Force-dependent isomerization kinetics of a highly conserved proline switch modulates the mechanosensing region of filamin Lorenz Rognonia, Tobias Mösta, Gabriel Žoldáka, and Matthias Riefa,b,1 a

Physik Department E22, Technische Universität München, 85748 Garching, Germany; and bMunich Center for Integrated Protein Science, 81377 Munich, Germany Edited by David Baker, University of Washington, Seattle, WA, and approved March 10, 2014 (received for review October 16, 2013)

Proline switches, controlled by cis–trans isomerization, have emerged as a particularly effective regulatory mechanism in a wide range of biological processes. In this study, we use single-molecule mechanical measurements to develop a full kinetic and energetic description of a highly conserved proline switch in the force-sensing domain 20 of human filamin and how prolyl isomerization modulates the force-sensing mechanism. Proline isomerization toggles domain 20 between two conformations. A stable cis conformation with slow unfolding, favoring the autoinhibited closed conformation of filamin’s force-sensing domain pair 20–21, and a less stable, uninhibited conformation promoted by the trans form. The data provide detailed insight into the folding mechanisms that underpin the functionality of this binary switch and elucidate its remarkable efficiency in modulating force-sensing, thus combining two previously unconnected regulatory mechanisms, proline switches and mechanosensing.

|

force sensing molecular memory optical tweezers

| protein folding | single molecule |

O

wing to its ring structure (Fig. 1A), the peptidyl-prolyl bond is special among the peptide bonds in that it can populate slowly exchanging cis and trans conformations with similar probability (1). Therefore, proline residues can drastically affect the kinetics of folding (2, 3) as well as conformational changes of proteins (4, 5). Although most native protein conformations require trans prolines, there are a number of examples where native cis prolines have been reported (1, 6–8). In a few cases, both cis and trans conformations can be accommodated in the folded state thus leading to two distinct native states (9–11). Beyond its importance for protein folding, peptidyl-prolyl isomerization has also been discussed as a molecular timer switch regulating different functional conformations of proteins (5, 12). Filamins are essential actin cross-linkers as well as mechanosignaling hubs in the cytoskeleton. Recently, a special conformation in the rod 2 region of filamin has been shown to act as a force-sensing region mediating the interaction to transmembrane receptors like integrins or the von Willebrand receptor glycoprotein Ib (GPIb) (13–15) (Fig. 1B). Lad et al. (13) have shown that the first β-strand of domain 20 (FLNa20) of human filamin A binds to the subsequent domain 21 (FLNa21) thus covering and autoinhibiting its transmembrane receptor-binding site forming the domain pair shown in Fig. 1B. This domain pair has a mechanosensor function (15, 16). Mechanical force transmitted through the cytoskeleton leads to the opening of the domain pair and an increase in receptor-binding affinity of up to 17-fold (15). The crystal structure of FLNa20 exhibits a native cis proline just in front of the last β-strand (Fig. 1B, Inset). Interestingly, a multiple-sequence alignment shows that this cis proline residue is strictly conserved in all available 3D structures of filamin Ig-like domains (Fig. S1). Although proline isomerization has been investigated in numerous bulk studies, the structural heterogeneity caused by a native-state isomerization cannot be detected in crystal structures. 5568–5573 | PNAS | April 15, 2014 | vol. 111 | no. 15

Even in NMR experiments, native-state isomers can only be detected if both populations are considerably populated at equilibrium. Recent advances allowing the folding of individual protein molecules to be observed over hours using optical trapping experiments make it possible to observe slow processes, such as proline isomerization events. Here we show that the proline P2225 in FLNa20 (Fig. 1B, Inset) exhibits native-state proline isomerization. Single-molecule force spectroscopy with optical tweezers allows us to monitor individual cis–trans isomerization events in real time and to study the influence of mechanical load on this equilibrium. We show that mechanically induced isomerization into the trans state may lead to a prolonged weakening of the autoinhibition of the filamin domain pair, thus leading to a strong activation of transmembrane receptor binding. Results To study the force-dependent folding and unfolding dynamics of FLNa20, we designed a construct where FLNa20 was sandwiched between two ubiquitin domains each carrying a cysteine residue to allow attachment of the construct in a single-molecule optical tweezer experiment following published protocols (15, 17) (Fig. 1C; for details, see SI Materials and Methods). A series of subsequent stretch (blue and orange) and relax (gray) cycles obtained on one molecule of FLNa20 is shown in Fig. 1D. In stretch cycles, unfolding events can be observed both at high forces around 15 pN (traces in blue in Fig. 1D) as well as at low forces around 5 pN (traces in orange in Fig. 1D). Refolding readily occurs at forces around 3 pN in all traces at the pulling velocity used (200 nm/s). The histogram of unfolding forces (on Significance Biological processes in the cell are highly dynamic and complex, and their correct interplay is ensured by a multitude of regulatory mechanisms. Among these, proline isomerization acts as a molecular switch that toggles two protein conformations, and thus functions, over time. In mechanosensing, mechanical stress is transduced into chemical signals. The molecular mechanisms underlying this regulation are crucial for understanding cell behavior and development. However, only a few experimental techniques are capable of studying force at the molecular level. In this paper, we use single-molecule mechanical experiments to investigate how the force-sensing region of the cytoskeletal cross-linker filamin is modulated by a proline switch. Author contributions: L.R. and M.R. designed research; L.R. and T.M. performed research; L.R., G.Ž., and M.R. analyzed data; and L.R. and M.R. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1

