The C2 Domains of Otoferlin, Dysferlin, and Myoferlin Alter the Packing of Lipid Bilayers

Materials

Lipids were obtained from Avanti polar Lipids (Alabaster, AL). Affinity media Ni2+-Sepharose high performance beads and amylose resin were obtained from GE-Amersham Biosciences (Pittsburgh, PA). Common reagents and primers were purchased from Sigma (St. Louis, MO).

Protein expression and purification

All primers were purchased from Sigma Aldrich. All constructs were verified by sequencing by the Center for Genome Research and Biocomputing core facility at Oregon State University. pcDNA4/TO/mGFP-dysferlin-myc-his (gift from K. Bushby Newcastle, U.K) pcDNA 3.1 myoferlin-HA (gift from W. Sessa Yale School of Medicine), and pcDNA3.1 otoferlin plasmid (gift from C. Petit, Institut Pasteur et Université Pierre et Marie Curie, France) were used as templates for amplification of Homo sapiens dysferlin (GenBank: AF075575.1), Homo sapiens myoferlin (AF075575.1), and Mus Musculus otoferlin (GenBank: AY586513.1), respectively. Forward and reverse primers were designed to amplify the coding sequences with 5′ and 3′ sites for LIC insertion into SspI/T4-treated pMCS69. Primers for the C2 domains of otoferlin, dysferlin and myoferlin were based upon the predicted domain boundaries reported previously²⁰. For multi-domain constructs, the primers were designed for the following amino acids; dysferlin-C2ABC 1-574aa otoferlin-C2ABC 1-553aa, otoferlin-C2DEF 960-1885aa, otoferlin C2ABCDEF 1-1885aa, myoferlin C2ABC 1-530aa. Dysferlin C2A D16A was constructed using Stratagene Quickchange site-directed mutagenesis kit with pMCSG9/6His-MBP-dysferlin C2A plasmid template. Resulting constructs were expressed as a fusion with 6his-MBP (maltose binding protein). Multi-domain constructs were transformed into Rosetta BL2, while single C2 domain constructs we transformed into BL21. Expressed 6his- tagged constructs were affinity-purified using nickel IMAC resin followed by buffer exchange into 20mM Tris-HCl, pH 7.5, 100 mM NaCl. The otoferlin construct consisting of all six C2 domains required an additional affinity purification step using amylose resin after nickel IMAC purification. A TEV protease site allowed for cleavage of the MBP purification tag when desired, however studies on the protein with and without the MBP tag indicated that the MBP did not affect the results. We note that a shorter dysferlin C2ABC construct with amino acids 1 – 528 resulted in degradation products during purification, necessitating the longer 1 – 574 construct.

Small unilamellar vesicle preparation

Vesicles were prepared as described previously^13,15. Briefly, chloroform dissolved lipids were mixed in the desired ratio and dried under vacuum until the solvent was removed. The dried lipids were then rehydrated in buffer to a concentration of 1mM and extruded using a membrane with a 50 nm cutoff. Extruder, syringes, and membranes were purchased from Avanti Polar Lipids.

Giant plasma membrane vesicle preparation

Vesicles were prepared as described previously using HEK293 cells²¹. Briefly, HEK293 cells cultured in DMEM media were washed with PBS and subsequently incubated with 2 mM DTT and 25 mM formaldehyde and small (micromolar) amounts of laurdan at 37 °C for 60 minutes. The resulting vesicles were imaged under bright field using a epi-fluorescent microscope to verify the formation of vesicles.

Protein fluorescence measurements

Steady state fluorescence measurements were conducted using a QM-40 with Glan Thompson polarizers (Photon Technology International, Birmingham, NJ) at 37 °C. All fluorescence measurements were conducted in samples containing 20mM Tris-HCl or HEPES buffer, pH 7.5, with 150 mM NaCl. Excitation at 350 nm was used for laurdan, and the generalized polarization value was calculated using the following equation

where I430 and I480 are the emission intensities at 430 and 480 nm. Anisotropy measurements were calculated using the following equation

where IVV and IVH correspond to the parallel and perpendicular emission intensities. Reported values represent the mean ± S.D. for three or more samples. Each sample was measured multiple times to ensure that the system was not changing over time.

