Dietary Omega-3 Fatty Acids and Ganglion Cell Function

Abstract

purpose. Diet-induced deficiencies in Omega-3 (ω-3) fatty acids are well known to alter photoreceptor function. In this study, the broader functional changes in a diversity of retinal neurons were considered.

methods. Sprague-Dawley dams were fed either ω-3-sufficient (ω-3⁺, n = 21) or -deficient (ω-3⁻, n = 19) diets 5 weeks before conception, with the pups continued on the mothers’ diet. After 20 weeks of age, electroretinograms (ERGs) were recorded by using protocols that isolate separate cellular generators, including; photoreceptors (PIII), ON-bipolar cells (PII), and ganglion/amacrine cells (STR). At the brightest energies, rod and cone responses were isolated with a paired-flash paradigm. Retinal tissue (ω-3⁺, n = 5; ω-3⁻, n = 5) was harvested at 23 weeks of age for fatty acid assays with thin layer and gas liquid chromatography.

results. Omega-3 deficiency caused a 48.6% decrease in total retinal docosahexaenoic acid (DHA). This change induced significant amplitude decreases only in the rod PII (−8.2%) and positive (p)STR components (−27.4%), with widespread delays in all signals (PIII 5.7%, PII 13.6%, pSTR 7.6%, and negative [n]STR 8.3%). Omega-3 deficiency exerted its greatest effects on signals originating in the inner retina (pSTR).

conclusions. Increasing dietary ω-3 has beneficial effects across the retina, with the greatest improvement occurring in ganglion cell function.

The ω-3 and ω-6 families of polyunsaturated fatty acids (PUFAs) are essential, as they are needed to sustain health and they must be ingested. ¹ The balance between these two families is central to normal cellular function and can protect against disease. ^{2 3 4} Despite this fact, the modern Western diet has grown relatively deficient in ω-3 PUFA because of the disproportionately high intake of ω-6-rich oils (e.g., many common vegetable oils, margarines, and meats), ⁵ such an imbalance may predispose to poor health in general and eye disease specifically. 

The retina is a particularly useful tissue in which to consider this deficiency. Retinal rod outer segments contain high levels of ω-3 PUFA, with the outer segments having the highest docosahexaenoic acid (DHA) content of any cell in the body and an unusually efficient DHA retention mechanism. ⁶ Moreover, this fatty acid has a functional benefit, as photoreceptors rich in ω-3 PUFA show improved performance in animal models ^{7 8 9 10 11 12 13 14} and humans. ^{15 16} In animals fed ω-3 PUFA, the electroretinogram (ERG) a-wave, which reflects photoreceptor function, is consistently faster, ^{7 8 9 10 11} with some studies also reporting better amplitudes. ^{9 10 11}  

A beneficial effect of ω-3 PUFA has also been shown for postreceptoral neurons. More specifically, a diet rich in ω-3 results in consistently faster ON-bipolar cell–dominated ERG b-wave responses, ^{7 8 17 18} and some studies also report an improvement in amplitude. ¹⁴ However, the effect of dietary ω-3 PUFA on ganglion cell functional integrity has yet to be investigated. 

Altered ω-3 PUFA may affect middle and inner retinal function in several ways. First, as the retina processes information in a serial manner, photoreceptoral deficits will lead to downstream postreceptoral dysfunction. It has long been known and more recently quantified ^{19 20} that PIII changes (or a-wave, photoreceptor response) can alter the PII (or b-wave, ON-bipolar cell dominated response). Given that ganglion cells are third-order neurons, it is likely that these cells also show downstream effects. 

In addition to downstream effects from the outer retina, it is also possible that ω-3 PUFA may directly affect the inner retinal neurons. Omega-3 PUFA may alter the function of membrane-bound proteins found on bipolar and ganglion cells, thereby influencing their activity, as has been demonstrated for photoreceptors. ^{12 13} Ganglion cells may be particularly susceptible to ω-3 PUFA deprivation, as they have high ω-3 uptake as evidenced by in vivo injections of radio-labeled DHA (³H-22:6). ^{21 22} In addition, ganglion cell axons distal to the lamina cribrosa become sheathed in myelin, which is composed of 70% lipid. ²³ Given this background, we posit that proximal retinal signals are susceptible to ω-3 modification. 

Methods

All experimental procedures conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Animals

Sprague-Dawley rats (Rattus norvegicus) were housed as described previously. ²⁴ Before experimentation, anesthesia was achieved with intramuscular injection of a cocktail of ketamine-xylazine, 60:5 mg/kg (Ketamil 100 mg/mL, Xylazil 100 mg/mL; Troy Laboratories, Smithfield, NSW, Australia). One drop of proxymetacaine hydrochloride (Ophthetic 5 mg/mL; Allergan, Frenchs Forest, NSW, Australia) and 0.5% tropicamide (Mydriacyl 5 mg/mL; Alcon Laboratories, Frenchs Forest, NSW, Australia) provided corneal anesthesia and mydriasis, respectively. Body temperature was maintained at 37 ± 0.5°C by a water heating pad. 

