Very long-chain fatty acids support synaptic structure and function in the mammalian retina

Elongation of Very Long chain fatty acids-like 4 (ELOVL4) is a fatty acid elongase responsible for the biosynthesis of very long chain (VLC; ≥ C26) fatty acids in the retina, brain, skin, Meibomian gland, and testes. Heterozygous inheritance of mutant ELOVL4 causes juvenile macular degeneration in autosomal dominant Stargardt-like macular dystrophy (STGD3). Retinal photoreceptors are enriched with VLC polyunsaturated fatty acids (VLC-PUFAs), which have been shown by our group and others to be necessary for the survival of rod photoreceptors. Our group performed a series of studies using mice conditionally depleted of retinal Elovl4 (KO) aimed at understanding the role of VLC-PUFAs in long-term retinal health and function, focusing on the role of these fatty acids in mediating synaptic function between the photoreceptors and the rest of the neural retina. The absence of VLC-PUFA from the retina of KO mice resulted in a marked decrease in retinal b-wave responses of the electroretinogram as well as a decrease in the amplitude of the oscillatory potentials mediated by the neural retina. Although there were no measureable differences between KO and wild type (WT) mice in either pre-synaptic rod calcium channel function or post-synaptic bipolar cell glutamate receptor responses, ultrastructural analysis revealed a marked decrease in the diameter of synaptic vesicles in rod terminals. Recent quantification suggests that this decrease in synaptic vesicle size due to the absence of VLC-PUFAs in KO mice, and the consequent decrease in glutamate content, could account for the decrease in b-wave response amplitudes that were previously measured in these animals.

Autosomal dominant Stargardt-like macular dystrophy (STGD3) is a juvenile form of progressive macular degeneration that begins with onset of vision loss as early as nine years of age and is characterized by loss of the macula and subsequent formation of a central scotoma.STGD3 is caused by a five base pair deletion and frameshift mutation in exon six of the ELOVL4 gene (Donoso et al., 2001;Edwards et al., 2001;Griesinger et al., 2000;Kniazeva et al., 1999;Zhang et al., 2001).The frameshift mutation induces a pre-mature stop codon and causes a premature termination of the transcript, resulting in a truncated ELOVL4 protein devoid of its ERretention motif.Since the ELOVL4 protein must be retained in the ER to perform its enzymatic function (Agbaga et al., 2008;Barabas et al., 2013;Harkewicz et al., 2012;Logan et al., 2014), loss of the ER-retention motif causes the ELOVL4 protein to be mislocalized within the cytosol (Agbaga et al., 2014).The mutant protein does not have any enzymatic activity of its own (Logan et al., 2013).However, using in vitro cell-based and cell-free microsomal assays, we found that coexpression of different forms of both wild-type and mutant ELOVL4 resulted in a significant dominant-negative effect of the mutant protein on both localization and enzymatic activity of the wild-type protein (Logan et al., 2013).This suggests that the retina phenotype observed in STGD3 results from a loss of VLC-PUFA products due to the dominant negative effect of an enzymatically inactive mutant protein.
It is known that VLC-PUFA are incorporated into phosphatidylcholine and are densely packed into photoreceptor outer segment membranes, but their presence and function in retinal synapses was relatively unknown until recently.To determine the role of VLC-PUFA in the structure and function of retinal synapses, we conditionally deleted Elovl4 from rod and cone photoreceptors in mice and evaluated inner retinal function, synaptic architecture, and the ultrastructure of VLC-PUFA-depleted photoreceptor terminals (Bennett et al., 2014).