Otoferlin alters the lipid order and fluidity of small unilamellar vesicles

Otoferlin is directly involved in exocytosis and neurotransmitter release through an as-yet unclear set of mechanisms^5,6,15,22,23. Synaptotagmin, a C2 domain protein also involved in neurotransmitter release actively manipulates membrane structure to achieve exocytosis¹³. It is possible that otoferlin operates through similar mechanisms. However, while co-floatation assays have demonstrated the membrane binding ability of otoferlin, no study has tested whether binding of otoferlin has a downstream effect on membrane structure^6,15. To determine if membrane bound ferlin proteins change the structure of lipid bilayers, small unilamellar vesicles (SUV) composed of 25% 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS), 74% 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), and ~ 1% laurdan were mixed with recombinant otoferlin constructs and monitored using a steady state fluorometer. Laurdan is a small, solvatochromic fluorescent membrane probe with an emission maxima that is sensitive to the ordering of lipids in bilayers^18,19, (Fig. 2a). When inserted into membranes, the laurdan fluorescent moiety is located at the level of the phospholipid glycerol backbone, and the emission spectrum is highly sensitive to hydration of this region of the bilayer. When immersed in a highly ordered lipid bilayer, the laurdan emission spectra possesses a maxima at 430 nm, while in more disordered (greater degree of hydration) bilayers the maxima occurs at longer wavelengths, typically in the range of 480 nm. The normalized intensity ratio at these two wavelengths is often referred to as the General Polarization (GP) value, and ranges from −1 to + 1, representing the most disordered and most ordered states respectively. GP values are not biased by phospholipid headgroup charge, nor are they affected by protein penetration into lipid bilayers^18,19,24. Further, laurdan does not associate preferentially with any specific lipids^18,19. These characteristics make it an ideal probe of the global properties of lipid bilayer organization. Unlike electron microscopy imaging of vesicles, which can suffer from artifacts due to dehydration and heavy metal staining, measurements using laurdan can be performed in solution.

As illustrated schematically in Fig. 2b, and summarized in Fig. 2c,d, addition of 10 μM of maltose binding protein (MBP) did not elicit any shift in the laurdan spectra while 2 μM of a recombinant otoferlin construct composed of the entire cytoplasmic region of the protein (C2ABCDEF) blueshifted the emission spectra, suggesting that the effects of otoferlin are protein specific. The magnitude of the observed shift was calcium enhanced, although in the absence of calcium an attenuated shift in the spectra was observed, suggesting some calcium independent activity (Fig. 2d). To test whether the observed changes in the spectra could arise due to SUV aggregation by otoferlin, we monitored the change in GP upon addition of 5 μM avidin to SUV harboring 1% 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-biotinyl (biotin-PE). No significant change in GP was observed due to biotin-avidin induced aggregation of vesicles (Fig. 2d). To further characterize the role of calcium in enhancing the effects of otoferlin on SUV, we conducted calcium and magnesium titrations (Fig. 2e). Calcium enhanced the effect of otoferlin on SUV in a dose dependent manner with a Kd of 9.3 ±0.3 μM. Under the range of concentrations tested (0 – 80 μM), magnesium did not elicit a large change in the GP, indicating that the effects of calcium on otoferlin activity is divalent cation specific. Based upon the MBP and biotin-avidin control studies, we conclude that otoferlin specifically interacts with SUV and directly perturbs the structural state of the lipid bilayers.

We next determined which regions of otoferlin harbor this functional activity by testing constructs composed of either the first three C2 domains (C2ABC) or the last three C2 domains (C2DEF) (Fig. 3a,b). When tested, both the C2ABC and C2DEF protein fragments blueshifted the spectra in a dose dependent manner, with the C2DEF construct appearing more potent within the concentration range tested (0–4 μM). Thus, domains within both the N- and C-terminus of the protein alter the lipid packing of synthetic SUV.