Dietary Manipulations

The diets were based on a rodent chow (AIN 93G) to which several sources of lipid were added. The ω-3-sufficient (ω-3⁺) diet was balanced in ω-3/ω-6 and comprised flaxseed, safflower, and tuna oils (5.5%:1.5%:0.5%), whereas the ω-3-deficient (ω-3⁻) diet contained safflower oil only (7%). The fatty acid, mineral, and vitamin composition have been detailed. ²⁴ Dams were placed on relevant diets from 5 weeks before impregnation, to ensure maximum tissue fatty acid change. ²⁵ Pups were weaned at postnatal day 21 and maintained on maternal diets to the end of the experiment. Although this dietary protocol yields an extreme ω-3 depletion, ²⁴ it has been adopted as a “proof of principle” in terms of the ω-3 effect. Experimental groups were assigned by a pseudorandom process that balanced for birth weight and sex. ²⁴ ERG procedures were performed on the entire cohort (ω-3⁺ n = 21; ω-3⁻ n = 19), whereas a subset was used for retinal tissue fatty acid analysis (ω-3⁺ n = 5; ω-3⁻ n = 5). 

Retinal Fatty Acid Analysis

Eyes were excised from deeply anesthetized animals at 23 weeks of age. Retinas were carefully dissected and removed and phospholipid extraction was performed by thin-layer and gas-liquid chromatography, as previously described by our group. ^{24 25}  

Electroretinography

Electroretinography was conducted at three ages (20, 21, and 22 weeks; two-way repeated-measures [RM]-ANOVA), with 1 week between each assay, to increase the experimental power. In all cases, the parameters were not significantly different across weeks; thus, the data from the three assays were averaged. Electrode configuration and flash characteristics have been described previously. ²⁶ To increase the range of luminous energy (∼8 log units from −5.95 to 1.52 log cd · s · m⁻² = 1 × 10⁻³ to 2.74 × 10⁴ photoisomerizations/rod), ²⁷ the voltage, current, and duration of the flash were altered. The flash duration was within the integrating time of the cellular generator elicited (determined by unpublished pilot data, Dang TM, 2007) and, with the exception of very dim flashes, was 1 ms. Note that all references to luminous energy indicate scotopic cd · s · m⁻². 

Rod responses were isolated for analysis as this pathway is better defined in terms of the ERG. As the mixed a-wave is rod dominated in rodents at our maximum flash exposure, ²⁸ cone extraction was not performed. Pilot data showed that the cone a-wave contributed 4.3% of the mixed a-wave amplitude at an exposure of 1.52 log cd · s · m⁻², which is well within the noise of our recording system, justifying our approach. However, at bright luminous energies, the PII has a large cone contribution, and the rod response has to be extracted with a paired flash paradigm. ²⁷ As the STR is collected at exposures below cone threshold, it represents the proximal cell dominated summation of rod inputs in rats, and no cone extraction is needed. ^{29 30 31}  

ERG Data Analysis

Implicit times for 80% of maximum amplitude were adopted to prevent intrusion of post-receptoral responses known to influence the later part of the a-wave. ³² For consistency, this same criterion was used to establish all ERG timing parameters at the single brightest luminous energy eliciting their response. 

Photoreceptoral Response (PIII).

The a-wave was modeled as an ensemble over two exposures (1.22 and 1.52 log cd · s · m⁻²) using a delayed Gaussian. ³³ This model is dependent on three parameters, a maximum amplitude (Rm_PIII, in microvolts), a sensitivity parameter, (S, m² · cd⁻¹ · s⁻³), and a delay term (t _d, in milliseconds). Because dietary manipulation has been shown to alter the time of activation of many stages in the phototransduction cascade, ^{12 13} t _d was floated. 

However, due to the interdependence of S and t _d, ²⁷ bias toward either parameter was minimized by implementing a bootstrap procedure ²⁶ (n = 500, Solver module of an Excel spreadsheet; Microsoft, Redmond, WA) with random parameter jitter (33%) for each diet group. This method returned a robust estimate of the average PIII parameters for each diet group, which was then used to seed the individual PIII fits required for determination of the PII (described later). In addition, the implicit time to 80% a-wave maximum was taken as an independent and unbiased measure of timing, as it encompasses both t _d and S changes. 

Bipolar Cell Response (PII).

