Cryo-EM tomography reveals insertion of unclustered integrins into liposomes. (A) Central section of the cryo-EM tomogram of frozen hydrated liposomes with integrin αIIbβ3 and THD co-incorporated. (B) Central section of the cryo-EM tomogram of liposomes containing only integrin αIIbβ3. The individual integrins were clearly visible in both A and B with a head density connected by a narrow leg density to the lipid densities, indicating that the integrins were incorporated into the lipid bilayers. Insets show magnified views of a few regions in both A and B and indicate that the integrins are well resolved. No gross clustering was observed in the presence or absence of THD. The red arrows point to the individual integrin densities.

THD induces the extended conformation in the absence of ligand or externally applied force. All panels are images of 50 × 50-nm fields. The number of images in each class is shown at the bottom left corner of each box. (A) Class averages of integrin nanodiscs in compact bent-over conformation. The class averages show a round-shaped nanodisc density and compact bent-over density for the integrins. All 18 classes are shown in Fig. S3. (B) Class averages of integrin nanodiscs in compact conformation in the presence of THD. The class averages show a round-shaped nanodisc density and compact density for the integrins. Although the integrin densities are compact, they appear slightly less compact than when THD is absent. The numbers of molecular images in each class are shown at the bottom left corner of each box. (C) Class averages of integrin nanodiscs in extended conformation in the presence of THD. All classes for the integrin nanodiscs with THD are shown in Fig. S4. (D) Total numbers of extended or compact integrin nanodiscs were obtained by summing the number of images in the classes showing extended or compact conformation. Only classes with clearly defined integrin density were included in the calculation of the percentages.

Compact integrin with and without THD roughly segregate into separate classes. Panels are images of 50 × 50-nm fields. Images of nanodiscs containing compact integrins from sample prepared in the presence or absence of THD were grouped together, and the pooled images were then classified and averaged iteratively. 20 classes were generated during the iterative process (parent classes) and at the final step, each class was split to two subclasses (daughter subclasses) to verify the internal consistency of the class averages. Each bracket contains two daughter subclasses from a parent class. Little or no difference between the daughter subclasses indicated robust alignment and classification by the iterative procedure. Within each class, the percentage of images from samples containing either integrin nanodiscs in the presence or absence of THD are depicted in a pie chart. Each resulting class average in which the integrin structure was clearly defined was greatly enriched in images from the integrin nanodiscs alone or from the THD-containing integrin nanodisc population. For example, the first row of images shows classes dominated by images from the integrin alone sample (up to 89%). In contrast, the second and third row are classes dominated by images from the sample containing added THD (up to 91%). In the classes in which the integrin structure is not clearly resolved, both samples were more evenly represented (fourth row). 8 classes of empty nanodiscs were omitted from this figure.

Embedding of integrin αIIbβ3 in phospholipid nanodiscs. (A) Model of integrin αIIbβ3 in a compact, bent conformation, inserted into a nanodisc. The membrane scaffold protein (MSP) is shown in solid surface view, lipid molecules are shown in colored wire, integrins are shown as ribbons. The proposed arrangement of the nanodiscs in solution and the likely arrangement of the integrin nanodiscs after drying and flattening in negative stain are shown. PDB coordinates for the integrin structures obtained from 3FCS (Zhu et al., 2008) (ectodomain), 2H7E (Wegener et al., 2007) (cytoplasmic domain), and 2K9J (Lau et al., 2009) (transmembrane domain). The nanodisc coordinates (Segrest et al., 1999) were obtained from Dr. S.C. Harvey’s website (http://rumour.biology.gatech.edu). (B) Negatively stained EM images of integrin nanodiscs. Several examples of integrin nanodiscs are highlighted with red arrows pointing to the nanodisc density and blue arrows pointing to the integrin densities. Bar, 100 nm. (C) Enlarged view of a representative gallery of single-integrin nanodiscs shown at higher magnification than in B. Examples of both compact bent-over and extended integrin conformations are included. Each panel in C and D is 50 × 50 nm. (D) Similar to C, except that occasional nanodiscs containing two integrins are shown.

Talin binding is sufficient to activate integrins. (A) Liposome size vs. PAC1 binding. Flow cytometric data were analyzed in 11 regions of forward scatter (FSC; a measure of particle size). THD increased PAC1 binding (▴) compared with integrin alone (■) or empty liposomes (♦). (B) Activation indices plotted against liposome size. The presence of THD (▴) increased activation relative to its absence (■). The inset depicts an SDS-PAGE of integrin liposomes in the presence and absence of THD. Data in A and B are from one of three independent experiments. (C) Activation indices of the liposomes at the mid-point of the forward scatter spectrum (mean FSC = 142) were averaged from three experiments. THD increased integrin activation. Unless otherwise indicated, error bars in the graphs shown here and below denote the standard error of three independent experiments. (D) THD activates an integrin mutant defective in kindlin binding. THD activated integrin αIIbβ3, and coexpression of kindlin-2 led to an additional increase in activation. In contrast, THD activated αIIbβ3(Y759A), a mutant defective in kindlin binding, whereas kindlin-2 did not augment the activation. αIIbβ3(Y747A), which is defective in talin binding, was not activated by THD nor by a combination of THD and kindlin 2. Activation index is calculated as in B. The bottom panel is a Western blot showing that the expression of THD and kindlin-2 are comparable in αIIbβ3, αIIbβ3(Y759A), or αIIbβ3(Y747A) cells.