Retinal synapses contain VLC-PUFA
Ribbon and conventional synapses were prepared from fresh bovine retinas by sucrose gradient centrifugation D113, page 2 of 7 Fig. 2. VLC-PUFAs are enriched in retinal synapses.(A) Representative Western blot of fractionated bovine whole retina (WR) showed that rhodopsin (Rho, 37 kDa) was enriched in rod outer segments (OS), whereas VGlut1 (60 kDa) was localized to the ribbon synaptosome (RS) and synaptic vesicle protein (SV2; 95 kDa) was localized to the conventional synaptosome (CS) fractions.The WR sample was the whole retinal homogenate before centrifugation.(B) Ribbon and conventional synaptosome (RS and CS, respectively) fractions contained VLC-PUFAs that were not derived from OS contamination (n = 4).(C) Vesicles in WT rod terminals were predominately 20 to 39 nm in diameter, whereas the vesicles within the KO mouse terminals (D) were more frequently 20 to 29 nm in diameter.(E) Representative micrograph of a WT rod terminal shows an example of vesicles measuring between 30 and 39 nm (arrows).(F) Representative micrograph of a KO rod terminal shows an example of vesicles measuring between 20 to 19 nm (thick arrowheads) and 30 to 39 nm (arrows).Abnormal vesicles (empty arrowheads) appearing "deflated" were frequently observed in the KO mouse spherules.Mice were 12 months of age.Scale bars: 500 nm.Data are expressed as mean ± SD. (This research was originally published in Investigative Ophthalmology and Visual Science.Bennett et al., (2014).c Association for Research in Vision and Ophthalmology.) (Redburn and Thomas, 1979).Fractions were confirmed by protein analysis, revealing clean separation of rod outer segments (ROS) from ribbon synapses (RS) from conventional synapses (CS) (Bennett et al., 2014).VLC-PUFA were found in all three membrane preparations.Normalization to rhodopsin, an integral outer segment membrane protein, confirmed that the VLC-PUFA detected within the RS and CS fractions were intrinsic to those membranes and not a result of contamination from the ROS, which are known to be enriched in VLC-PUFA (Figs. 2A and 2B).Lipidomic analysis revealed significant differences in the PC molecular species distribution between the three fractions, indicating the presence of different phospholipid molecular species within each membrane fraction (Tab.1).

Loss of VLC-PUFA from retinal synapses results in a decrease in both synaptic vesicle diameter and number
Transmission electron microscopy (TEM) was used to evaluate ultrastructural changes within the retinal synapses of Elovl4 KO and control mice.A randomized blind study revealed that synaptic vesicle diameter was significantly reduced in KO mice at 12 months of age.WT mice had an average vesicle diameter of 29.5 ± 0.93 nm, whereas KO mice had an average vesicle diameter of 24.5 ± 0.62 nm, with the majority measuring less than 29.0 nm (Figs.2C and 2D).In addition, the number of tethered presynaptic vesicles (within 40 nm of the synaptic ribbon) was significantly reduced in the KO mice with an average of 3.6 ± 0.2 vesicles/µm of the presynaptic ribbon, while WT mice had an average of 4.7 ± 0.3 vesicles/µm of the presynaptic ribbon (Figs.2E and 2F).

Loss of VLC-PUFA results in synaptic reorganization
Immunohistochemistry performed on WT and KO retinas revealed a notable change in synaptic organization in Elovl4 KO mice (Figs.3A and 3B).The rod glutamatergic terminals (marked by VLGUT1 staining) make synaptic connection with bipolar cells (marked by PKC-α staining) within the outer plexiform layer (OPL).In the case of the KO mice, the rod photoreceptor terminals (green) appear to withdraw their terminals from the OPL where they should be making connections with the bipolar cell dendrites (orange).This results in a loss of VGLUT1 positive terminals in the OPL of KO mice and an increase in PKC-α staining within the outer nuclear layer (ONL) as the bipolar cell dendrites appear to extend downward in an attempt to re-connect with the withdrawn presynaptic terminals.Therefore it is likely that the decreased presynaptic vesicle size and number resulted in decreased synaptic efficiency and drove the subsequent reorganization of the rod terminals and the bipolar dendrites within the OPL.

Loss of VLC-PUFA results in rod-mediated functional deficits within the neural retina
Electroretinography (ERG) provides a non-invasive means to assess the electrophysiological responses of the photoreceptor outer segments (a-wave) and the neural retina (b-wave) in response to varying intensities of photostimulation.ERG was performed as described (Bennett et al., 2014) in order to analyze the various stages of the retinal response to light.These studies revealed that the outer segment-mediated a-wave was significantly reduced in the KO mice compared to WT mice.This loss of a-wave response could be explained by the loss of rod photoreceptor cells.In addition, the b-wave response induced by the photoreceptor pre-synaptic terminals and mediated by the neural retina was also significantly reduced in KO mice compared to WT (Figs. 4A and 4B).This reduction in b-wave amplitude, however, was greater than predicted from the loss or rod photoreceptor cells and represented specific changes in the retinal synapses in the KO mice.Using a Butterworth filter (30 and 80 Hz) to remove a-and b-wave contamination allowed for the isolation of oscillatory potentials (OPs), which are mediated by the synaptic feedback responses of amacrine, horizontal, and bipolar cells to the initial rod photoresponse (Wachtmeister, 1998).OP amplitudes were significantly decreased in KO mice compared to WT mice, suggesting a decrease in synaptic efficiency within the neural retina following depletion of VLC-PUFA (Figs. 4C and 4D).