To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1319448111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1319448111

Cδ

A

B

Pro2222

24

Cα

-ΔG 0

trans

1

15

cis

ω = 180° ω = 90° ω = 0°

rod 2

rod 1 avidin-biotin

disulfide bond 0.2

dig-antidig

unfolding forces

7s per cycle

20 200 nm/s

37s brea

k

te

state stable

20

Force [pN]

receptor binding site

Probability

25

15

19

ble sta

0.1

0.0

D

actin

20

DNA

Pro2225

21

ubiquitin

less sta

C

trans

cis

10 5 50 nm

0 0.4

0.2

0.0

E

Probability 200 nm/s

50 nm/s

10 nm/s

P2225G

5

0

50 nm

Fig. 1. FLNa20 shows a bimodal stability in single-molecule unfolding experiments. (A) Structural and energetic scheme of trans and cis isoform of a peptidylprolyl bond. (B, Left) Cartoon representation of a filamin dimer attached to actin. In the rod 2 region domains 19, 20, and 21 are color-coded dark gray, green, and orange, respectively. (B, Lower Right) Crystal structure of domains 19–21 of human filamin A (13). The receptor-binding region of domain 21 is marked in purple. (B, Inset) Zoom into the loop region in the G strand of domain 20 where P2222 and P2225 are located. (C) Schematic of the experimental setup. (D) Consecutive stretch-and-relax cycles of one single FLNa20 molecule separated by an x offset and moving-average filtered with a time window of 0.6 ms. Two different unfolding stabilities are observed. Extension traces are colored blue and orange for the stable and less stable states, respectively. Retraction traces with the refolding step are colored in light gray. (E) Stretch-and-relax cycles of FLNa20 with the proline-to-glycine mutation P2225G. The traces were recorded at three different speeds (200, 50, and 10 nm/s) and moving-average filtered with adequate time windows (0.6, 2.4, and 12 ms). At lower speeds consecutive unfolding and refolding events are observed during one extension cycle.

the left in Fig. 1D) confirms that the observed unfolding events fall into two distinct classes. Both classes of unfolding events exhibit the same contour length increase (Fig. S2). From this sample series, the switching kinetics between stable and unstable conformations can be estimated to several tens of seconds. Switching between conformations is reversible, hence permanent mechanical/chemical damage to the molecule cannot be an explanation for the observed effect. The observed switching kinetics is consistent with the proline isomerization kinetics seen in bulk experiments for oligopeptides (18). To test this hypothesis, we removed the cis proline in position 2225 in the mutant P2225G. In this mutant mimicking the trans state of P2225, we only observed the low-force population (Fig. 1E and Fig. S3). As a control, removing the trans proline P2222 in the mutant P2222G led to no observable difference to the wild type, and retained the bimodal stability distribution (Fig. S4). To observe proline cis–trans isomerization kinetics in real time with a single molecule, we performed experiments in the socalled passive mode. Herein, the trap centers were held at constant separation thus allowing the molecule to fluctuate between folded and unfolded states while the traps impose a constant average force bias (Fig. 2A, Upper). Due to the length change of the protein chain during unfolding and the involved bead relaxation in the trap potential, the average force bias will be different for the unfolded and folded protein. This difference is accounted for during analysis (for details see the supporting information in ref. 16). We find that the molecule switches Rognoni et al.