THD activates integrins in nanodiscs. (A) THD increases the binding of PAC1 to captured integrin nanodiscs. The inset shows the configuration of the assay, which is described in the Materials and methods section. The activation indices were calculated as (L − L₀) / (L_max − L₀) where L = luminescence, L₀ = luminescence in presence of 1 µM eptifibatide, and L_max = luminescence in the presence of anti-LIBS6 activating anti-β3 antibody. (B) Mutations of THD reduce its capacity to activate integrins in nanodiscs. The increase in activation is calculated as in Fig. 3 B. A mutation that reduces membrane binding (K322D) or one that blocks binding to the integrin-β3 tail (L325R) reduced capacity of THD to activate αIIbβ3 in nanodiscs. (C) EM images of negatively stained fibrin and integrin nanodiscs alone. Relatively few integrin nanodiscs are bound to the fibrin. Bar, 100 nm. (D) EM images of negatively stained fibrin and integrin nanodiscs in the presence of 5 µM THD showing an increase in the number of bound integrin nanodiscs when THD is present. The top insets show enlarged views of integrin nanodiscs bound to the fibrin. Single fibrin-bound integrin nanodiscs are clearly visible, indicating that the increase in binding is due to increased monomer affinity rather than clustering. The bottom insets show enlarged views of unbound integrin nanodiscs. Both compact and extended conformations are visible. Bar, 100 nm. (E) The local concentrations of nanodiscs were estimated from EM images by manual enumeration of nanodiscs/µm². The number of bound integrin nanodiscs per 10-nm length of fibrin strand was plotted for each concentration. The presence of THD increased the number of bound integrin nanodiscs. Error bars = SE (n = 4) for each concentration.

THD activates purified αIIbβ3 by binding to the β3 tail. (A) THD induced an increase in PAC1 binding (▴), whereas THD-W359A (■), a mutant deficient in integrin tail binding, failed to do so. The inset on the right depicts SDS-PAGE showing that integrin and THD are comparable. Data are from one of three independent experiments. (B) Activation indices of the subset of liposomes at the mid-point of the forward scatter spectrum (mean FSC = 142) were averaged over three independent experiments. Increases in integrin activation (calculated as AI_{with_THD} − AI_{integrin_alone}, where AI stands for activation indices) are depicted. THD increased integrin activation compared with THD-W359A. (C) Requirement of the β3 tail for THD activation. Using the ELISA described in the Materials and methods section, calpain abolished anti–β3-c tail binding but had no effect on anti-β3 or anti–αIIb-c tail binding. SDS-PAGE revealed that calpain did not markedly shift mobility of β3 or αIIb. (D) THD fails to activate αIIbβ3 lacking the β3 C terminus. Increase in integrin activation was calculated as in B. (E) SDS-PAGE shows that integrin and THD incorporation are comparable for calpain-digested and intact integrins. (F) THD mutants do not disrupt protein folding. Differential scanning calorimetry of THD, THD-W359A, and THDF-K322D. All three proteins exhibited similar narrowly defined melting points, indicating that they were well folded.

The importance of lipid embedding for THD activation. (A) A model of talin head–integrin-β3 tail interaction (Wegener et al., 2007). β3 tail is shown in ribbon diagram and the F3 domain of talin is shown in solid surface. The two mutations, W359A (blocks integrin binding) and K322D (predicted to reduce membrane binding), are highlighted in blue and red, respectively. (B) THD-K322D (♦) caused markedly reduced PAC1 binding relative to THD (▴). Data are from one of three independent experiments. The SDS-PAGE on the right confirms that the integrin and THD incorporation are comparable for THD-K322D and THD (both lanes were excised from the same gel image and were juxtaposed for clarity). (C) Activation indices of the subset of liposomes at the mid-point of the forward scatter spectrum were averaged over three independent experiments. Increases in integrin activation were calculated as in Fig. 3 B. THD but not THD-K322D induced increased integrin activation. (D) The K322D mutation reduces F3–membrane interaction. Co-sedimentation of F3 or F3-K322D with liposomes was assessed by SDS-PAGE. Densitometry was used to quantify these results, which are depicted in the bar graph to the right. Mass spectrometry revealed that both F3-WT and F3-K322D were within 0.1% of their predicted masses; hence, the increased mobility of F3-K322D is ascribed to its increased negative charge. (E) Schematic of assays for measuring activation of detergent-solubilized αIIbβ3. (F) THD failed to activate captured detergent-solubilized integrin. αIIbβ3 was mixed with various concentrations of THD and was captured on a plate coated with AP3, and activated αIIbβ3 was detected by PAC1 binding. Captured αIIbβ3 was detected by the binding of polyclonal anti-β3 antibody (anti-β3). PAC1 binding was low and did not change even with ∼1,000-fold molar excess of THD. The panel on the right shows that Mn++ activated the αIIbβ3 in the same experiment. (G) The plate was coated with PAC1 and αIIbβ3 mixed with various concentrations of THD was then added to the plate. The bound integrins were detected with polyclonal anti-β3 antibody. At a 1,000-fold molar excess, THD failed to induce binding of PAC1 to the αIIbβ3; in contrast, 1 mM Mn++ did so, and the binding was inhibited by a competitive antagonist, eptifibatide.