As the PII response has a large cone contribution the rod response was isolated by using a paired flash protocol. ^{27 34} Briefly, putative cone waveforms were obtained from a probe flash of 1.52 log cd · s · m⁻² preceded by a saturating flash of 1.88 log cd · s · m⁻² with a 500-ms interstimulus interval. As only a single response was measured, the cone response was analyzed in terms of its maximum amplitude and an implicit time to 80% of maximum amplitude. The rod-isolated waveform was extracted by subtracting the cone response from a mixed response to the same flash energy (1.52 log cd · s · m⁻²). Individual rod PII components were derived from this extraction by digitally subtracting the modeled PIII response and low-pass filtering (−3 dB at 46.9 Hz, Blackman window). 

The intensity-response function of the rod PII peak amplitude was modeled by using a Naka-Rushton function fit to the linear data and returned three parameters: maximum amplitude (V _max, in microvolts), a semisaturation constant (k, log cd · s · m⁻²), and a slope parameter (n). ³⁵ The intensity response series of the mixed b-wave (peak-to-peak) is biphasic (Fig. 1 , filled circles) reflecting the contribution of multiple mechanisms. First, the interaction between the negative PIII and the positive PII will create a plateau (Fig. 1 , arrow) at moderate intensities. ³⁶ Digitally subtracting the PIII minimizes this confounding factor ^{36 37} (Fig. 1 , open circles) and exposes the PII generator. However, even after this extraction, a remnant intrusion was still apparent at brighter energies (>−1.81 log cd · s · m⁻²). We believe that this arises from the emerging cone PII ^{38 39} and have excluded these luminous energies from the rod analysis. The maximum rod response was defined by the upper 95% limit of the paired-flash derived rod isolated PII amplitude (Fig. 1 , open square). This approach yields outcomes consistent with the intensity response observed in genetically modified rod-only mice. ³⁸  

Proximal Retinal Response.

To increase the signal-to-noise ratio of the small amplitude scotopic threshold response (STR), we averaged multiple waveforms (20 repeats, 2-second interstimulus interval) at each flash energy and then subjected to a low-pass filter post hoc (46.9 Hz, −3 dB, Blackman window). The three dimmest luminous energies (−5.95 to −5.26 log cd · s · m⁻²) were averaged to determine the peak (pSTR) and trough (nSTR) amplitudes. 

Statistics

Group data have been expressed as the mean (±SE). The Grubbs’ test was used to trim up to one outlier from any data set. Data normality was tested by the Kolmogorov-Smirnov test (Prism, ver. 4; GraphPad, San Diego, CA) and homogeneity was established with a variance ratio (Bartlett’s test) that was evaluated by F-test. Student’s t-tests were used to compare ERG parameters across diet groups (unpaired). A one-way RM-ANOVA was used to compare parameters (PIII, PII, pSTR, nSTR) across the within-diet groups. 

Retinal fatty acid analysis had an α of 0.05, whereas a more conservative α of 0.035 was adopted to protect against type 2 errors of multiple ERG parameter comparisons and in cases of nonhomogenous variances. ⁴⁰  

Results

Retinal Tissue Fatty Acid Analysis

The dietary manipulation achieved a fourfold increase in the ratio of ω-6 to ω-3 present in the retina from 0.41 (ω-3⁺) to 1.81 (ω-3⁻). Specific changes to key metabolites are shown in Figure 2 , where it is apparent that the ω-3⁻ diet resulted in a 48.6% decrease in retinal DHA (ω-3⁺, 34.1% ± 0.8%; ω-3⁻, 17.5% ± 0.5%; t ₈ = 17.96, P < 0.001). Other significant (t ₈ = 6.44, P < 0.001) retinal fatty acid changes for the ω-3⁻ diet group were an increase in 22:4ω-6 (74.6%) and 22:5ω-6 (3817%, docosapentaenoic acid, DPA) and a decrease in 22:5ω-3 (65.7%). Fatty acid compounds not shown in Figure 2were detected in trace amounts. 

Diet Effect on Retinal Function

Figure 3shows the average rod-isolated ERG waveforms for each diet group over a wide range of luminous energies (∼8 log units). Note that each panel has different amplitude scales and the top and bottom panels are shown on different time scales to emphasize key components. At the dimmest light levels (Fig. 3 , bottom), the STR consists of a small positive component (peak ≈120 ms, reference line) followed by a negative component (trough ≈220 ms, reference line). With increasing intensity (>−4.91 log scot cd · s · m⁻²), the ERG becomes dominated by the corneal positive b-wave (or PII). This continues to grow until intrusion from the corneal negative a-wave becomes apparent (≈−2.35 log cd · s · m⁻²). At the brightest energy, the a-wave has a trough at approximately 10 ms and the b-wave peaks at ∼70 ms (dotted reference lines). The oscillatory potentials (OPs) usually seen on the leading edge of the b-wave are reduced in these waveforms as the averaging process acts as a low-pass filter. The ω-3⁺ waveforms (Fig. 3 , thin trace) are faster and larger than the ω-3⁻ waveforms (thick trace) across all luminous energies. 