between two states: one where the unfolded state (Fig. 2A, gray) dominates with interspersed short-lived folding events (Fig. 2A, orange) and another stable state where the native state dominates (Fig. 2A, blue) with only rarely populated unfolded states; the assignment of the respective cis and trans states is marked in the bar above the trace with the trans states colored in dark gray and cis states in light gray, respectively. A zoom into the transition region between a cis and a trans state illustrates the differences in protein stability associated with isomerization (Fig. 2A, Inset). It is important to note that both folded and unfolded states are populated in both cis and trans conformation, albeit with drastically different population probabilities. We can identify four different states (Ncis, Ucis, Ntrans, and Utrans) with the lifetimes τNcis , τUcis , τNtrans , and τUtrans , respectively, which are shown in the extension vs. dwell-time scatter plot in Fig. 2A (Lower Left). In the cis state, we consistently observe the population of a short-lived intermediate state (arrow in Inset of Fig. 2A), which is off-pathway for the unfolding transition and hence, we lump this state into the lifetime of the native state. A full kinetic and thermodynamic description of proline isomerization and its coupling to folding requires knowledge of all rates shown in the thermodynamic box model given in Fig. 2B. The protein can apparently exist in two distinct folded (Ncis and Ntrans) as well as unfolded states (Ucis and Utrans). The horizontal transitions describe the folding/unfolding kinetics in the cis and trans states, whereas the vertical transitions describe proline isomerization rate constants in the native and unfolded state, respectively. PNAS | April 15, 2014 | vol. 111 | no. 15 | 5569

BIOPHYSICS AND COMPUTATIONAL BIOLOGY

Force [pN]

10

dependence of the cis–trans population on mechanical force can be observed. At high loads (Fig. 3A, Top) where the molecule is predominantly in the unfolded state, the trans state prevails. At lower loads, this population shifts strongly toward the cis state (Fig. 3A, Middle and Bottom). Apparently, the folding state of the protein strongly affects the cis–trans equilibrium of residue P2225. The assigned dwells in the cis and trans states now enable us to determine the load-dependent isomerization kinetics. It is important to note that the single-molecule traces do not directly allow the observation of isomerization events but rather we

cis

5 nm

Ncis

5 nm

U trans

kUN(F )

Utrans

0

5

10

15

20

0

5

10

15

20

0.0 0 1000 100 10 1 0.1 0.01 0.001 25 0

25

Force [pN]

2

4

6

8

10

7.4 pN

10 nm

0.0

1000 100 10 1 0.1 0.01 0.001

A

100 s

4.9 pN

2

4

6

8

10

Force [pN]

100 s

3.8 pN

100 s

10 0

B

U

kct(F ) 10-1

cis

Ncis

10-2

0

5

10

U

5570 | www.pnas.org/cgi/doi/10.1073/pnas.1319448111

trans

Ntrans

10-5

C

kct(F )

N

kct

Ucis

cis

kUN(F ) N

10-3 10-4

15

kNU(F )

Utrans

trans

kUN(F ) 20

25

10 0 N

10-1 ktc 10-2

app

ktc [1/s]

We first characterized the horizontal transitions, i.e., the force-dependent folding/unfolding kinetics in both cis and trans states (Fig. 2C). To separate cis from trans states, we used the vastly different lifetimes of their respective native states. The dwell-time scatter plot shows that, in the native state, two clouds (blue and orange in Fig. 2A) can be observed with lifetimes different by at least 2 orders of magnitude. Based on this criterion, we split the passive-mode traces and analyzed forcedependent folding kinetics as well as equilibrium energetics for both states separately (for details, see SI Materials and Methods and Fig. S5). Force-dependent chevron plots as well as population probabilities are shown in Fig. 2C. To enhance the accessible force range of our measurements to both the higher and lower ends for the cis-state folding kinetics, we also performed force-jump experiments, where the molecule was rapidly (within 10 μs) stepped to a high load for measurement (circles in Fig. 2C, Left). For details on jump experiments, see SI Materials and Methods and Fig. S6. For comparison, the passive-mode unfolding and folding results for the mutants P2225G and P2222G are given in Figs. S3 and S4. A summary of thermodynamic and kinetic parameters is in Table S1. We now turn to the vertical transitions in our kinetic model (Fig. 2B), describing the kinetics of cis–trans isomerization in native and unfolded states, respectively. Fig. 3A shows the population of cis (light gray) and trans states (dark gray) in passivemode traces obtained at different biasing forces (see Fig. S7, Right for zooms into the cis–trans transition region). A strong

kNU(F )

kct

Fig. 2. Force-dependent equilibrium folding kinetics of FLNa20. (A) Passivemode trace of FLNa20 at a force bias of 4.6 pN (acting on the unfolded protein). The colored trace corresponds to 5-kHz data where the stable longlived native state Ncis is colored in blue, the less stable short-lived Ntrans in orange, and both unfolded states Ucis and Utrans in light gray. The black trace is moving-average filtered with a 4.2-ms time window. (A, Upper) The light and dark gray line denotes the cis and trans states. The Inset (A, Inset) shows a cis-to-trans transition and illustrates the different folding stabilities. A short-lived off-pathway intermediate observed in the cis state is marked with an arrow. (A, Lower Left) A deflection vs. dwell-time scatter plot shows that the two native states Ncis and Ntrans are kinetically well separated. (B) Kinetic model for the folding and prolyl isomerization of FLNa20. (C) Unfolding and folding of cis (Left) and trans state (Right). (C, Upper) Forcedependent equilibrium probabilities for the native (colored) and unfolded state (gray). (C, Lower) Force-dependent unfolding (colored) and folding (gray) rates measured in passive mode (triangles) and jump experiments (circles). Error bars represent the SEM.

