Factor XIII topology: organization of B subunits and changes with activation studied with single-molecule atomic force microscopy

Summary.

Background:

Factor XIII (FXIII) is a precursor of the blood plasma transglutaminase (FXIIIa) that is generated by thrombin and Ca²⁺ and covalently cross-links fibrin to strengthen blood clots. Inactive plasma FXIII is a heterotetramer with two catalytic A subunits and two non-catalytic B subunits. Inactive A subunits have been characterized crystallographically, whereas the atomic structure of the entire FXIII and B subunits is unknown and the oligomerization state of activated A subunits remains controversial.

Objectives:

Our goal was to characterize the (sub)molecular structure of inactive FXIII and changes upon activation.

Methods:

Plasma FXIII, non-activated or activated with thrombin and Ca²⁺, was studied by single-molecule atomic force microscopy. Additionally, recombinant separate A and B subunits were visualized and compared with their conformations and dimensions in FXIII and FXIIIa.

Results and Conclusions:

We showed that heterotetrameric FXIII forms a globule composed of two catalytic A subunits with two flexible strands comprising individual non-catalytic B subunits that protrude on one side of the globule. Each strand corresponds to seven to eight out of 10 tandem repeats building each B subunit, called sushi domains. The remainder were not seen, presumably because they were tightly bound to the globular A₂ dimer. Some FXIII molecules had one or no visible strands, suggesting dissociation of the B subunits from the globular core. After activation of FXIII with thrombin and Ca²⁺, B subunits dissociated and formed B₂ homodimers, whereas the activated globular A subunits dissociated into monomers. These results characterize the molecular organization of FXIII and changes with activation.

Introduction

Blood coagulation factor XIII (FXIII) is the precursor of transglutaminase (FXIIIa) that covalently crosslinks fibrin and other plasma proteins. FXIIIa catalyzes formation of an isopeptide bond between the γ-carboxy-amide group of a glutamine residue and the ε-amino group of a lysine residue in adjacent molecules. Fibrin crosslinking by plasma FXIIIa increases stiffness of individual fibrin fibers [1–3] and the whole clot fibrin network [4–8], changes the extensibility of fibrin fibers [1,2], compacts protofibrils within a fibrin fiber [9] and changes the overall fibrin clot structure [6]; it also prevents red blood cell fallout from clots [10,11] and affects platelet-driven clot contraction [12–17]. FXIIIa-catalyzed incorporation of fibrinolysis inhibitors into fibrin increases its proteolytic stability [18–21]. In summary, FXIIIa-catalyzed crosslinking is crucial for the lytic and mechanical stability of blood clots and congenital FXIII deficiency is associated with bleeding disorders [22].

Structurally, inactive plasma FXIII is a 326-kDa heterotetramer (A₂B₂) with two 83-kDa catalytic A subunits (FXIII-A) and two non-catalytic 80-kDa B subunits (FXIII-B) [23–25]. Besides plasma FXIII (pFXIII), there is a cellular form of FXIII (cFXIII) composed of two A subunits without B subunits (FXIII-A₂) [23–25]. cFXIII is present in platelets, megakaryocytes, monocytes and macrophages and has been shown to play a role in hemostasis, wound healing, angiogenesis, maintaining pregnancy, etc. [23,26,27].

In (patho)physiological conditions, the inactive pFXIII is converted into a catalytically active form (pFXIIIa) by thrombin and Ca²⁺ in two steps [28]. First, thrombin cleaves off a 37-residue-long activation peptide (AP) from the N-terminus of each A subunit, rendering them to the potentially active intermediate form (A′):

$${\text{A}}_2{\text{B}}_2\stackrel{\text{Thr}}{\to }{\text{A}}_2^{\prime }{\text{B}}_2+2\text{AP}$$ (1)

Next, in the presence of Ca²⁺ the inhibitory B subunits dissociate from the intermediate inactive complex A′₂B₂, resulting in formation of the active pXIIIa-A*₂:

$${\text{A}}_2^{\prime }{\text{B}}_2\stackrel{{\text{Ca}}^{2+}}{\to }{\text{A}}_2^{*}+{\text{B}}_2$$ (2)

Equations (1) and (2) could be summarized as follows:

$${\text{A}}_2{\text{B}}_2\stackrel{\text{Thr},{\text{Ca}}^{2+}}{\to }{\text{A}}_2^{*}+{\text{B}}_2+2\text{AP}$$ (3)

Equation (3) reflects a commonly accepted notion that the active form of pFXIIIa is a dimer of the A* subunits (A*₂) [29–31]. However, in contrast to the existing paradigm, a recent study has suggested that during activation of FXIII-A₂ the complex dissociates into monomers (A*) [32]. Despite being at variance with biochemical studies [29,31], this idea is in line with the monomeric crystal structure of activated cFXIIIa [33] and with molecular dynamics simulations, suggesting weakening of the A : A inter-subunit interface after activation of pFXIII [34]. Thus, whether the active FXIIIa is dimeric (FXIII-A*₂) or monomeric (FXIII-A*) has been a matter of controversy.