Deficits in synaptic transmission due to the absence of vlc-pufa are not due to deficits in pre-synaptic calcium currents or post-synaptic glutamatergic currents
Whole-cell patch clamp recordings were used to evaluate the amplitude and voltage dependence of pre-synaptic rod photoreceptor inward calcium currents (I Ca ) as well as glutamatemediated post-synaptic currents in retinal slices from both WT and KO mice.Recordings were performed under standard conditions as described previously (Bennett et al., 2014;Van Hook and Thoreson, 2013).There was no significant difference in I Ca between WT and KO mice, indicating that loss of presynaptic I Ca are not responsible for any decreases in synaptic transmission in KO mice (Figs.5A-5C).Post-synaptic glutamatergic currents mediated by mGluR6 were evaluated by measuring responses to the mGluR6 antagonist CPPG applied in the presence of the metabotropic glutamate receptor group 3-selective agonist L-2-amino-4-phosphonobutyric acid (L-AP4).Rod bipolar cells were voltage clamped at −60 mV and responses were evoked by transient and localized CPPG application.Inward rod bipolar cell mediated currents were measured and no significant differences were detected between WT and KO mice, indicating that changes in the b-wave were not mediated by dysregulation of post-synaptic glutamate receptor currents (Figs.5D-5F).A lack of significant differences between WT and KO mice in both pre-synaptic I Ca and postsynaptic rod bipolar cell glutamate receptor currents indicates that the decrease in synaptic transmission is most likely due to deficits in pre-synaptic release downstream of I Ca , but upstream of post-synaptic metabotropic glutamate receptor responses.This is consistent with the possibility that the deficit in the b-wave arises from decreases in presynaptic release of glutamate, perhaps due to a small number of pre-synaptic vesicles or a decrease in vesicle size.

Decreases in B-wave neural retina responses correlate with decreases in pre-synaptic release of glutamate
The diameters of synaptic vesicles from KO mice were significantly smaller than those from WT mice, 24.5 nm vs. 29.5 nm (Figs.2C and 2D).Because volume scales with the cube of radius, this relatively small change in diameter would result in a reduction in volume (and thus glutamate content) of ∼57% (43% of control).This reduction in glutamate content could explain the roughly 50% decrease in b-wave amplitude in KO mice.

Discussion
Our studies demonstrate a clear role for ELOVL4 in supporting the structural and functional integrity of neuronal synapses within the mammalian retina.VLC-PUFA are not exclusively expressed in photoreceptor outer segments as was previously thought, but rather are present within both the large ribbon synapses made up of the photoreceptor and bipolar cells as well as the smaller conventional synapses utilized by the rest of the neural retina.The depletion of these very long chain fatty acids clearly provoked changes in the membrane structure of presynaptic photoreceptor terminals, causing them to withdraw into the ONL away from the bipolar cell dendrites, creating a larger distance for glutamate to diffuse before reaching its post-synaptic target.This retraction of pre-synaptic terminals from their post-synaptic targets, combined with a decrease in synaptic vesicle diameter and number, results in less glutamate being released and translates into a gross dysregulation of synaptic efficiency, which can be measured directly by de-creases in the electrophysiological responses of the scotopic system.This dysregulation cannot be accounted for by changes in inward pre-synaptic I Ca , which are necessary for synaptic vesicle docking and release (DeLorenzo and Freedman, 1978;Katz and Miledi, 1967) or by changes in post-synaptic glutamate receptor-mediated currents.This further supports the idea that the changes measured within the scotopic system are mediated by the decrease in pre-synaptic vesicle diameter and number, translating to a smaller pool of releasable vesicles as well as a decrease in the quantal size of individual vesicles.The reduction in vesicle diameter results in an average volume reduction of 57%, which correlates with a decrease in scotopic b-wave responses of ∼50% at higher stimulus intensities.It is important to note that these studies were conducted on 12-month-old mice and that this is an age-dependent phenotype where over time the reduction of these VLC-products results in retinal degeneration, synaptic remodeling, and dysregulation of synaptic function.
In an earlier publication from our group, (Brush et al., 2010), we found very long chain saturated fatty acids (sum of 26:0 + 28:0 + 30:0) in the neutral sphingolipids of the rat and bovine retina, as well as in bovine ROS.Since ELOVL4 is responsible for the synthesis of all very-long chain fatty acids (≥ C26) regardless of their degree of unsaturation, another possible scenario is that these very long chain saturated fatty acids, which exist as components of sphingolipid molecules, are providing a significant level of structural support for synaptic membrane size, while the very-long chain polyunsaturated fatty acids, which exist as components of phosphatidylcholine molecules, are providing a significant level of structural support for synaptic membrane morphology and fluidity.The longer the fatty acid chain, the more dynamic its influence can be on the structural and biophysical properties of a membrane.The biophysical properties of these two types of VLC-FA are very different, so it is possible that their ratios must be carefully balanced to achieve the proper size and curvature of a synaptic vesicle membrane.The loss of the VLC-PUFA could explain the significant changes in synaptic vesicle morphology, such as the lack of curvature, while the loss of the VLC-SFA could explain the significant reduction in synaptic vesicle size, as both ultrastructural changes were found in Elovl4 KO mice (Bennett et al., 2014).