10 nm

ktc(F )

kNU(F )

0.5

10 nm

Ntrans

U

kct(F )

N trans

0.5

U

Ucis

cis

kUN(F )

kct

ktc

Ncis

1.0

ktc(F )

kNU(F )

1.0

N

C

cis

ktc

B

N

2s

10 100

Dwell time [s]

1 pN

0.01 0.1 1

1 pN

Ucis and Utrans

42

Probability

46

Ntrans

Rate [s-1]

50

1 pN

Deflection [nm]

50 s

1 pN

1 pN

4.6 pN

app

trans

kct [1/s]

A

U

ktc(F )

10-3 10-4 10-5 0

2

4

6

Force [pN]

8

10

Fig. 3. Force-dependent equilibrium isomerization kinetics. (A) Passivemode traces at different force biases (given for the unfolded protein). Graycolored traces represent 5-kHz data where the cis and trans states are colored in light and dark gray, respectively. With rising force bias (from Bottom to Top) a drastic shift in the cis–trans population from cis to trans is observed. The black trace is moving-average filtered with a 4.2-ms time window. (B) Force-dependent cis-to-trans isomerization rates measured in passive mode (triangles) and jump experiments (circles) plotted against the force bias acting on the unfolded protein. The dashed line corresponds to U N ðFÞ, the dotted line to kct , and the solid line to Eq. 1. (C) Force-dependent kct trans-to-cis isomerization. Error bars represent the SEM.

Rognoni et al.

N U kapp ct ðFÞ = ð1 − PUcis ðFÞÞ · kct + PUcis ðFÞ · kct ðFÞ

[1]

N U kapp tc ðFÞ = ð1 − PUtrans ðFÞÞ · ktc + PUtrans ðFÞ · ktc ðFÞ

[2]

and

where PUcis ðFÞ and PUtrans ðFÞ are the measured force-dependent population probabilities for the unfolded protein shown in Fig. U N U 2C. kN ct , kct ðFÞ, ktc , and ktc ðFÞ are the microscopic isomerization rate constants for the unfolded (U) and native (N) states. The N isomerization rate constants for the native state (kN ct and ktc ) can be assumed to be force-independent because in the folded structure the prolyl bond does not bear much load as force propagates through multiple parallel β-sheets. However, in the unfolded structure the force acts directly on the prolyl bond and is therefore force dependent. Hence, the apparent isomerization rates app app (kct and ktc ) are plotted against the force acting on the unfolded protein. U The force dependence of kU ct ðFÞ and ktc ðFÞ can be described with the Bell model (19):   F · Δxct U;0 U [3] kct ðFÞ = kct · exp − kB T and U;0 kU tc ðFÞ = ktc

  F · Δxtc · exp − kB T

[4]

where F is the force applied to the unfolded protein, kB is the Boltzmann constant, and T the absolute temperature. kU;0 ct and kU;0 tc are the isomerization rates for the unfolded protein at zero load. Finally, the distances from the cis and trans states to the transition state are given by Δxct and Δxtc , respectively. app In Fig. 3B we plot the observed apparent cis–trans rate (kct ) app and in Fig. 3C the apparent trans–cis rate (ktc ). To compensate for deviations in trap stiffness calibration among different molecules, we corrected the forces acting in the measurement by using the load-dependent population probabilities of the trans state as determined in Fig. 2C (Upper Right). This recalibration procedure reduced the force scatter in the rate plots of Fig. 3 B and C considerably. As described above, in addition to the passive-mode experiments (triangles in Figs. 2C and 3B) we also performed force-jump experiments to enhance the accessible force range (circles in Fig. 3B; for details see SI Materials and Methods, Figs. S6 and S8). A fit to Eqs. 1 and 2 yields the forceU dependent vertical rate constants kU ct ðFÞ and ktc ðFÞ and the . Because in Fig. 2B the force-independent rate constant kN tc lowest accessible data point is around forces of 4 pN, a fit cannot provide a very good estimate for kN ct . However, because in a cyclic thermodynamic path at equilibrium, the product of the equilibrium constants must equal 1, kN ct is not a free parameter anymore and can be directly calculated from the principles of detailed balance (20). A summary of all rate constants including their force dependence can be found in Table S2. We find that the folding state of the protein exclusively affects the cis–trans isomerization. When the protein is folded, cis–trans isomerization (kN ct Þ is slower by more than 3 orders of magnitude compared N U with the unfolded state [kU ct ðFÞ]. In contrast, ktc and ktc ðFÞ are Rognoni et al.