Cellular FXIII (cFXIII-A₂) can be activated with thrombin and Ca²⁺ as pFXIII-A₂B₂, but it can also be slowly activated by Ca²⁺ without proteolysis [35]. Crystallographically, the inactive cFXIII comprises an A₂ dimer with a cylinder-like structure [36–38], whereas the non-proteolytically activated cFXIIIa in complex with a covalently bound inhibitor existed as an A monomer undergoing conformational rearrangement upon activation to expose the catalytic center [33].

Unlike isolated A subunits, the atomic structures of the heteromeric pFXIII-A₂B₂ complex and isolated B subunits remain unknown. The B subunit of pFXIII is a glycoprotein consisting of 10 tandem repeats, called “sushi domains”, each containing about 60 amino acid residues held together by disulfide bonds. Low-resolution structural data on FXIII-B and full pFXIII-A₂B₂ were obtained using transmission electron microscopy [39]. Isolated B subunits appeared as thin, flexible ~30-nm-long and 2–3-nm-thick strands. Based on sedimentation analysis, these B subunits were suggested to exist as monomers; however, this assumption contradicts other evidence [29,40]. Electron microscopy of glutaraldehyde-crosslinked pFXIII-A₂B₂ (but not uncrosslinked pFXIII-A₂B₂) revealed asymmetric globular particles, in which individual subunits were not discernible. Based on these data, B subunits of pFXIII-A₂B₂ are thought to be tightly wrapped around the globular core formed by the A₂ dimer. Unfortunately, the biophysical studies of reversibly denatured pFXIII did not reveal details of the subunit assembly in pFXIII-A₂B₂ [41–43]. Therefore, the structure of individual pFXIII-B subunits and their topology in the pFXIII-A₂B₂ tetramer remain largely unknown.

Here, we visualized and characterized quantitatively the morphology of the entire pFXIII-A₂B₂ molecules and individual A and B subunits before and after proteolytic activation using high-resolution single-molecule atomic force microscopy (AFM) [44,45]. We showed that pFXIII consisted of a globular A₂ dimer and thin flexible strands of B subunits that typically protrude from the globular core. In activated pFXIIIa, the B subunits dissociated from the A₂ globules and existed as B₂ dimers, whereas the dimeric A₂ globules fell apart into catalytically active A* monomers.

Materials and methods

Preparations of plasma-purified (pFXIII) and recombinant (rFXIII) human factor XIII

pFXIII and pFXIIIa were purchased from Enzyme Research Laboratories (South Bend, IN, USA). The inactive pFXIII was > 94% pure in SDS-PAGE (Fig. 1A) with a potential 2950 Loewy U mg⁻¹ specific transglutaminase activity (1 Loewy U mL⁻¹ is defined as the highest dilution of the enzyme capable of forming an insoluble clot under the conditions described in [46]). Preactivated pFXIIIa was > 94% pure (Fig. 1A) and had a specific activity of 2778 Loewy U mg⁻¹. In addition, we used: (i) inactive rFXIII-A subunits from ZymoGenetics (Seattle, WA, USA), > 95% pure (Fig. 1A) with a potential 3049 Loewy U mg⁻¹ specific activity; (ii) inactive rFXIII-A subunits from Zedira (Darmstadt, Germany), > 95% pure (Fig. 1A) with a potential 3343 Loewy U mg⁻¹ specific activity (the latter preparation contained maltodextrin as a stabilizer, which formed a non-uniform background in AFM images; see Fig. S2); and (iii) inactive rFXIII-B subunits from Zedira, > 95% pure (Fig. 1A).

Fig. 1.