Omega-3’s Effect on the Receptoral Response (PIII).

Figure 4Ashows the group average a-wave to emphasize the photoreceptoral loss with dietary manipulation. The corresponding PIII fits, ³³ and implicit times are shown. Note that the PIII model is fit to an ensemble across two luminous energies, but for clarity, only the brightest is shown. 

The a-wave implicit times were delayed in the ω-3⁻ group (Fig. 4B ; ω-3⁺, 8.61 ± 0.07 ms; ω-3⁻, 9.10 ± 0.14 ms; t ₃₈ = 3.24, P < 0.01) which is consistent with the significant reduction in sensitivity (Fig. 4D ; S, ω-3⁺, 3.58 ± 0.01 m² · cd⁻¹ · s⁻³; ω-3⁻, 3.52 ± 0.02 log m² · cd⁻¹ · s⁻³; t ₃₈ = 3.01, P < 0.01) and an increased latency (Fig. 4E ; t _d; ω-3⁺, 4.58 ± 0.03 ms; ω-3⁻, 4.82 ± 0.07 ms; t ₃₈ = 2.98, P < 0.01). There is a nonsignificant trend toward lower PIII amplitude (Fig. 4C ; Rm_PIII; ω-3⁺, −584 ± 20 μV; ω-3⁻, −536 ± 21 μV; t ₃₈ = 1.63, P = 0.11). 

Omega-3’s Effect on the Postreceptoral Response (PII).

The rod PII response was quantified by the Naka-Rushton function (Fig. 5A ; ω-3⁺, filled circles, thick curve; ω-3⁻, open circles, thin curve). The rod-derived PII maximum amplitude was reduced (Fig. 5C ; V _max: ω-3⁺, 1320 ± 40 μV; ω-3⁻, 1210 ± 40 μV), but just fails to reach significance (t ₃₈ = 1.99, P = 0.05) based on our conservative α value. However, when a one-sided t-test is conducted, for reasons outlined in the Discussion, it gives P = 0.027. There is no significant difference in the semisaturation constant (Fig. 5D ; k: ω-3⁺, −3.33 ± 0.01 log cd · s · m⁻²; ω-3⁻, −3.32 ± 0.02 log cd · s · m⁻²; t ₃₈ = 0.23, P = 0.82) but a significantly steeper slope with ω-3 deficiency (Fig. 5E ; n: ω-3⁺, 0.89 ± 0.01; ω-3⁻, 0.93 ± 0.01; t ₃₈ = 3.74, P < 0.01). The 80% criterion implicit time shows a significant delay in the ω-3⁻ group (Fig. 5B ; ω-3⁺, 47.6 ± 0.7 ms; ω-3⁻, 54.0 ± 1.1 ms; t ₃₈ = 5.51, P < 0.001). 

With ω-3 deficiency, the cone PII did not exhibit a significant change in amplitude (Fig. 6B ; ω-3⁺, 347.1 ± 14.1 ms; ω-3⁻, 319.5 ± 10.0 ms; t ₃₈ = 1.57, P = 0.13) nor timing (Fig. 6C ; ω-3⁺, 52.6 ± 0.5 ms; ω-3⁻, 54.9 ± 1.0 ms; t ₃₈ = 2.09, P = 0.04). 

Omega-3’s Effect on the STR.

The STR response at dim light levels is represented by the gray areas in Figure 7 . Figure 7Ahas been expressed on a log–log scale to improve visualization of the small STR amplitudes (see Fig. 5Afor semilog axes). The average of these parameters (Fig. 7C)shows that the ω-3⁻ group had a significantly smaller pSTR than did ω-3⁺ group (ω-3⁺, 13.1 ± 0.7 μV; ω-3⁻, 9.5 ± 0.8 μV; t ₃₈ = 3.51, P < 0.01). However, there is no significant difference in nSTR amplitude (Figs. 7B 7D ; ω-3⁺, −16.6 ± 0.6 μV; ω-3⁻, −15.5 ± 0.4 μV; t ₃₈ = 1.57, P = 0.12). The ω-3⁻ group also experienced significant delays in both pSTR (Fig. 7E ; ω-3⁺, 109 ± 1 ms; ω-3⁻, 117 ± 2 ms; t ₃₈ = 3.16, P < 0.001) and nSTR implicit times (Fig. 7F ; ω-3⁺, 195 ± 2 ms; ω-3⁻, 211 ± 4 ms; t ₃₈ = 3.16, P < 0.001). 