Fig. 1 .
Fig. 1.Structure of VLC-PUFA.Free FA form of VLC-PUFA 32:5 n6 (A) and 34:5 n3 (B).Note the polyunsaturated methyl end and the saturated carboxyl terminal ends.(C) A typical phospholipid containing VLC-PUFA esterified to the sn-1position of the glycerol backbone.LC-PUFA, either 22:6 n3 or 20:4 n6, or others can occupy the sn-2 position.The sn-3 position in this scenario is occupied by phosphocholine (This research was originally published in Journal of Lipid Research.Agbaga et al., Retinal very long-chain PUFAs: new insights from studies on ELOVL4 (2010).c The American Society for Biochemistry and Molecular Biology).

Fig. 3 .
Fig. 3. Reduced VLC-PUFAs caused synaptic terminals to mislocalize in mouse retinas.(A) Glutamatergic vesicles labeled with VGlut1 (green) preferentially localized to the OPL in the WT mice but were found predominantly in the ONL of the KO mice.Bipolar dendrites labeled with PKCα (red) in the WT and KO mice.Scale bars: 20 µm.VLC-PUFA-deficient mice had a loss of photoreceptors and synaptic reorganization.(B) Vesicles labeled with VGlut1 (green) were localized with the bipolar cell dendrites labeled with PCK-α (orange) in the photoreceptor synaptic layer (OPL) in the WT, but were found in the ONL and OPL in the KO retina.The OPL in the KO retina was disorganized compared to WT with bipolar cell dendrites (orange) extending down into the ONL.Similar results were observed from eight different 9-month-old mice per genotype.Scale bars: 10 µm.(This research was originally published in Investigative Ophthalmology and Visual Science.Bennett et al. (2014).c Association for Research in Vision and Ophthalmology.)

Fig. 4 .
Fig. 4. Rod-mediated function deteriorated in VLC-PUFA-deficient mice.Representative 12-month-old WT (A) responses to increasing light intensities (−3.0 log scot cd•s/m 2 to 3.0 log scot cd•s/m 2 ) had higher a-and b-wave amplitudes than 9-month-old KO (B) responses.Vertical bar at upper right in each case is y-axis amplitude = 200 µV.(C) WT oscillatory potentials (OPs) had amplitudes higher than the KO (D) OP amplitudes.Vertical bar at upper right in each case is y-axis amplitude = 20 µV.The numbers between the traces in (A, C) are the light intensities (log scot cd•s/m 2 ) at which the responses were elicited.The latency of the OP responses was not different between the mice.(This research was originally published in Investigative Ophthalmology and Visual Science.Bennett et al. (2014).c Association for Research in Vision and Ophthalmology.)

Fig. 5 .
Fig. 5. Rod calcium currents (I Ca ) and rod bipolar cell (RBC) glutamate receptor signaling were similar between 12-month-old WT and KO mice.(A) Representative I Ca recorded with a voltage ramp (−90 to +60, 0.5 mV/ms) in a rod from a WT retina.The displayed trace is an average of two traces from a single cell.(B) I Ca from a KO rod.The displayed trace is an average of three traces from a single cell.(C) Left: group data, showing that the I Ca amplitude was similar in WT and KO rods.Likewise, the voltage dependence, as indicated by the peak voltage (V peak ) and voltage of half-maximal activation (V 50 ), was similar in rods from WT and KO mouse retinas.(D) Response of an RBC from a WT retina to a 1-second puff of CPPG (600 µM) in the presence of L-AP4 (4 µM).The RBC was voltage clamped at −60 mV, and the displayed trace is an average of five traces.(E) Response of an RBC from a KO retina.The displayed trace is an average of five traces.(F) Amplitudes of RBC responses to CPPG puffs were similar in WT and Elovl4 KO mouse retinas.Data are mean ± SEM. ns, not significant.(This research was originally published in Investigative Ophthalmology and Visual Science.Bennett et al. (2014).c Association for Research in Vision and Ophthalmology.)

Table 1 .
Phosphatidylcholine molecular species are different in retinal synaptosomes compared to photoreceptor outer segments.