Polypeptide chain composition of plasma-purified (pFXIII) and recombinant (rFXIII) preparations of factor XIII. (A) SDS-PAGE in reducing conditions of the FXIII preparations used in this study. Lanes from left to right: molecular mass markers, pFXIII from Enzyme Research Laboratories containing A and B subunits, rFXIII-A subunits from Zedira, rFXIII-A subunits from ZymoGenetics, rFXIII-B subunits from Zedira, and preactivated pFXIIIa from Enzyme Research Laboratories. The insert shows an SDS-PAGE of pFXIII in non-reducing conditions. (B) Activation of pFXIII with thrombin and Ca²⁺ for 10 min, 30 min and 2 h. (C) Activation of rFXIII-A₂ (ZymoGenetics) with thrombin and Ca²⁺ for 10 min, 30 min and 3 h. Letters A, A*, B and rB at the bottom of the lanes indicate the subunit composition of each sample. All figures show entire gels.

Subunit composition of the FXIII preparations was evaluated by 7.5% SDS-PAGE (Fig. 1A) in Laemmli buffer [47]. In reducing SDS-PAGE, the A and B subunits were overlapping, as has been shown before [24], and separated only after partial cleavage of the A subunits with thrombin because activated FXIII-A* became 4-kDa smaller and appeared as a separate band (Fig. 1B); A* and B subunits were initially separated in the commercially available preactivated pFXIIIa (Fig. 1A). In non-reducing conditions, the A and B subunits of inactive pFXIII were revealed as clearly separated bands (Fig. 1A, insert); some degree of heterogeneity of FXIII-A subunits seen in non-reducing conditions is probably a result of partial cleavage or glycosylation [48]. Composition of the two preparations of rFXIII-A subunits (from ZymoGenetics and Zedira) were similar to each other and identical to the A subunits of pFXIII. rFXIII-B subunits had smaller molecular weight than their plasma-purified counterparts because of a different glycosylation pattern (Fig. 1A).

Activation of FXIII with thrombin and Ca²⁺

pFXIII or rFXIII-A₂ were diluted to a 0.267 mg mL⁻¹ concentration with 20 mm HEPES buffer, pH 7.4, containing 150 mm NaCl and 5 mm CaCl₂. Human thrombin (T4393, Sigma-Aldrich, St. Louis, MO, USA) was added at a final concentration of 0.4 NIH U mL⁻¹ and the mixture was incubated for 10 min, 30 min, 2 h or 3 h at 37 °C. In 2 h pFXIII was fully activated by the partial cleavage, as determined by the conversion of the A subunits to the smaller A* subunits (Fig. 1B). Cleavage of rFXIII-A subunits from ZymoGenetics was slower and was almost complete in 3 h (Fig. 1C). The same results were obtained with rFXIII-A subunits from Zedira (Fig. S1). At 30 min after addition of thrombin, about half of the A subunits of both pFXIII and rFXIII-A₂ were partially cleaved to the activated state (Fig. 1B–C). We used 5 mm CaCl₂, which can potentially induce very slow non-proteolytic activation of FXIII that would not exceed 1% at 30 min [49], so our AFM images should not detect non-proteolytic activation of FXIII.

The BioVision Colorimetric FXIIIa Activity Assay Kit (BioVision, Milpitas, CA, USA) was used to determine specific transglutaminase activity of activated FXIII preparations normalized by the protein concentration determined at λ = 280 nm, assuming ${\text{A}}_{\text{lcm}}^{0.1\%}=1.49$ for rFXIII-A and ${\text{A}}_{1\text{cm}}^{0.1\%}=1.38$ for pFXIII [32,50]. The activity assay utilizes the transglutaminase activity of FXIIIa to crosslink an amine-containing substrate to a glutamine-containing substrate, resulting in the release of the ammonium cation, which is quantified with a detection reagent that has a decreased absorbance at 340 nm upon reaction with NH4⁺.

Sample preparation for AFM

Preparations of FXIII and its derivatives were diluted to 2 μg mL⁻¹ with 20 mm HEPES buffer, pH 7.4, containing 150 mm NaCl and 5 mm CaCl₂. Typically 2 μL of the diluted protein solution was applied on a substrate and kept for 5–15 s. Then 200 μl of fresh milli-Q water was carefully placed over the sample, kept for 10 s and removed with a flow of air to dry the surface. All the protein samples were adsorbed on the highly oriented pyrolytic graphite coated with an amphiphilic graphite modifier (GM-graphite), used earlier for high-resolution single-molecule AFM imaging of proteins and nucleic acids [44,45,51].