identical within resolution of our measurement indicating that the folding state of the protein domain does not influence trans– cis isomerization. The slopes of the plots in Fig. 3 B and C at high loads (where the protein is unfolded) also allow estimating how force couples directly to the isomerization kinetics in the unfolded state. The extremely shallow slope (dashed line in Fig. 3B) indicates a very weak coupling to force, consistent with transition-state positions on the order of around 1 Å (21, 22). Discussion Proline isomerization is a ubiquitous phenomenon in classical assays of protein folding (1). It may therefore be surprising that proline isomerization has so far been elusive in single-molecule protein-folding studies. A simple reason may be that the accessible observation timespans in single-molecule studies have often been limited to less than a minute. The development of ultrastable optical traps has made it possible to observe single molecules over minutes up to hours at time resolutions in the submillisecond range (17, 23–26). Hence this technique is ideally suited to study, on the single-molecule level, the slow-timescale heterogeneity that proline isomerization imposes on protein folding. Furthermore, the unambiguous extraction of rates in complex thermodynamic reaction schemes can be tedious and is, in many cases, impossible using bulk methods. For the cyclic scheme in Fig. 2B, all eight kinetic rate constants cannot be extracted without assumptions (27). We would like to point out that the direct access to rate constants possible in single-molecule experiments overcomes such limitations, as we show in this study. Proline isomerization rates have been studied extensively for peptides using spectroscopic methods (28). Whereas the details of isomerization kinetics depend on the sequence of the flanking amino acids, in general, the kinetics is in the range of several peptide minutes near 15 °C with ktc being a factor of 1 to 9 slower peptide than kct (28). The kinetics of proline isomerization we find at −1 −1 28 °C for the unfolded state (kU;0 and kU;0 ct = 0:09  s tc = 0:019  s ) lie well within this range if we consider the acceleration of the isomerization due to higher temperature. Although in folded protein structures the majority of prolines are found in the trans state, there are a number of reported protein structures where the proline exhibits cis conformation. Among those, a small subset tolerates both proline conformers, albeit with different folding free energies and sometimes even different structures (9). Such conformational heterogeneities are difficult to detect by structural methods if the relative population of the two nativestate conformations differ significantly (29). For FLNa20, we find that the isomerization state of P2225 mostly affects the unfolding −1 rate of the native state (kcis;0 vs. ktrans;0 = 2:1  s−1 ), NU = 0:023  s NU whereas folding rates are similar for cis and trans conformations −1 (kcis;0 vs. ktrans;0 = 89  s−1 ). Hence, the proline residue is UN = 410  s UN likely not structured in the transition state of folding. The difference in free energy between the two states is 7.4 kBT making Ncis more than three times more stable than Ntrans. In contrast, in a study of the N2 domain of the gene-3-protein of the filamentous phage fd by Jakob et al. (9), the isomerization state of proline 161 was reported to alter folding kinetics while leaving unfolding kinetics unaffected. Hence, in this protein, the proline residue is structured in the transition state of folding. Proline isomerization not only affects folding, but in turn the isomerization process is strongly influenced by the folding state of the protein (9). Hence, in our case, the free energy liberated during the folding process can stabilize the cis over the trans app conformation by either speeding up kapp tc or slowing down kct or app a mixture of both. In FLNa20, we find that only kct is slowed −5 −1 down by more than 3 orders of magnitude (kN vs. ct = 5:5 × 10   s app −1 kU;0 = 0:09  s ) whereas k is unaffected. Even though we do ct tc not have a structural interpretation for the clear partitioning of this effect into only one rate constant, the contrasting force dependence PNAS | April 15, 2014 | vol. 111 | no. 15 | 5571

BIOPHYSICS AND COMPUTATIONAL BIOLOGY

deduce an isomerization event indirectly through the altered folding/unfolding kinetics as described above. Therefore, we cannot determine whether an isomerization event has occurred in the native or unfolded state of the protein. Hence, the isomapp erization kinetics we observe is an apparent rate kct reflecting the sum of the rates of isomerization in the unfolded and in the native state according to