It is generally believed that laurdan is not sensitive to peripheral membrane binding to lipid headgroups, but rather to the physical state of the membrane^18,19,24. As an additional test of whether the changes in laurdan spectra are due to alterations in lipid bilayer structure, we measured the anisotropy of the laurdan probe in the presence and absence of the otoferlin proteins. Whereas the laurdan GP is a reflection of the order of the bilayer, anisotropy values reflect the fluidity of the bilayer²⁵. For both constructs, an increase in the fluorescence anisotropy upon addition of 5μM protein was observed, indicating that the rotation of the laurdan probe was slowed (Fig. 3c). In accordance with the observed GP values for C2ABC and C2DEF, the change in anisotropy was greater for the C2DEF region compared to the C2ABC construct. By contrast, no change in anisotropy was observed when either 5μM bovine serum albumin or 5 μM MBP were tested. The decreased rotation of the laurdan probe upon protein binding to SUV supports the conclusion that the proteins directly alter the physical properties of the lipid bilayer including the fluidity.

The experiments in Fig. 2 and 3a–c were conducted with SUV composed of POPC and POPS. However, the compositions of cellular membranes are significantly more diverse. To ascertain whether the effects of the protein constructs are relevant in more complex systems, laurdan was added to giant plasma membrane vesicles (GPMV) derived from HEK293 cells and mixed with each of the multi-domain constructs. GPMV are large micrometer sized unilamellar vesicles chemically blebbed from cells, and possess many of the endogenous lipids found in the plasma membrane, although it has been demonstrated that blebbing induces mixing of the inner and outer leaflet lipids of the plasma membrane bilayer^21,26. Visual inspection of the GPMV spectra in the absence of otoferlin show a peak in the 440–450nm range, suggesting that unlike the SUV, the GPMV membranes possess a greater degree of lipid packing, most likely due to the ordering effect of cholesterol (Fig. 3d). However, addition of 5 μM otoferlin C2ABC or C2DEF further shifted the emission maxima to shorter wavelengths, (Fig. 3d). Addition of 10 μM MBP did not shift the laurdan emission profile, and served as a negative control. These results are in qualitative agreement with the conclusions based upon measurement with SUV, and demonstrate that even in a more complex “cell-like” lipid mixture, the lipid ordering activities of otoferlin are still observed.

Conservation of activity among otoferlin, dysferlin and myoferlin C2 domains

Otoferlin is one member of the ferlin family, and we next sought to determine if the results obtained with otoferlin are generalizable to other members. We therefore extended our laurdan measurements to include dysferlin and myoferlin. Specifically, we restricted our studies to the comparison of the N-terminus C2ABC region of these proteins, as a recent study found greater sequence divergence within this region as compared to the C-terminus of the proteins²⁰. When tested, 5 μM of constructs composed of the first three C2 domains (C2ABC) of both myoferlin and dysferlin blueshifted the spectra of laurdan labeled SUV in a calcium enhanced manner similar to that of otoferlin C2ABC (Fig. 4). Thus, the membrane altering activity observed with otoferlin appears to be a general property common to members of the ferlin family.

We next compared the GP values of SUV mixed with 5 μM C2ABC to SUV samples mixed with 5μM C2AB and 5μM C2C. The GP of laurdan SUV in the presence 5 μM otoferlin C2ABC was determined to be only slightly higher than that of samples containing both 5 μM otoferlin C2AB and 5 μM C2C suggesting that a small amount of lipid ordering activity is lost by the separation of C2AB from C2C (Fig. 4a). However, in similar experiments with dysferlin and myoferlin, the GP is shifted similarly in the presence of 5 μM C2ABC compared to 5 μM C2AB with 5μM C2C (Fig. 4b, c). Overall, it appears that separation of C2C from C2AB results in a reduction in activity roughly equivalent to the activity of isolated C2C, and thus the effects of C2C are additive to those of C2AB.