Acquisition and processing of AFM images

AFM imaging was performed using a MFP-3D microscope (Asylum Research, Goleta, CA, USA) in a tapping mode with a typical scan rate of 0.5 Hz. Images were taken in air using sharpened silicon cantilevers, SSS-SEIHR (Nanosensors, Neuchâtel, Switzerland), with guaranteed tip radius < 5 nm or standard cantilevers, OMCL-AC200TS (Olympus, Tokyo, Japan), with a typical tip radius of 7 nm. FemtoScan Online software (http://www.femtoscanonline.com) was used to filter, analyze and present the AFM images. SPM Image Magic software (https://sites.google.com/site/spmimagemagic) was used for a semi-automatic measurement of the height of visualized objects.

Statistical analysis

Statistical analysis was performed in R (https://www.r-project.org) using the package Psych (https://cran.r-project.org/web/packages/psych). The results are presented as mean ± SD.

Results

Molecular structure of inactive FXIII and its subunits

Structure of tetrameric pFXIII-A₂B₂

Individual molecules of pFXIII adsorbed on a GM-graphite surface were visualized in AFM as a globule with two, one or no thread-like extensions (Fig. 2A,B). Two strands (typically, both on one side of the globule) were seen in 67% of the molecules, one in 14%, and no strands were seen in 19% of the pFXIII-A₂B₂ molecules (n = 418). The size (height) of the globular portions had a symmetrical distribution with a maximum at 3.5 ± 0.5 nm (n = 2511, Fig. 2C). The shape of the globular portion of pFXIII was characterized quantitatively using circularity (i.e. the ratio of the projected area of an object to an area of a circle with the same perimeter; the circularity of a circle is equal to one, whereas for elongated structures it is smaller). We measured the perimeter and area of projections of 40 randomly selected globular cores of pFXIII molecules with two visible extensions. The calculated average circularity was equal to 0.86 ± 0.05, suggesting moderate ovality of these structures. The dimensions of the strands were characterized by histograms with a peak contour length of 20 ± 6 nm (n = 186; Fig. 2D) and a height of 0.4 ± 0.1 nm (n = 162; Fig. 2E). Based on the known polypeptide composition of pFXIII-A₂B₂, the observed molecular images could be interpreted as an A₂ globule with one or two individual thread-like B subunits.

Fig. 2.

Atomic force microscopy (AFM) images and morphometric characterization of plasma-purified factor XIII (pFXIII) molecules. (A) A representative wide-field AFM image of pFXIII. (B) Individual pFXIII molecules showing characteristic structures with a globular portion and two strands. (C) A distribution of heights of the globular portions of the pFXIII molecules shown in A and B. (D) A distribution of contour lengths of the protrusions of strands of the pFXIII molecules shown in B. (E) A distribution of heights of the protrusions of strands of the pFXIII molecules shown in B.

The absolute values of heights of the A₂ globules and thread-like B subunits measured with AFM are smaller than the dimensions of the corresponding atomic structures of A₂ globules [36–38] and homologous sushi domains [52] determined using X-ray crystallography. This discrepancy of dimensions observed in the AFM images and full-atom crystal structures is likely to be a result of variable orientation possibilities of the FXIII-A₂ complex, as well as potential surface effects of GM-graphite and sample drying that affect the overall behavior of the molecule in AFM and may affect the measurements.

Structure of rFXIII-A₂

The rFXIII composed of two A subunits (ZymoGenetics) was visualized as globules with an average height of 3.7 ± 0.3 nm (n = 634; Fig. 3). The height value was close to the height of the globular portion of pFXIII-A₂B₂ (Table S1), so we concluded that in rFXIII each globule corresponded to an A₂ dimer, which is in agreement with the literature on the prevalent oligomeric state of free FXIII-A subunits [24,25,36]. A similar height value (3.3 ± 0.4 nm, n = 338) was obtained for the globular rFXIII-A₂ from Zedira, despite the presence of maltodextrin on the substrate that somewhat complicated the quantitative image analysis, because of the uneven background (Fig. S1, Table S1). An average circularity of rFXIII-A₂ molecules was equal to 0.91 ± 0.05 (n = 40). This value was significantly larger than the circularity of the globular cores of pFXIII molecules with two visible extensions (P < 0.005, Wilcoxon rank sum test). Thus, the shape of the globular portion of tetrameric pFXIII-A₂B₂ was slightly elongated compared to rFXIII-A₂.

Fig. 3.

