The neurovascular retina in retinopathy of prematurity

Abstract

The continuing worldwide epidemic of retinopathy of prematurity (ROP), a leading cause of childhood visual impairment, strongly motivates further research into mechanisms of the disease. Although the hallmark of ROP is abnormal retinal vasculature, a growing body of evidence supports a critical role for the neural retina in the ROP disease process. The age of onset of ROP coincides with the rapid developmental increase in rod photoreceptor outer segment length and rhodopsin content of the retina with escalation of energy demands. Using a combination of non-invasive electroretinographic (ERG), psychophysical, and image analysis procedures, the neural retina and its vasculature have been studied in prematurely born human subjects, both with and without ROP, and in rats that model the key vascular and neural parameters found in human ROP subjects. These data are compared to comprehensive numeric summaries of the neural and vascular features in normally developing human and rat retina. In rats, biochemical, anatomical, and molecular biological investigations are paired with the non-invasive assessments. ROP, even if mild, primarily and persistently alters the structure and function of photoreceptors. Post-receptor neurons and retinal vasculature, which are intimately related, are also affected by ROP; conspicuous neurovascular abnormalities disappear, but subtle structural anomalies and functional deficits may persist years after clinical ROP resolves. The data from human subjects and rat models identify photoreceptor and post-receptor targets for interventions that promise improved outcomes for children at risk for ROP.

Fig. 1

Human rhodopsin growth curve and ROP onset. The smooth curve is normalized to the median adult rhodopsin content, 7.19 nmols/retina (Fulton et al., 1999a). Rhodopsin content reached 50% of the adult value at 5 weeks post-term (95% confidence interval: 0 to 10 weeks). The arrow indicates the age of ROP onset (technically pre-threshold ROP) at 32 weeks gestational age (Palmer et al., 1991). ROP onset coincides with a period of rapid developmental increase in rhodopsin content.

Fig. 2

Rod mediated ERG records and model fits of a-wave and b-wave responses in a 10 week old infant (left) and an adult (right). (A) Dark adapted ERG responses to a series of short-wavelength flashes. Flash intensity in log scot td s is indicated to the right of the traces; for clarity, only alternate flashes are labeled. (B) Solid lines re-plot the first 40 ms of the response to the seven brightest flashes; dashed lines represent the Lamb and Pugh model (Eq. 2) fit to the a-waves. The parameters S_ROD, R_ROD, and t_d for these fits are indicated. (C) b-wave amplitude plotted as a function of stimulus intensity. The smooth curve represents Eq. 4 fit to the b-wave data; responses to higher intensities were not included in the curve fit. The parameters log σ and V_MAX are indicated.

Fig. 3

Growth curves for rod photoreceptor model parameters, S_ROD and R_ROD, in human subjects. In each panel, the smooth curve is Eq. 3 fit to the data. The dashed lines indicate the 5^th and 95^th prediction limits of normal (Whitmore, 1986). Left panels (A, C): Development of S_ROD and R_ROD plotted as a function of age for 126 term born subjects with normal eyes (age 23 days to 52 years). Infants were recruited for testing at 4 or 10 weeks of age. Age₅₀ for S_ROD is 10.7 weeks (95% CI: 8.9 to 12.5) and for R_ROD is 11.2 weeks (95% CI: 9.4 to 13.0). Right panels (B, D): S_ROD and R_ROD for subjects with Severe ROP (N = 10), Mild ROP (N = 40), and No ROP (N = 18). Subsets of the data from both the term born (N = 71) and the preterm subjects (N = 21) were previously reported (Fulton and Hansen, 2000; Fulton et al., 2001).

Fig. 4

Deactivation of the rod photoresponse in term born subjects. (A) Rod isolation. A sample response in a 10 week old infant is shown. The amplitude of the a-wave response to a photopically matched red flash (dashed line) was subtracted from the a-wave response to the probe flash (solid line). The amplitude used in the analysis was measured 8 ms after the flash (vertical line). (B) Sample rod isolated a-wave responses from a 10 week old infant. The interval between probe and test flashes (ISI) is indicated for each response. (C) Deactivation in a 10 week old was summarized by t₅₀, the time to recover to half the amplitude of the a-wave to a single flash as determined by linear interpolation. R/R_MAX (see text) is plotted as a function of ISI. (D) The distribution of t₅₀ values for term born 10 week old infants (N = 15) and adults (N = 8). The horizontal lines represent the mean for each group. Re-plotted from Hansen and Fulton, 2005b.