Comparing Outer and Inner Retinal Changes

To determine whether ω-3 deficiency preferentially affects generators of rod-derived ERG components we expressed our changes normalized to the average control values (ω-3⁺) in Figure 8 . Variability for the control group is expressed as the 96.5% confidence limits for the mean as shown by gray bars in the figure. Treated group means falling outside this area are statistically significant (P < 0.035). The ω-3⁻ group exhibited nonsignificant decreases in PIII amplitude (−7.5% ± 3.6%) and nSTR amplitude (−7.2% ± 2.7%; P > 0.035 for both). In contrast the ω-3⁻ group showed a significant −8.2% ± 2.9% decrease in PII amplitude and a −27.4% ± 5.7% decrease in pSTR amplitude (P < 0.035 for both). Moreover, the diet-induced percentage loss was significantly greater in the pSTR than the PIII, PII, or nSTR (F _3,18 = 13.7, P < 0.001). 

The implicit times of all components were significantly delayed (P < 0.035) in the ω-3⁻ group: PIII 5.7% ± 1.6%, PII 13.6% ± 2.3%, pSTR 7.6% ± 1.6%, and nSTR 8.3% ± 2.1% (Fig. 8B) . Furthermore, the diet induced delay in timing is significantly greater in the PII than the PIII, pSTR, or nSTR (F _3,18 = 5.32, P < 0.001). 

Discussion

Retinal Fatty Acids

Animals fed the ω-3⁻ diet had a 48.6% reduction in retinal DHA compared with the ω-3⁺ group, as well as a reduction in 22:5ω-3, an increase in 22:4ω-6, and a dramatic 39.2-fold increase in 22:5ω-6 (DPA, P < 0.05). These findings are in agreement with the literature. ^{8 9 10 11} Indeed, the substitution of DHA for DPA achieved with an ω-3⁺ diet has been shown to improve phototransduction efficiency and magnitude. ¹² Given these fatty acid changes, we would anticipate alterations to retinal function. 

It should be acknowledged that our dietary manipulation was extreme in terms of human consumption and that our approach was adopted to achieve a high level of tissue fat modulation. Elsewhere, we have shown that this is also the case in the ciliary body, ²⁴ and herein we report substantial retinal changes. However, the relevance of this diet to humans is not clear, as the effect of less severe ω-3 deprivations, typical in humans, over long durations and even over a lifetime, is not well known. Nonetheless, large-scale human clinical trials imply that even modest dietary modifications of ω-3 intake by humans can lead to substantial health benefits. ^{41 42}  

Electroretinography

Omega-3’s Effect on the Receptoral Response (PIII).

ω-3 Dietary deprivation caused delays in the photoreceptor response. This delay manifested as a decrease in sensitivity and an increase in t _d, resulting in an increase in implicit time. These findings are in agreement with the literature, which consistently reports a timing delay in photoreceptor responses. ^{7 8 9 10 11} In contrast, the Rm_PIII trend in the raw amplitude was not statistically significant (P = 0.11). Indeed, the literature reports various effects on PIII amplitude, with some studies showing no change ^{7 8} and others a decrease in the ω-3⁻ group. ^{9 10 11} In vitro studies suggest that both timing and amplitude changes should occur with ω-3 dietary deprivation as the kinetics and the concentration of metarhodopsin II-G_t formation, a key element in the activation of the transduction cascade, is reduced. ¹² In addition Na⁺/K⁺ ATPase dysfunction has been reported ^{43 44 45} with ω-3 deficiency, which would alter the state of cellular excitability and reduce the Rm_PIII. 

That we failed to find an Rm_PIII deficit may reflect variability in our experimental groups, as the effect size of −7.5% is modest at best. To detect this difference given the variability of our data (SD = 92.2 μV), a sample size of 47 would have been required, whereas with our current sample size (ω-3⁺ n = 21, ω-3⁻ n = 19), the power of this experiment is only 0.37. Given the sensitivity of in vitro experiments, the decreased Rm_PIII trend found with ω-3 deficiency is likely to be real. 

Omega-3’s Effect on the Postreceptoral Response (PII).