Atomic force microscopy (AFM) images and morphometric characterization of recombinant factor XIII -A₂ molecules (rFXIII-A₂). (A) A representative wide-field AFM image of rFXIII-A₂. (B) Individual rFXIII-A₂ molecules seen as globules. (C) A distribution of heights of the rFXIII-A₂ globules shown in A and B.

Structure of rFXIII-B₂

rFXIII-B subunits were visualized as strands with an average contour length of 33 ± 3 nm (n = 175; Fig. 4A–C) and a height (thickness) of 0.6 ± 0.1 nm (n = 175; Fig. 4D). Importantly, not only was the contour length increased, but also the height of the strands was significantly greater compared to the thickness of the flexible strands observed in pFXIII-A₂B₂ (P < 0.001, Wilcoxon rank sum test; Table S1). The increased length of rFXIII-B subunits could be potentially attributed to either elongation resulting from dissociation of the monomeric B subunits from the A₂ dimer or to their dimerization. An additional argument for dimerization is an increase of thickness of B subunits separated from the A₂ globule (Table S1), suggesting that the separated rFXIII-B subunits exist in a dimeric form [40,53], whereas in the whole pFXIII-A₂B₂ tetrameric molecules the B subunits protrude from the globular portion of the protein as two disjointed single polypeptides.

Fig. 4.

Atomic force microscopy (AFM) images and morphometric characterization of recombinant factor XIII-B₂ molecules (rFXIII-B₂). (A) A representative wide-field AFM image of rFXIII-B₂. (B) Individual rFXIII-B₂ molecules visualized as flexible strands. (C) A distribution of contour lengths of the rFXIII-B₂ strands shown in A and B. (D) Distribution of heights of the rFXIII-B₂ strands shown in A and B.

Structural changes associated with activation of FXIII

To follow structural perturbations of FXIII during its activation, we compared the polypeptide chain compositions and AFM single-molecule images of pFXIII-A₂B₂ and rFXIII-A₂ before and after treatment with thrombin and Ca²⁺. In addition, we imaged commercially available preactivated pFXIIIa.

Activation of pFXIII-A₂B₂

AFM samples of pFXIII prepared at a 30-min time-point of incubation with thrombin revealed that most of the B subunits dissociated from the globular portion of pFXIII (Fig. 5A,B) and appeared as free flexible strands with an average 33 ± 4 nm contour length (n = 148; Fig. 5D) and 0.6 ± 0.1 nm height (n = 148; Fig. 5E). These dimensions were significantly distinct from the corresponding parameters of the B subunit flexible strands seen in inactive pFXIII molecules that had a visible contour length of 20 ± 6 nm and a height of 0.4 ± 0.1 nm (P < 0.001 for both parameters, Wilcoxon rank sum test). These differences suggest that upon activation of pFXIII with thrombin in the presence of Ca²⁺, B subunits dissociate from the A₂ dimer and self-associate to form B₂ dimers. We also measured the heights of globular particles remaining separated after 30-min activation of pFXIII (Fig. 5A–C). The height distribution (n = 327) was bimodal with two peaks: one at3.3 ± 0.6 nm corresponded to the whole tetrameric pFXIII-A₂B₂ and, potentially, dimeric FXIII-A₂, whereas the other peak at 2.1 ± 0.3 nm probably represented FXIII-A and/or FXIII-A* monomers. As a matter of fact, the number of globular particles in the AFM images reduced with increasing incubation time of pFXIII-A₂B₂ with thrombin: at 2 h after addition of thrombin, only a few globular particles were present. We attribute this effect to aggregation of the activated pFXIII previously observed by other researchers [29,32,54].

Fig. 5.

Atomic force microscopy (AFM) images and morphometric characterization of plasma-purified factor XIII (pFXIII) activated with thrombin and CaCl₂. (A) A representative image showing separation of globules and strands. Arrowheads point to activated (likely monomeric) A* and/or A subunits; an arrow points to an intact non-activated pFXIII molecule. (B) Individual pFXIII-A* (and/or pFXIII-A) subunits visualized as globules and pFXIII-B₂ dimers seen as strands. (C) A distribution of heights of the globules after activation of pFXIII shown in A and B. (D) Distribution of contour lengths of the strands formed after activation of pFXIII shown in A and B. (E) Distribution of heights of the strands formed after activation of pFXIII shown in A and B.

