacking δ-granules sometimes showed RAB27B distributed throughout the cell, indicating that failure to form δ-granules may allow this protein to associate with other cellular structures.

Counting δ-and α-granules via object analysis
Platelet α-granules are relatively uniform in size and morphology, making them good candidates for evaluation via object-based image analysis.Images stained with WGA and for one or more cargo proteins were analyzed using the Imaris spot function.Spot counts per platelet (Supplemental Figure S7a) showed similar results for WGA, SRGN, and TSP1, with lower counts for VWF and SPARC consistent with their concentration in localized regions, some of which may have fallen below the detection threshold.FGN yielded highly variable and often higher spot counts, which is likely attributable to the detection of extragranular FGN (see above).The mean spot counts for SRGN, TSP1, and WGA were similar at 37.0 (range 14-69; s.d.14.4), 34.8 (range 18-54; s.d.9.8), and 38.2 (range 27-62; s.d.11.5), respectively.These values are somewhat below granule counts reported in EM studies, which may be attributable to the relatively lower resolution of SIM (especially in the z-axis) failing to distinguish between closely-associated granules.
3D rendered images can also be used to derive surface objects.To compare the effectiveness of WGA spot and surface objects in counting α-granules, both were scored for two sets of platelets from a normal donor.Concordant results were observed between the datasets (Supplemental Figure S7b), as well as strong correlation of counts for individual cells (Supplemental Figure S7c).This indicates that spot and surface object counting are equivalent for counting α-granules.Counts of RAB27B-stained surface objects also gave consistent results for replicate sets from multiple donors (Supplemental Figure S8), which yielded counts consistent with visual scoring (Figure 3d).Of the two methods, object analysis is considerably easier, faster and less subject to observer bias.
Based on these observations, we established a workflow for rapid counting of δ-and α-granules via surface object analysis of platelets stained with WGA and for RAB27B (Figure 4).As already mentioned, the α-granule counts are consistently low relative to expectations from EM studies of platelet sections.It is difficult to make an accurate assessment here.While EM studies are potentially more accurate, they typically examine sections from which granule scores are extrapolated to whole platelets, of which small numbers are analyzed.For example, 10 platelets from each of the three donors were examined in a recent 3D-EM study [54], which reported mean numbers of 50 α-granules and 5.5 dense granules per platelet.The same study reported a mean maximum/minimum α-granule dimensions of 251 and 297 nm, respectively, a size range where SIM provides a clear advantage in resolution relative to diffractionlimited microscopy.3D immunofluorescence microscopy studies are also difficult to assess owing to varied methodologies.For example, a recent study used automated analysis of SIM images to assess VWF puncta in large numbers of platelets from normal donors and VWD patients [31] and reported mean counts of VWF objects in normal platelets that were >50% lower than our VWF and WGA spot/object counts.This may highlight the importance of validating image acquisition and analysis methods when moving from assessment of individual cell images (as done here) to automated evaluation of entire image fields.

Conclusions
We have used SIM imaging to replicate observations from immunoelectron microscopy studies regarding the localization of cargo proteins within α-granules, including the concentration of VWF and SPARC in distinct regions.Consistent with mouse studies, we localized SRGN to the human α-granule lumen, where it presumably plays a similar role in stabilizing PF4 and other cargo.We established that WGA can serve as a rapid stain for α-granules and RAB27B as a marker of δ-granules, and we developed a workflow to count both granule types based on surface object analysis of SIM images of platelets stained with WGA and/or for RAB27B.This method may prove useful in situations where EM is not available, but SIM or other high-resolution immunofluorescence microscopy modalities can be applied to the diagnosis of platelet abnormalities.The ability of SIM to support high-resolution imaging and analysis is also applicable to studies of platelet protein localization and function.For example, we identify a possible association of LRBA with SELP and/or α-granules that is presently of unknown functional significance.

Highlights
SIM imaging confirms serglycin is present in human platelet alpha granules, where it likely stabilizes granule cargo and facilitates secretion.
RAB27B is concentrated at dense granule membranes, and staining for this protein supports the assessment of dense granules via SIM image analysis.
Wheat germ agglutinin stains alpha granules and is useful in rapidly assessing granule distribution, morphology, and number.

Figure 1 .
Figure 1.Imaging and localization of α-granule proteins.(a) Representative SIM imaging of a fixed resting human platelet stained for cytoskeletal alpha tubulin (magenta), granule cargo proteins thrombospondin-1 (TSP1; blue) and Von Willebrand Factor (VWF; green), and the membrane protein P-selectin (SELP; red).Granule proteins are localized within the confines of the tubulin ring and show visible overlap with each other and SELP (merge).3D maximum intensity renders; bar = 300 nm (see Supplemental Figure S1 and Video 1).(b) Representative platelet shows α-granular staining pattern VWF (red), TSP1 (blue) and serglycin (SRGN; green); (c) overlap channels and values for Pearson's correlation (PCC) and Manders' coefficients (MCA, MCB) confirm strong overlap of SRGN and TSP1.Bar = 400 nm.(d) PCC values for multiple individual platelet images (n = 10-41; bars indicate means and 95% confidence intervals) indicate strong overlap of SRGN and TSP1 comparable to that obtained for double-staining of SELP with goat and mouse (G&M) antibodies (see Supplemental Figure S2), plus similar overlap of SRGN, TSP1 and VWF with SELP (see Supplemental Figure S3a for MCC values).

Figure 2 .
Figure 2. Wheat germ agglutinin (WGA) stains α-granules.(a) Representative platelet stained with WGA (red) and immunostained for VWF (magenta), SELP (green) and surface membrane ITGB3/CD61, showing overlap channel for WGA with VWF.Under these conditions WGA staining of cell outer membranes is weak relative to α-granules; bar = 400 nm.(b) PCC values for individual platelets (n = 11-53) show strongest association of WGA with VWF and SRGN, weakest with TSP1; association with SELP is similar to that of SELP with VWF.(c) Pairwise MCC data indicates that WGA overlap with SRGN and SELP is symmetrical, and asymmetrical with VWF and TSP1.

Figure 3 .
Figure 3. RAB27B as a marker of dense granules.(a) Representative normal donor platelet stained for RAB27B (red), CD63 (green) and WGA (magenta) shows tendency for RAB27B and CD63 to stain the same irregular puncta with differing intensities.(b) PCC values indicate strong overlap of RAB27B and CD63 with each other relative to WGA.(c) Representative platelets from a normal male volunteer (NMV) and a dense granule deficient Hermansky-Pudlak syndrome (HPS) patient show a striking difference in RAB27B-positive puncta.(d) Visual scoring of puncta gave counts and distributions consistent with their identification as dense granules (the NMV platelet shown in C was scored 8, the HPS platelet 1).Bars = 300 nm.

Figure 4 .
Figure 4. Counting dense and α-granules in platelets stained for RAB27B and with WGA.(a) Image of a representative platelet showing RAB27B associated with dense granules and WGA associated with α-granules; maximum intensity render (left) and Imaris surface objects (right).(b) Pooled counts for surface objects in platelets from 2 normal male donors, E and C (see Supplemental Figure S12 for replicate data) indicate this method supports rapid counting of both types of platelet granules.Mean (standard deviation) values from left to right: 9.6 (3.29); 7.8 (2.73); 33.2 (10.9); 26.2 (8.86).Numbers of cells scored were 68 for donor E and 57 for donor C.