In addition to the expected photoreceptoral effects, we also find changes to the rod PII response which is thought to be a measure of ON-bipolar cell activity. ⁴⁶ The ω-3⁻ group exhibits a delay in timing of the PII response (P < 0.001) consistent with the literature. ^{7 8 17 18} A two-sided t-test did not show a significant reduction in the PII amplitude (P = 0.05) consistent with studies in monkeys ^{7 11 17} and rats. ^{8 47} However, a significant decrease in PII amplitude was reported in ω-3⁻ guinea pigs. ¹⁴ As the probability (P = 0.05), using our conservative approach of a two-tailed t-test is very close to our criterion of α = 0.035, we cannot be confident that this finding is real. Indeed, the power of this experiment is 0.56, which is less than desirable (power = 0.8). Given the a priori expectation from the literature that ω-3 deficiency should either decrease or leave the PII amplitude unaltered, a one-sided t-test can be adopted to increase power. Note that a similar approach does not salvage the Rm_PIII. This analysis returns P = 0.027; thus, it is likely that the PII amplitude decrease found with ω-3 deficiency is also real. Therefore, for both the PIII and PII finding the ω-3⁻-induced loss is frustrated by the small magnitude of change coupled with the modest experimental variability limiting experimental power. It is also worth noting that when normalized (Fig. 8A) , the PII loss was significantly removed from control variability (96.5% CI), supporting our contention. 

The cone PII was not significantly affected with ω-3 deficiency. This finding is in contrast to the loss in amplitude found using a 30-Hz flicker stimulus to elicit cone responses in guinea pigs. ¹⁴ This difference may arise as the flicker stimulus used in past work elicits a complex interaction between the ON- and OFF-bipolar cells, ⁴⁸ whereas our paired-flash methodology favors ON-bipolar cell isolation. ²⁷ Nevertheless, it is of interest that the cone PII trend with ω-3 deficiency followed a tendency similar to that of the rod PII. More specifically, the trend for a decrease (−8.0% ± 2.9%) in cone PII amplitude was consistent with that of the rod PII loss (P = 0.92). However, the trend in cone PII delay (4.3% ± 1.9%) was less than that of the rod PII (13.6% ± 2.3% P < 0.001). Indeed, a preferential effect of ω-3 on rods over cones is consistent with studies in infants. ^{15 16} The cause of this preference is unknown and requires further investigation. 

Omega-3’s Effect on the STR.

We report for the first time a dietary ω-3 PUFA effect on the STR which comprises two components: the pSTR and nSTR. In rodents, the pSTR has major ganglion cell contributions, and the nSTR reflects both amacrine and ganglion cell function. ^{30 31 49} The pSTR amplitude was significantly decreased in the ω-3⁻ group (Figs. 7A 7C ; P < 0.01) and the nSTR was unaffected (Figs. 7B 7D , P = 0.13). That the nSTR was not affected indicates a ganglion cell dysfunction rather than a general amacrine/ganglion cell inner retinal deficit. This finding also argues against a Müller cell dysfunction, as this would decrease both pSTR and nSTR. ^{50 51} Both the pSTR and nSTR experienced significant delays in their implicit times (Figs. 7E 7F , P < 0.001). 

Although some studies have indicated that ω-3 PUFA may have neuroprotective effects on ganglion cells ⁵² these benefits have only been found after exposure to stress. Ours is the first study to find that ω-3 deprivation alters ganglion cell function in the absence of experimentally induced insult. 

Comparing Outer and Inner Retinal Changes

The reported dietary ω-3 PUFA-induced outer and middle retina dysfunction are typically very subtle, ^{7 8 14 17 18} and consequently it is of interest to determine whether the inner retina is more severely affected. As ERG components arising from different retinal layers vary in amplitude over at least 2 log units, ERG changes are expressed relative to the control data (Fig. 8) . This approach allows all ERG components to be expressed and compared by using a common metric. We found a larger pSTR loss (−27.4% ± 5.7%; Fig. 8A ) than for the other retinal components (7.2%–8.2%; F ₃ = 11.6, P < 0.001). 

A larger pSTR than PII loss is consistent with the steeper Naka-Rushton slope (n) found in the ω-3⁻ group (Fig. 5E , P < 0.001). In mice, it has been shown that PII energy–response functions can be approximated better with a slope of 1, after inhibition of inner retinal responses using NMDA and GABA, which acts to remove the STR. ⁴⁹ Thus, the steeper slope is internally consistent, with ganglion cells being more affected than bipolar cells. 

Although both the pSTR and nSTR are thought to have ganglion cell contributions their differing effects may reflect contributions from neurons other than ganglion cell to the nSTR in rodents. ^{29 30} Moreover, chronic models of glaucoma have shown a preferential loss of pSTR over nSTR, consistent with the greater ganglion cell contribution to this waveform. ⁵³ An alternative explanation for the absence of nSTR dysfunction, is less opposition from the corneal positive waveforms, such as the PII and pSTR onto the nSTR response, ^{31 49} thus masking the true nSTR loss. 