Fig. 5

Growth curves for rod-driven post-receptor function, log σ and V_MAX, in human subjects. In each panel, the smooth curve is Eq. 3 fit to the data. The dashed lines indicate the 5^th and 95^th prediction limits of normal (Whitmore, 1986). Left panels (A, C): log σ and V_MAX plotted as a function of age for 211 subjects with normal eyes (age 23 days to 52 years). Age₅₀ for log σ is 11.0 weeks (95% CI: 9.3 to 12.7) and for V_MAX is 10.0 weeks (95% CI: 8.3 to 11.7). Right panels (B, D): log σ and V_MAX for subjects with Severe ROP (N = 10), Mild ROP (N = 40), and No ROP (N = 18). Fifty-eight (85%) of the ROP subjects have log σ below the normal mean for age and 30 (44%) have V_MAX below the normal mean for age. Subsets of the data from both the term born (N = 142) and the preterm subjects (N = 21) were previously reported (Fulton and Hansen, 2000; Fulton et al., 2001).

Fig. 6

Relationship between rod photoreceptor sensitivity, S_ROD, and post-receptor b-wave log σ. Deficits in log σ are plotted as a function of deficits in S_ROD. The solid diagonal line has slope of 1; dashed lines represent the range of values for each parameter found in healthy, mature control subjects. (A) Results from term born 10 week olds and adults. During normal development, deficits in rod sensitivity (S_ROD) predict deficits in post-receptor b-wave log σ. That is, the points fall near the diagonal line. (B) Results from ROP subjects. Many ROP subjects have deficits in log σ that are greater than predicted by deficits in S_ROd. Subsets of the data from both the term born (N = 71) and the preterm subject (N = 21) were previously reported (Fulton and Hansen, 2000; Fulton et al., 2001).

Fig. 7

Cone ERG. (A) Sample responses from a 10 week old infant to a 2.1 log unit range of red stimuli (λ > 610 nm) presented on a steady, white, rod saturating background (∼+3.0 log phot td). (B) Fit of the cone photoreceptor model to the infant’s a-waves. The parameters S_CONE and R_CONE are indicated. In these fits, t_d was 3.0 ms and τ was 1.8 ms. (C) S_CONE [left axis: (phot td_-1) s_-3; right axis: percent adult] for infants and older subjects. Data from subjects with Severe ROP, Mild ROP, and No ROP are compared to data from term born 10 week olds and adults. The means (±SEM) are plotted. Data in panel C re-plotted from Hansen and Fulton, 2005a and Fulton et al., 2008.

Fig. 8

Cone-driven b-wave responses as a function of stimulus intensity. (A) Results from term born 10 week old infants (N = 28) and adults (N = 13). A photopic hill, which peaks at approximately +2.3 log phot td s, is apparent in the adults but not in the infants. (B) Results from ROP infants (Severe, Mild, None) tested at corrected age 10 weeks compared to results from term born infants. (C) Results from former pre-terms tested at older ages compared to results from adult controls. Re-plotted from Hansen and Fulton, 2005a and Fulton et al., 2008.

Fig. 9

Oscillatory potentials. (A) OP energy in term born infants, ROP infants, and older ROP subjects plotted as percent of the normal adult mean (±SEM). Energy is the area under the Gaussian fit to the power spectrum and is related to the square of the summed OP amplitude (Akula et al., 2007b). (B) OP energy in older ROP subjects and mature normal controls stratified by spherical equivalent. Among the ROP subjects, high myopia ranged from -8.40 D through -14.00 D (N = 4); myopia ranged from -0.37 D through -5.75 D (N = 11); and emmetropia ranged from plano through +2.44 D. Data from 14 myopic and five emmetropic controls are shown; there were no controls with high myopia. A subset of the OP data was reported by Akula et al., 2007b.

Fig. 10

Multifocal ERG responses in term born infants. (A) The traces represent the response from each of the 61 hexagons in the unscaled array from an infant. (B) The mean (±SEM) amplitude of the positive component of the wave, P1, in term born infants (N = 18) and adults (N = 6) is plotted as a function of ring number. Data in panel B re-plotted from Hansen et al., 2009.

Fig. 11

Multifocal ERG responses in subjects with a history of Mild ROP. (A) The traces represent the response from each of the 103 hexagons in the scaled array from a 13 year old subject. (B) The mean (±SEM) amplitude of the positive component of the wave, P1, is plotted as a function of ring number for Mild ROP subjects (N = 11) and healthy controls (N = 9). Data in panel B re-plotted from Fulton et al., 2005.