Having established the activity of multi-domain constructs, we next compared equivalent individual C2 domains across ferlin family members. Starting with C2A, we observed that otoferlin did not exhibit appreciable lipid ordering activity at the three concentrations tested (5, 15, 30 μM), while the C2A domains of dysferlin and myoferlin did alter the laurdan signal in a dose-dependent manner, with dysferlin C2A appearing slightly more potent than myoferlin (Fig. 5a). All C2B and C2C domains of otoferlin, dysferlin, and myoferlin shifted the laurdan spectra of SUV in a dose dependent manner, with the C2C domains appearing more potent than the C2B domains (Fig. 5b, c). Thus, with the exception of otoferlin C2A, all the examined C2 domains shifted the laurdan spectra in a calcium enhanced manner (Fig. 5a–c).

A previous report predicted that a mutation to residue D16 in the C2A domain of dysferlin would result in loss of function, and it is hypothesized that this aspartate is involved in calcium the binding activity of the domain¹⁷. As an additional test to determine the role of calcium on the lipid ordering activity of the C2 domains, a dysferlin C2A D16A mutation was generated and tested for activity. Relative to wild-type, this mutant displayed significantly attenuated activity regardless of the presence of calcium (Fig. 5d). Further, a aspartate mutation predicted to effect calcium binding for the myoferlin C2A domain was also tested (D71A). Like the dysferlin mutant, the myoferlin C2A mutant did not shift the laurdan spectra regardless of calcium (Fig. 5e).

The interaction between ferlin C2 domains and membranes is composed of both electrostatic and hydrophobic contributions

The C2 domains of many proteins bind membranes with a high degree of lipid specificity. For instance, the C2 domains of Golgi localized proteins typically bind to zwitterionic lipids, while proteins that target the plasma membrane harbor C2 domains that bind acidic lipids¹². To determine if otoferlin interacts preferentially with specific lipids, we repeated the laurdan measurements with SUV of varying anionic lipid compositions. When tested, the C2ABC region of otoferlin failed to alter the measured GP when mixed with SUV comprised of 0% POPS or 10% POPS (Fig. 6a). However, the construct altered the GP of liposomes composed of 25% POPS and 50%POPS. To ascertain whether sensitivity to acidic lipids is a general property of ferlin C2 domains, we conducted similar GP measurements on SUV composed of 0 or 25% POPS mixed with dysferlin or myoferlin C2ABC. For both proteins, and increase in the GP value was observed for SUV composed of 25% POPS but not 0%. Fig. 6b summarizes the results of these measurements. To explicitly test whether the interactions are electrostatic in origin, or due to highly specific interactions between the POPS headgroup and proteins, we carried out GP measurements on SUV composed of 25% POPG and 75% POPC. Like POPS, POPG is negatively charged, but possesses a phosphoglycerol headgroup instead of a phosphatidylserine headgroup. As shown in Fig. 6c, all three ferlin constructs increased the GP values when mixed with POPG containing SUV, supporting a generalized electrostatic interaction. Based upon these results, we conclude that the activities of the ferlin constructs require negatively charged lipids.

A previous report on the C2 domains of synaptotagmin I and VII determined that the mechanism of membrane binding for synaptotagmin I was predominantly electrostatic and ionic strength sensitive, while the membrane binding properties of synaptotagmin VII were insensitive to ionic strength and mediated by both electrostatic and hydrophobic contributions²⁷. To determine if electrostatics are the prevailing force driving interaction between SUV and the ferlin constructs, we measured the GP values of the C2ABC region of otoferlin, myoferlin, and dysferlin at various NaCl concentrations (Fig. 6d). Interestingly, the measured GP values did not change appreciably at the NaCl concentrations tested, suggesting additional non-electrostatic interactions contribute to the interaction between ferlin C2 domains and membranes. To ensure that the observed salt insensitivity was not due to an artifact of the system, we also measured otoferlin C2ABC induced changes of dansyl-PE anisotropy at 150 and 500 mM NaCl (Fig. 6e). When vesicles composed of 25 % POPS, 74 % POPC, and 1 % dansyl-PE were mixed with 5 μM otoferlin, an increase in anisotropy was observed at both 150 mM and 500 mM NaCl conditions. By contrast, 10 μM of synaptotagmin I C2B increased the dansyl-PE anisotropy at 150 mM but not 500 mM NaCl. This further suggests a difference between the mechanisms otoferlin and synaptotagmin use to interact with lipid membranes.