AFM of commercially available preactivated pFXIIIa also revealed the presence of monomeric A* subunits visualized as globules with an average height of2.5 ± 0.5 nm (n = 413; Fig. 6C), substantially smaller than the dimeric A₂ globules in pFXIII (Fig. 2C), as well as separated FXIII-B₂ dimers visualized as strands with a contour length of 36 ± 5 nm (n = 194; Fig. 6A–B,D) and a height of 0.6 ± 0.1 nm (n = 194; Fig. 6E). Intact pFXIII molecules were virtually missing in the preparation of preactivated pFXIIIa.

Fig. 6.

Atomic force microscopy (AFM) images and morphometric characterization of pre-activated plasma-purified factor XIII (pFXIIIa). (A) A representative image showing separate globules and strands. (B) Individual pFXIII-A* and/or A subunits visualized as globules and pFXIII-B₂ dimers visualized as strands. (C) A distribution of heights of the globules shown in A and B. (D) A distribution of contour lengths of the strands shown in A and B. (E) A distribution of heights of the strands shown in A and B.

Activation of rFXIII-A₂

According to SDS-PAGE, at 30 min after treatment with thrombin and Ca²⁺, about half of rFXIII-A₂ molecules were partially cleaved (Fig. 1C). In the AFM images of rFXIII-A₂ from ZymoGenetics activated for 30 min, we saw two types of globular particles: most of the particles had an average height of 2.0 ± 0.4 nm, corresponding to the monomeric A and/or A* subunits, whereas the minority had an average height of 3.4 ± 0.2 nm, corresponding to the intact dimeric rFXIII-A₂ (n = 439; Fig. 7). Similar results were obtained with activated rFXIII-A₂ from Zedira, namely the AFM images revealed small A and/or A* monomeric globules with an average height of 2.3 ± 0.4 nm (n = 540; Fig. S3).

Fig. 7.

Atomic force microscopy (AFM) images and morphometric characterization of recombinant factor XIII-A₂ molecules (rFXIII-A₂) activated with thrombin and CaCl₂. (A) A representative image showing individual multiple small globules comprising monomeric A* and/or A subunits. (B) Individual monomeric A* and/or A subunits. (C) A distribution of heights of the globules shown in A and B.

Discussion

The crystal structures of FXIII-related molecules are limited to active monomeric and inactive dimeric A subunits without flexible B subunits [33,36–38]. Transmission electron microscopy of glutaraldehyde-crosslinked pFXIII-A₂B₂ did not allow for identification of individual subunits and did not reveal their spatial arrangement [39]. So, this work is aimed at providing structural information on various forms of individual molecules of FXIII and its subunits with an unprecedented resolution using single-molecule AFM. Our findings provide new information about the structural arrangement of inactive pFXIIIA₂B₂, as well as the changes in subunit composition following activation.

Tetrameric, dimeric and monomeric subunits of FXIII

Tetrameric pFXIII-A₂B₂

The inactive pFXIII exists in three structural variants, namely a globule with two, one or no thin flexible extensions. This structural diversity may reflect the equilibrium between association and dissociation of the B subunits from the A₂ dimer. Alternatively, the B subunits may be partially or fully wrapped around the globular A₂ dimer, making them invisible in AFM images. The first explanation is supported by the following calculation. Based on K_d = 4.17 × 9 10⁻¹⁰ M for A-B interactions [55] and FXIII concentration ~6.3 nM, in equilibrium 24% of A and B subunits dissociate, whereas 76% exist as a complex. According to our AFM data, 26% of the B subunits dissociate from the FXIII globular core, whereas 74% are connected to the globules, in excellent agreement with the calculated values.

Remarkably, two spread-out B subunits were typically both seen on one side of the globule, perhaps because FXIII-A₂ is asymmetrical and the B subunits might attach to asymmetrically positioned binding sites. Each of these binding sites is presumably located in the vicinity of the activation peptide [34] and includes critical amino acid residues Gly562 and Tyr283 [56]. This natural structural asymmetry might be even more pronounced in AFM because of a non-random preferential orientation of FXIII molecules during adsorption on the substrate.

rFXIII-A₂ homodimer

Although a crystallographic structure of FXIII-A₂ is available, its AFM imaging is important for identifying the A₂ dimer within inactive and activated pFXIII molecules. By AFM, rFXIII-A₂ is visualized as a globule with a size close to that of the pFXIII globular core (Fig. 3C, Fig. S2C), implying that the molecular dimensions of the globular portion of pFXIII are determined by the A₂ dimer and a contribution of B subunits is negligibly small.