app

app

of kct and ktc might suggest that the structural environments around P2225 are so distinct that the transition of trans to cis is not directly catalyzed, whereas the structured environment around cis P2225 directly increases the lifetime of the cis state. Another exapp planation rationalizing the lack of effect of folding on ktc may be the superposition of compensating effects. An increased freeapp energy bias toward cis would increase ktc , but the constraints of a folded protein would slow down conformational changes. In the app case of kct , these effects would reinforce each other to reduce the rate. A clear distinction between the two mechanisms is not possible at this point. It has been suggested that mechanical forces may affect the function of mechanosensitive proteins by inducing cis to trans switching of proline residues (21). Because the backbone length of a cis state of a proline is slightly shorter than in the trans state (on the order of around 1 Å) (30), one would expect that, in an unfolded polypeptide, mechanical load shifts the equilibrium even further toward the trans state. This effect can be observed in the high-force region of the plots in Fig. 3 B and C, where app app U 1. A slight slope kct ≈ kU ct and ktc ≈ ktc because PN ≈ 0 and PU ≈ app of the rate constants can be observed while kct increases with app force and ktc decreases. Consistent with the small ∼1-Å length change associated with proline switching, these slopes are very shallow. A more detailed analysis would require the application of much higher loads as has been shown in molecular dynamics experiments (21). As discussed above, we assume the isomeriN zation rates for the folded domain (kN ct and ktc ) to be force independent. If we apply an alternative model where, in the folded state, cis–trans isomerization has the same force dependence as in the unfolded state, the conclusions as well as extrapolated rates do change only very marginally. The reason is that the protein populates a folded state significantly only at forces below 3–5 pN. A coupling by a length of 1.7 Å would change the extrapolated zero-force values by as little as a factor of 1.1, which is indeed negligible. Changing pulling direction (26, 31) might give additional insight into the force dependence of isomerization in the folded state. Because the forces acting on filamin in its cytoskeletal environment are likely lower than tens of pN, we exclude force regulation through direct mechanical action on prolyl bonds. Our measurements suggest, however, that a much more sensitive regulation of protein states through mechanical forces can be achieved by the mechanical unfolding of a protein. The plot of app the force-dependent change of kct in Fig. 3B shows that in the region between 3 and 7 pN, where the cis conformation of FLNa20 unfolds, the switching rates of P2225 changes by more than 3 orders of magnitude. In this context it is important to remember that FLNa20 together with its adjacent domain FLNa21 constitute a mechanosensor domain pair as described in the Introduction (Fig. 4, Lower Left). At forces around 4 pN, the domain pair 20–21 opens up and thus exposes and activates the integrin/GPIb-binding site on domain 21 while both domains still remain folded (Fig. 4, Lower Center). At higher force, domain 20 will unfold and destroy the environment that keeps P2225 stably in its cis conformation (Fig. 4, Lower Right). If forces act for a prolonged period of time, domain 20 will remain unfolded and a cis-to-trans switch can occur within a few seconds. Now, already very low forces will shift the equilibrium of domain 20 far into the unfolded state, preventing the autoinhibition of the binding site on domain 21 over tens of seconds until the proline switches back and FLNa20 folds into the more stable state Ncis, which allows the autoinhibition through domain-pair formation. Interestingly, a recent study using magnetic tweezers has suggested that FLNa20 may exist in two different unfolded forms (32). Here, we provide evidence that the underlying mechanism is in fact proline isomerization. Even though it cannot be ruled out, there is currently no evidence for a direct influence of the 5572 | www.pnas.org/cgi/doi/10.1073/pnas.1319448111

F

F F

21 21

19

c 20 no stress binding site blocked

21

c 20 low stress competition of receptor and blocking A-strand

c

t

20

high stress binding site accessible for minutes

Fig. 4. The proline switch modulates force sensing. When unstrained FLNa20 covers the receptor-binding site present in FLNa21 (Lower Left). Mechanical force in a strained cytoskeleton will lead to domain pair opening in the rod 2 segment of filamin, allowing binding to membrane receptors (Lower Center). Under high stress FLNa20 unfolds and P2225 can isomerize to trans, leading to a native state (Ntrans ) with drastically reduced stability (Lower Right). The binding site would stay accessible even at lower forces up to minutes until P2225 isomerizes back into the cis state.

proline state on the interaction of FLNa20’s A strand with domain 21. Intriguingly, the presence of a cis proline within the G strand is not exclusive to human FLNa20. A closer look at the structures of 36 different filamin domains, including domains 4–5 of Dictyostelium discoideum reveal a cis proline in the same domain region. A multiple-sequence alignment shows that this proline is the only residue that is 100% conserved in filamins (Fig. S1). It is often (80%) followed by a bulky aromatic residue (Tyr or Phe) known to stabilize the proline cis isomer (33). This strong conservation is remarkable given an overall sequence identity of only 26% on average, and because cis prolines occur in less than 5% of the proteome (34). For example, Ig-like domains from other proteins like fibronectin, myomesin, or titin do not exhibit this cis proline. In the case of trans prolines, a study investigating the sequence conservation in fibronectin type III domains showed that proline-to-alanine mutations had no effect on the folding stability, but increased aggregation (35). The proline-to-glycine mutations of FLNa20 presented here had no observable effect on aggregation in our experiments, at least on a timescale of days and at room temperature. Even though at this point we cannot assess the relevance of the cis–trans switch for all of the other 23 domains of filamin, our results for domain 20 highlight the potential consequences of such a switch for modulating protein stability and the mechanosensing of this important cytoskeletal protein. A proline-dependent switch mechanism has also been suggested for the shear-sensing domain A2 of von Willebrand factor (VWF), albeit with no direct experimental evidence (36). A2 contains a native cis proline, which may isomerize to trans after shear stress has induced unfolding of domain A2. This isomerization would lead to greatly delayed refolding, thus expediting cleavage by ADAMTS13 of VWF during vessel repair. These examples suggest an emerging general role for proline cis–trans switches in mechanosensing. Rognoni et al.