Fig. 12

Optical coherence tomography (OCT) images. (A) The retinal lamina are as indicated: ILM, inner limiting membrane; NFL, nerve fiber layer; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; ELM, external limiting membrane; CC, connecting cilium; RPE, retinal pigment epithelium; C, choroid; IS, inner segment layer; OS, outer segment layer; t, temporal; n, nasal. (B) Cross-sectional images from one eye of five control subjects (left) and five subjects with a history of Mild ROP (right). In ROP subjects, the fovea was shallower and broader than in controls. Adapted from Hammer et al., 2008; copyright Association for Research in Vision and Ophthalmology.

Fig. 13

Retinal vasculature in OCT images in a control subject (top) and a subject with Mild ROP (bottom). En face views showing vessels in the outer plexiform layer (A, C) were created by averaging the axial slices shown between the horizontal lines in the corresponding cross-sectional images (B, D). Adapted from Hammer et al., 2008; copyright Association for Research in Vision and Ophthalmology.

Fig. 14

Fundus photographs of children (4 to 8 years old) representing each of the ROP categories, Severe (A), Mild (B), None (C) and a term born control (D). The caliber of the retinal arterioles (arrows) was similar in the Mild, None, and term born control eyes. The Severe ROP eye had attenuated arterioles. Also, consistent with the OCT results (Ecsedy et al., 2007; Hammer et al., 2008), the foveal dimple appeared blunted in eyes with ROP, suggesting a shallower foveal pit.

Fig. 15

Dark adapted scotopic thresholds in young infants. (A) Dark adapted thresholds (scot td s) of young infants plotted as a function of age. Results from several studies are summarized (Hansen and Fulton, 1981; Powers et al., 1981; Hamer and Schneck, 1984; Brown, 1986). The average adult threshold is represented by the solid horizontal line; the dashed lines show ±SD. (B) Spectral sensitivity functions of dark adapted 10 week old infants and adults. Thresholds were measured at eight wavelengths in each subject. Mean (±SEM) threshold is plotted as a function of wavelength for infants and adults. The smooth waves are V_λ' (Wyszecki and Stiles, 1982). Re-plotted from Hansen and Fulton, 1993.

Fig. 16

Threshold development at parafoveal and peripheral sites in term born infants. Each infant (N = 9) was tested at both sites at age 10, 18, and 26 weeks. The longitudinal data from each individual are connected by line segments. (A) Threshold at 10° eccentricity as a function of age. (B) Threshold at 30° eccentricity as a function of age. (C) Difference between parafoveal and peripheral thresholds (Δ_10-30). The mean adult value (triangle) and its range (dotted lines) are shown in each panel. Re-plotted from Hansen and Fulton, 1999.

Fig. 17

Threshold development in subjects with Mild ROP (A) and No ROP (B). Thresholds were measured longitudinally at parafoveal (10° eccentric) and peripheral (30° eccentric) sites. Results for individual infants are connected by line segments. Each infant was tested at both sites at every visit. The difference between parafoveal and peripheral thresholds (Δ_10-30) is plotted in the lower panels. In every panel, the dashed lines represent the 99% prediction interval for normal threshold development derived from the results in Fig. 16. Adapted from Barnaby et al., 2007; copyright Association for Research in Vision and Ophthalmology.

Fig. 18

Background adaptation. Threshold is plotted as a function of background intensity on log-log axes. (A) The smooth curve plots the equation log T = log T_DA + log [(A_o + I)/ A_o] (Eq. 6). The model parameters, dark adapted threshold (T_DA) and the eigengrau (A_O), are indicated by the arrows. The horizontal dashed line asymptotes the curve at T_DA; the oblique dashed line intersects the horizontal line at A_O. Values of T_DA and A_O were determined for each subject at the parafoveal (10° eccentric) and peripheral (30° eccentric) sites. Increment threshold functions from (B) a 10 week old term born infant, (C) a 13 year old Mild ROP subject, and (D) a 14 year old control are shown. The values of T_DA (log scot td s) and A_O (log scot td) at the peripheral and parafoveal sites are shown in each panel. Panels B, C, and D re-plotted from Hansen and Fulton, 2000a, .