It is also of interest to determine whether the timing of the inner retina is more delayed than the outer, as this may reflect subtle changes in mechanisms such as protein activation. ¹² The PII implicit time was relatively more delayed than other retinal components (Fig. 8B ; F ₃ = 13.7, P < 0.001) indicating a preferential effect of ω-3 deficiency on the efficacy of ON-bipolar cell activity. Nevertheless, when the retinal system was viewed as a whole, the ganglion cells were still the most severely affected, as their maximum capability to function was decreased, and furthermore, the change (−27.4%) was greater than the bipolar cell timing delay (13.6%; F ₇ = 5.97, P < 0.035). 

The ganglion cells were preferentially affected by ω-3 deficiency, and two mechanisms may explain this effect. First, the greater pSTR deficit may indicate a direct effect of ω-3 PUFA deficiency on ganglion cells. Systemic injections of radio-labeled DHA (³H-22:6) results in heavy labeling of photoreceptors, and moderate labeling of the nerve fiber layer, both of which were greater than that in the inner nuclear layer. ^{21 22} The relatively high uptake in these two regions may underlie the pattern of dysfunction seen in this study. Large amounts of DHA may be needed by ganglion cells to support synthesis of long axons or myelin, thus increasing susceptibility to ω-3 deficiency. Unlike other retinal cells, ganglion cell axons are myelinated distal to the lamina cribrosa. Myelin increases the speed of neurotransmission and is composed of 70% lipid. ²³ As cell culture studies show that ω-3 PUFA can stimulate myelin production ⁵⁴ and dietary ω-3 deficiency can alter myelin fatty acid composition, ⁵⁵ it is possible that a lack of ω-3 PUFA influences the myelin of ganglion cells, leading to a preferential ganglion cell dysfunction. 

Second, the greater ganglion cell dysfunction may simply reflect outer retinal changes, which are amplified by nonlinear processing in the inner retina. It has been demonstrated that linearity exists between PIII and PII amplitudes. ^{19 20} However, the same cannot be assumed for the relationship between the PII and the STR, as at the inner retina, there is additional lateral communication and inhibitory feedback. ^{56 57} For example, a 50% reduction in PIII leads to a 50% reduction in PII, but it is not clear what effect this has on the STR. Indeed, the data of Saszik et al. ⁴⁹ show that the PII and STR have different sensitivities to background light, suggesting a nonlinear gain at the inner retina. This finding is supported by the nonlinear processing found at the ganglion cell level by white-noise analysis. ^{58 59} The exact relationship between the PII and the STR requires further investigation. 

What is important is that, regardless of whether the ω-3 effect on ganglion cells occurs via a direct effect and/or through amplification of outer retinal losses due to nonlinear processing, this class of cells is preferentially affected by dietary manipulation of ω-3 PUFA. 

Summary of ω-3’s Effect on Retinal Function

Omega-3 PUFA improves the function of photoreceptors and bipolar cells, consistent with the literature. We report for the first time an improvement in ganglion cell function that is greater than for responses arising from the outer and middle retina. Although the mechanism through which this occurs requires further investigation, these data suggest that ω-3 PUFA may have a role to play in diseases that affect ganglion cells. 

 Supported by an Elizabeth and Vernon Puzey Scholarship (CTON), a C. J. Martin Postdoctoral Fellowship (BVB), and National Health and Medical Research Council (NHMRC) Project Grants 400127 (BVB) and 350224 (AJV).

 Submitted for publication January 14, 2008; revised March 20 and April 21, 2008; accepted June 19, 2008.

 Disclosure: C.T.O. Nguyen, None; A.J. Vingrys, None; B.V. Bui, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Intensity response series of the peak-to-peak amplitude (filled circles, arrow indicating plateau) and the PII amplitude after PIII extraction (open circles). At the brightest luminous energy a paired-flash isolates the rod PII_max (open square). The upper 95% CL identifies waveforms with cone intrusion (>−1.81 log cd · s · m⁻²). PII data are modeled with a Naka-Rushton function, with V _max the saturated amplitude; k, 1/sensitivity, and n the slope of the function.

Figure 2.

 

Retinal fatty acid profiles after 20 weeks of dietary manipulation (ω-3⁺, n = 5 □; ω-3⁻, n = 5 ▪) expressed as a percentage of the total phospholipid quotient. Where fatty acid species are not shown, these were found in trace amounts that failed to return values above baseline noise. Data are expressed as the mean ± SEM. *P < 0.05.

Figure 3.