attached. For the single-molecule mechanical measurements, a dumbbell conformation was generated by attaching the biotin/digoxigenin functionalized end of the handles to micrometer-sized streptavidin/anti-digoxigenin silica beads (Fig. 1C). The beads were trapped in the foci of a custom-built dual beam optical-tweezer setup and subjected to stretch-and-relax cycles or a constant force bias with fixed trap positions. To extend the accessible force range a jump protocol was used. All measurements were performed in PBS (10 mM phosphate buffer, 2.7 mM potassium chloride, and 137 mM sodium chloride, pH 7.4). Transitions between states were detected using a hidden Markov model analysis on the unfiltered raw data of the difference signal of the two traps. A complete description of the methods used is given in SI Materials and Methods.

Materials and Methods The investigated constructs of human filamin A were genetically inserted between two ubiquitins with terminal cysteines that served as spacers. To the terminal cysteines thiol-modified DNA handles of a length of 180 nm were

ACKNOWLEDGMENTS. This work was supported by an Sonderforschungsbereich 863 A2 Grant and ZO 291/1-1 Grant from the Deutsche Forschungsgemeinschaft. M.R. acknowledges financial support through an instrument grant from the Institute for Advanced Study of Technische Universität München.

1. Wedemeyer WJ, Welker E, Scheraga HA (2002) Proline cis-trans isomerization and protein folding. Biochemistry 41(50):14637–14644. 2. Brandts JF, Halvorson HR, Brennan M (1975) Consideration of the Possibility that the slow step in protein denaturation reactions is due to cis-trans isomerism of proline residues. Biochemistry 14(22):4953–4963. 3. Brandts JF, Brennan M, Lung-Nan Lin (1977) Unfolding and refolding occur much faster for a proline-free proteins than for most proline-containing proteins. Proc Natl Acad Sci USA 74(10):4178–4181. 4. Lummis SCR, et al. (2005) Cis-trans isomerization at a proline opens the pore of a neurotransmitter-gated ion channel. Nature 438(7065):248–252. 5. Eckert B, Martin A, Balbach J, Schmid FX (2005) Prolyl isomerization as a molecular timer in phage infection. Nat Struct Mol Biol 12(7):619–623. 6. Kiefhaber T, Grunert HP, Hahn U, Schmid FX (1990) Replacement of a cis proline simplifies the mechanism of ribonuclease T1 folding. Biochemistry 29(27):6475–6480. 7. Tugarinov V, Zvi A, Levy R, Anglister J (1999) A cis proline turn linking two betahairpin strands in the solution structure of an antibody-bound HIV-1IIIB V3 peptide. Nat Struct Biol 6(4):331–335. 8. Charbonnier JB, Belin P, Moutiez M, Stura EA, Quéméneur E (1999) On the role of the cis-proline residue in the active site of DsbA. Protein Sci 8(1):96–105. 9. Jakob RP, Schmid FX (2008) Energetic coupling between native-state prolyl isomerization and conformational protein folding. J Mol Biol 377(5):1560–1575. 10. Pappenberger G, et al. (2003) Kinetic mechanism and catalysis of a native-state prolyl isomerization reaction. J Mol Biol 326(1):235–246. 11. Sarkar P, Reichman C, Saleh T, Birge RB, Kalodimos CG (2007) Proline cis-trans isomerization controls autoinhibition of a signaling protein. Mol Cell 25(3):413–426. 12. Lu KP, Finn G, Lee TH, Nicholson LK (2007) Prolyl cis-trans isomerization as a molecular timer. Nat Chem Biol 3(10):619–629. 13. Lad Y, et al. (2007) Structure of three tandem filamin domains reveals auto-inhibition of ligand binding. EMBO J 26(17):3993–4004. 14. Pentikäinen U, Ylänne J (2009) The regulation mechanism for the auto-inhibition of binding of human filamin A to integrin. J Mol Biol 393(3):644–657. 15. Rognoni L, Stigler J, Pelz B, Ylänne J, Rief M (2012) Dynamic force sensing of filamin revealed in single-molecule experiments. Proc Natl Acad Sci USA 109(48):19679–19684. 16. Ehrlicher AJ, Nakamura F, Hartwig JH, Weitz DA, Stossel TP (2011) Mechanical strain in actin networks regulates FilGAP and integrin binding to filamin A. Nature 478(7368):260–263. 17. Stigler J, Ziegler F, Gieseke A, Gebhardt JCM, Rief M (2011) The complex folding network of single calmodulin molecules. Science 334(6055):512–516. 18. Grathwohl C, Wüthrich K (1981) NMR studies of the rates of proline cis–trans isomerization in oligopeptides. Biopolymers 20(12):2623–2633. 19. Bell GI (1978) Models for the specific adhesion of cells to cells. Science 200(4342): 618–627.