Free B₂ homodimers and monomeric B subunits within pFXIII

Recombinant dimeric B subunits (rFXIII-B₂) appear as thin flexible strands with an average contour length of 33 nm, which is in good agreement with a 30-nm estimation obtained by transmission electron microscopy [39]. Because both the length and thickness (height) of the free B₂ dimers were significantly larger compared to the flexible strands protruding from the globular core of pFXIII-A₂B₂ molecules (Table S1), we conclude that these strands in pFXIII-A₂B₂ comprise monomeric B subunits. This deduction agrees with earlier studies showing that modified B subunits that are unable to dimerize with each other can still participate in formation of the A₂B₂ heterotetramer [40,53] if they possess intact sushi domains 1 and 2 [40,55].

Based on this reasoning, we can tentatively estimate how many sushi domains are seen in the single B subunits extending from the globular part of pFXIII and having a contour length of 20 nm in the AFM images (Fig. S5A). If the contour length of B₂ dimers (about 33 nm) corresponds to a previously suggested structural model, in which two B subunits have an antiparallel orientation as a result of inter-subunit interactions between sushi domains 4 and 9 [40], then the length of a B monomer should be about 27.5 nm (Fig. S5B). If that is true, then the 20-nm strands extending from the globular part of pFXIII molecules correspond to seven or eight freely exposed sushi domains of the B subunit (Fig. S5A) and only sushi domains 1 and 2, which are critical for pFXIII heterotetramer formation, comprise the shorter portions of B subunits that are tightly bound to the globule formed by two A subunits and remain invisible [40,55].

FXIII-A monomers versus dimers

Our measurements strongly suggest that the oligomeric pFXIIIa-A₂B₂ and rFXIII-A₂ disassemble upon activation and form monomeric active A* subunits. In all four species of activated FXIII studied, the monomeric A or A* subunits were visualized as compact globules that were significantly smaller than A₂ dimers (~2.1 nm for a monomer vs. ~3.5 nm for a dimer, Table S1). To exclude the possibility that this difference in height was solely a result of spatial rearrangement of the A₂ dimers, we also calculated and compared the molecular volumes of these two types of globules. The 2.3-fold difference (185 ± 23 nm³, n = 20, for a monomer, and 428 ± 58 nm³, n = 20, for a dimer) strongly confirms that the observed variations in the height of A-subunit-containing globules reflect their monomeric vs. dimeric structure.

Structural mechanism of proteolytic activation of pFXIII

It has been generally accepted that activated pFXIIIa or cFXIIIa contain dimeric A* subunits (A*₂) [29,31]. However, a few recent studies have suggested that the active A* subunits are monomeric and that their dissociation from dimers to monomers is an inherent part of FXIII activation [32–34]. Similarly, the existing data on the oligomeric vs. monomeric state of free B subunits are also controversial [23].

Our work provides direct evidence for the formation of monomeric FXIII-A* subunits and dimeric FXIII-B₂ subunits after activation (proteolytic cleavage of the activation peptide, AP) of pFXIII-A₂B₂ with thrombin and Ca²⁺. This paradigm-changing concept is reflected by the equation:

$${\text{A}}_2{\text{B}}_2\stackrel{\text{Thr},{\text{Ca}}^{2+}}{\to }2{\text{A}}^{*}+{\text{B}}_2+2\text{AP}$$ (4)

It is conceivable that the conversion of A2B2 tetramer to A* monomers and B2 dimer can go through the intermediate heterodimeric complexes AA?, AB or A?B, although our experimental data do not provide evidence for that. In the samples prepared for AFM imaging 30 min after addition of thrombin and Ca2+ to pFXIII-A2B2 or rFXIII-A2, only about 1/2 of the A subunits were cleaved as judged from the SDS-PAGE (Fig. 1B–C). However, on the AFM images at this time-point of activation the vast majority of the A subunits were monomeric with very few A2 dimers, suggesting that cleavage of only one AP from the A2 dimer makes the two A subunits dissociate because the AA? (hetero)dimers are thermodynamically unstable in the presence of Ca2+. In the case of pFXIII-A2B2, dissociation of B subunits precedes separation of the AA? complex into monomeric A* and A subunits.