 

Averaged group ERG waveforms showing the effect of ω-3 fatty acids. Numbers to the left of the waveform indicate the luminous energy. The range of luminous energy isolates separate components, in order from dimmest to brightest flashes: scotopic threshold response, b-wave, and rod-isolated a/b-wave combination. Note that the low intensities are on an expanded time scale. Reference lines (dotted lines) are shown for low intensities at the approximate peak (120 ms) and trough (220 ms) times of the STR, and for the bright intensity at the approximate a-wave trough (10 ms) and b-wave peak (70 ms). Also note that averaging the waveforms across animals leads to cancellation of OPs. The ω-3⁺ animals (thin line, n = 21) have waveforms that are faster and larger than those of the ω-3⁻ animals (thick line, n = 19) at all energies.

Figure 4.

 

(A) The effect of dietary ω-3 fatty acids on the photoreceptoral response. Averaged a-wave (ω-3⁺, ○, n = 21; ω-3⁻, •, n = 19) with corresponding PIII model (solid lines). Dashed vertical lines: implicit time at an 80% amplitude criterion. Note that a single waveform for each diet is shown for clarity, although the PIII model is optimized over an ensemble of two waveforms. (B) The a-wave implicit time is delayed in the ω-3⁻ group (average ± SEM, P < 0.01) PIII model parameters returned by modeling the response to individual animals (average ± SEM, ω-3⁺, open; ω-3⁻, filled). (C) There is no significant change in PIII amplitude (Rm_PIII, P = 0.11). (D) A significant loss in sensitivity (S, P < 0.01); (E) an increase in delay in the ω-3⁻ group (t _d, P < 0.01). *Significant differences.

Figure 5.

 

The effect of dietary ω-3 fatty acids on the extracted rod PII response. (A) The intensity response for the rod PII for both diet groups (average ± SEM; ω-3⁺, ○; ω-3⁻, •) with their corresponding Naka-Rushton functions (ω-3⁺, thin line; ω-3⁻, thick line) whose parameters are shown in (C) and (D). (B) The average (± SEM) rod PII implicit time is delayed in the ω-3⁻ group (P < 0.001; ω-3⁺, □; ω-3⁻, ▪). (C) V _max (maximum bipolar cell function) shows a significant decrease in the ω-3⁻ group with a one-sided t-test (P < 0.035). (D) There is no significant difference in the k (1/sensitivity) of the bipolar cells (P = 0.82). (E) The slope of the Naka-Rushton (n) is significantly larger in the ω-3⁻ group (P < 0.001). *Significant differences.

Figure 6.

 

The effect of dietary ω-3 fatty acids on the cone PII response. (A) Averaged cone waveforms (ω-3⁺, thin line, n = 21; ω-3⁻, thick line, n = 19), (B) cone PII amplitudes (P = 0.13), and (C) 80% criterion implicit times (P = 0.04) were not significantly altered with diet.

Figure 7.

 

The effect of dietary ω-3 fatty acids on the STR amplitude. (A) The average (±SEM) intensity response shown on a log–log scale to highlight the pSTR response (shaded area) for both diet groups (ω-3⁺, ○; ω-3⁻, •). Naka-Rushton functions were optimized using slope (n) fixed to one (ω-3⁺, thin line; ω-3⁻, thick line) to illustrate the more prominent pSTR presence in the ω-3⁺ group evidenced by the greater departure from the rod PII model. (B) The intensity response series to the trough amplitude with the nSTR-dominated response highlighted (shaded area) for both diet groups on a semilog axis. (C) pSTR was reduced in the ω-3⁻ group (P < 0.01; ω-3⁺, □; ω-3⁻, ▪). (D) There was no significant difference in the nSTR response amplitude between diets (P = 0.12). (E) pSTR timing was delayed in the ω-3⁻ group (P < 0.001). (F) nSTR timing was delayed in the ω-3⁻ group (P < 0.001). *Significant differences.

Figure 8.

 

Differential dietary effect on outer to inner retinal ERG components. (A) In terms of amplitude when expressed relative to the average ω-3⁺ value (± 96.5% CI, shaded area), the ω-3⁻ group (average ± SEM, •) exhibited decreases in PIII (−7.5%), PII (−8.2%), pSTR and nSTR (−27.4% and −7.2%, respectively), although only the PII and pSTR losses are significant. *P < 0.035. Diet-induced percentage loss is greater in the pSTR than the PIII, PII, or nSTR. #P < 0.001. (B) In terms of timing the ω-3⁻ group (average ± SEM, •) exhibits significant delays (*P < 0.035) in PIII (5.7%), PII (13.6%), pSTR, and nSTR (7.6% and 8.3%, respectively). Diet-induced percentage delay in timing is greater in the PII than in the PIII, pSTR, or nSTR. #P < 0.001.

The authors thank Professor Andrew Sinclair for providing technical advice regarding fatty acid analysis.