20. Wegscheider R (1911) Über simultane Gleichgewichte und die Beziehungen zwischen Thermodynamik und Reactionskinetik homogener Systeme. Monatshefte für Chemie/ Chemical Monthly 32(8):849–906. 21. Chen J, Edwards SA, Gräter F, Baldauf C (2012) On the cis to trans isomerization of prolyl-peptide bonds under tension. J Phys Chem B 116(31):9346–9351. 22. Valiaev A, Lim DW, Oas TG, Chilkoti A, Zauscher S (2007) Force-induced prolyl cis-trans isomerization in elastin-like polypeptides. J Am Chem Soc 129(20):6491–6497. 23. Moffitt JR, Chemla YR, Izhaky D, Bustamante C (2006) Differential detection of dual traps improves the spatial resolution of optical tweezers. Proc Natl Acad Sci USA 103(24):9006–9011. 24. Wang MD, Yin H, Landick R, Gelles J, Block SM (1997) Stretching DNA with optical tweezers. Biophys J 72(3):1335–1346. 25. Cecconi C, Shank EA, Bustamante C, Marqusee S (2005) Direct observation of the three-state folding of a single protein molecule. Science 309(5743):2057–2060. 26. Shank EA, Cecconi C, Dill JW, Marqusee S, Bustamante C (2010) The folding cooperativity of a protein is controlled by its chain topology. Nature 465(7298):637–640. 27. Ikai A, Tanford C (1973) Kinetics of unfolding and refolding of proteins. I. Mathematical analysis. J Mol Biol 73(2):145–163. 28. Zoldák G, et al. (2009) A library of fluorescent peptides for exploring the substrate specificities of prolyl isomerases. Biochemistry 48(43):10423–10436. 29. Jakob RP, Schmid FX (2009) Molecular determinants of a native-state prolyl isomerization. J Mol Biol 387(4):1017–1031. 30. Reimer U, Fischer G (2002) Local structural changes caused by peptidyl-prolyl cis/trans isomerization in the native state of proteins. Biophys Chem 96(2-3):203–212. 31. Dietz H, Berkemeier F, Bertz M, Rief M (2006) Anisotropic deformation response of single protein molecules. Proc Natl Acad Sci USA 103(34):12724–12728. 32. Chen H, et al. (2013) Mechanical perturbation of filamin A immunoglobulin repeats 20-21 reveals potential non-equilibrium mechanochemical partner binding function. Sci Rep 3:1642. 33. Meng HY, Thomas KM, Lee AE, Zondlo NJ (2006) Effects of i and i+3 residue identity on cis-trans isomerism of the aromatic(i+1)-prolyl(i+2) amide bond: Implications for type VI beta-turn formation. Biopolymers 84(2):192–204. 34. Lorenzen S, Peters B, Goede A, Preissner R, Frömmel C (2005) Conservation of cis prolyl bonds in proteins during evolution. Proteins 58(3):589–595. 35. Steward A, Adhya S, Clarke J (2002) Sequence conservation in Ig-like domains: the role of highly conserved proline residues in the fibronectin type III superfamily. J Mol Biol 318(4):935–940. 36. Zhang Q, et al. (2009) Structural specializations of A2, a force-sensing domain in the ultralarge vascular protein von Willebrand factor. Proc Natl Acad Sci USA 106(23): 9226–9231. 37. Schiene-Fischer C, Aumüller T, Fischer G (2013) Peptide bond cis/trans isomerases: a biocatalysis perspective of conformational dynamics in proteins. Top Curr Chem 328:35–67.

Rognoni et al.

PNAS | April 15, 2014 | vol. 111 | no. 15 | 5573

BIOPHYSICS AND COMPUTATIONAL BIOLOGY

The combination of high sensitivity and stability that can be achieved with state-of-the-art optical trapping has allowed us to study the full kinetics and energetics of the coupling of a highly conserved proline switch to protein folding in a mechanosensor protein. More complex and physiologically relevant protein systems can now be studied in single-molecule mechanical assays. A whole enzyme machinery is involved in catalyzing cis–trans isomerization of proteins in vivo (11, 12, 37). Therefore, it will be an important task for the future to study the influence of cis/trans-prolyl isomerases (e.g., FK506 bindig protein, cyclophilin, parvulin) on peptide chains directly on the singlemolecule level.