Importantly, in the circulation, binding of pFXIII to fibrinogen is mediated predominantly by the interaction between the B subunit of FXIII and the C-terminus of fibrinogen’s gamma-prime-chain [57]. Therefore, there is no conceivable spatial obstacle that would prevent dissociation of dimeric A* subunits into monomers upon activation of pFXIII that we observed in purified pFXIIIa.

Conclusions

In summary, our results provide single-molecule images that reflect conformational and orientational possibilities of the individual subunits of inactive and activated human plasma FXIII. Although the results of AFM depend on surface-related conditions that are different from solution and molecular dimensions are smaller from drying, they can capture fundamental structural features underlying the function of plasma-derived and cellular transglutaminases. Based on the AFM images, the inactive plasma FXIII (pFXIII-A₂B₂) consists of a globular A₂ dimeric core and two individual B subunits extending out from one side as flexible strands. The size of the globular part was found to be 3.5 nm and the extending filamentous strands were about 20 nm long. Recombinant inactive FXIII-A subunits in the absence of FXIII-B subunits were shown to form a globular homodimer about3.5 nm in size, recapitulating the structure of cellular FXIII. Recombinant FXIII-B subunits in the absence of FXIII-A subunits formed flexible homodimeric strands that were 0.6 nm thick and 33 nm long. The surface molecular topography revealed by AFM suggests that activation of pFXIII-A₂B₂ with thrombin and Ca²⁺ includes at least three concerted structural rearrangements: (i) dissociation of the initially monomeric B subunits from the globular core of the molecule; (ii) association of the released B subunits to form B₂ homodimers; and (iii) dissociation of the initially dimeric A subunits into enzymatically active monomers. Accordingly, activation of the B-subunit-free dimeric rFXIII-A₂ with thrombin and Ca²⁺ includes only one step, namely dissociation into monomeric enzymatically active A subunits.

Supplementary Material

supplement

Fig. S1. Polypeptide chain composition of rFXIII-A₂ (Zedira) activated with thrombin and Ca²⁺ for 10 min, 30 min and 3 h. Letters A and A* at the bottom of the lanes indicate the subunit composition of each sample.

Fig. S2. AFM images and morphometric characterization of rFXIII-A₂ molecules (Zedira). (A) A representative wide-field AFM image of rFXIII-A₂. (B) Individual rFXIII-A₂ molecules seen as globules. (C) Distribution of heights of the rFXIII-A₂ globules shown in A and B. The background is covered by maltodextrin added to the commercial preparation as a stabilizer.

Fig. S3. AFM images and morphometric characterization of rFXIII-A₂ (Zedira) activated with thrombin and CaCl₂. (A) A representative image showing individual small globules comprising monomeric A* and/or A subunits. (B) Individual rFXIII-A* and/or A subunits. (C) Distribution of heights of the rFXIII-A* globules shown in A and B. The background is covered by maltodextrin added to the commercial preparation as a stabilizer.

Fig. S4. Comparative vertically paired distributions of heights of the globular portions of FXIII preparations before (top panels) and after (bottom panels) activation with thrombin and CaCl₂ for 30 min. (A) Inactive pFXIII, N = 2511. (B) inactive rFXIII-A₂ (ZymoGenetics), N = 634. (C) Inactive rFXIII-A₂ (Zedira), N = 338. (D) Activated pFXIIIa, N = 327. (E) Activated rFXIII-A* (ZymoGenetics), N = 439. (F) Activated rFXIII-A₂ (Zedira), N = 540.

Fig. S5. Characteristic AFM images correlated with the proposed models of B subunit topology. (A) An AFM image of a heterotetrameric pFXIII-A₂B₂ molecule in which monomeric B subunits extend from the A₂ globular dimeric core (on the left). The corresponding cartoon (on the right) shows sushi domains 1 and 2 tightly attached to the globular core, while domains 3–10 are freely extending outside. (B) A B₂ homodimer after separation from pFXIII-A₂B₂ upon activation (on the left). The corresponding cartoon shows two B subunits with antiparallel orientation due to inter-subunit interactions between sushi domains 4 and 9.

Table S1. Morphometric parameters of globules and filaments seen with AFM in FXIII preparations before and after activation.

Essentials.

• Factor XIII is a heterotetramer with 2 catalytic A subunits and 2 non-catalytic B subunits.

• Structure of active and inactive factor XIII was studied with atomic force microscopy.

• Inactive factor XIII is made of an A₂ globule and 2 flexible B subunits extending from it.

• Activated factor XIII separates into a B₂ homodimer and 2 monomeric active A subunits.