Fig. 19

Scotopic spatial summation and interaction. (A) Dark adapted spatial summation in term born 10 week old infants and adult controls. Threshold in scot td s is plotted as a function of stimulus area. Points represent the mean ±SEM. The oblique and horizontal lines intersect at the critical areas for complete summation. The average critical area for summation in infants and adults are indicated by the arrows. The corresponding critical diameters are 4.42° and 2.32°, respectively. (B) Average Westheimer functions for infants and adults. Infants were tested using 2°, 20 ms stimuli presented on backgrounds of 2.7 to 6.1° diameter. Adults were tested with 10′, 10 ms stimuli on backgrounds of 0.3 to 3.1° diameter. The background diameters which produced the maximum threshold elevation are about four times larger in infants (3 to 3.5°) than in adults (0.64 to 0.75°). Panel B re-plotted from Hansen and Fulton, 1994.

Fig. 20

Mean (+SEM) arteriolar integrated curvature, IC_A (left), and rod photoreceptor sensitivity, S_ROD (right), in infants with a history of ROP and in rat models, plotted as percent of normal for age (dotted line). In both the human and animal subjects, mean IC_A is nearly two times higher in ROP and rod sensitivity (S_ROD) is reduced by ∼25%. Data re-plotted from Fulton et al., 2001; Gelman et al., 2005; Moskowitz et al., 2005a; Fulton and Hansen, 2006; and Akula et al., 2007a.

Fig. 21

Development of rat rod photoreceptor outer segments. (A) Dashed curve: Developmental elongation of rod outer segments (ROS) follows a logistic growth function (Fulton et al., 1995). The ROS are 50% of the adult length at P12.5 (95% CI: P12.1–P12.8). Rhodopsin concentration (weight of rhodopsin per dry weight of retina) follows a similar course (Timmers et al., 1999), reaching half the adult value at P13.4 (95% CI: P12.1–P14.7). Solid curve: The amount of rhodopsin extracted from the whole retina is approximately 50% of that in adults at age P18.7 (95% CI: P18.2–P19.2 days). (B) Rhodopsin absorbances in rat ROS determined by microspectrophotometry (Dodge et al., 1996). Mean (±SEM) of absorbances in dark adapted ROS are shown at four ages: 13, 19, and 34 days and adult. In 19 day old rats, there is a gradient of absorbance along ROS length. The oldest, least mature disks at the tip have absorbance similar to that found in the 13-day-old ROS. The newest, most mature disks at the base have absorbance similar to that found in adult ROS. In the adult ROS, absorbance is similar along the entire length of the ROS. Adapted from Dodge et al., 1996; copyright Association for Research in Vision and Ophthalmology.

Fig. 22

Rat control sample records. Sample ERG records of a control rat tested at P17 and P31. ERG responses to a range of brief (<1ms), full-field strobe flashes were recorded. The stimuli were controlled in intensity by calibrated neutral density filters and were increased in 0.3 log unit steps. The unattenuated flash (5.2 log μW/cm²) produced approximately 10^4.8 isomerizations per rod per flash. The numbers to the left of each row of traces indicate the stimulus intensity (log R*). For clarity, only every third record is shown. The start of each trace coincides with stimulus onset. Adapted from Liu et al, 2006a; copyright Association for Research in Vision and Ophthalmology.

Fig. 23

Deactivation of the rat rod photoresponse. (A) The amplitude of the a-wave response to the probe flash, expressed as proportion of the amplitude of the response to the test flash alone, plotted as a function of ISI. The data are from the infant and the adult who had the median t₅₀ for the 4.0 log μW/cm² flash. (B) Mean t₅₀ plotted as a function of stimulus strength, expressed as estimated proportion of rhodopsin molecules isomerized/rod/flash. The infants' and adults' functions do not differ significantly. Data re-plotted from Fulton and Hansen, 2003.

Fig. 24

Induction of retinopathy. Two models of ROP were created by exposing rat pups to different ambient oxygen environments during the ages when the retina was immature and rhodopsin content increasing. ERG responses and fundus photographs were obtained at P20, P30, and P60 (gray bars). Top: The 75 model: oxygen was maintained at 75% from P7 to P14. Middle: The 50/10 model: starting on the day of birth, ambient oxygen concentrations were alternated every 24 hours between 50% and 10% for 14 days. Bottom: Room air is 21% oxygen. Adapted from Akula et al., 2007a; copyright Association for Research in Vision and Ophthalmology.

Fig. 25

Sample records from P20 rats and a-wave model fits. (A) Responses from 50/10 model, 75 model, and control rats. Flash intensity in μW/cm² is shown to the left of the traces. Arrows indicate stimulus delivery. (B) Fits (black lines) of a model of the activation of phototransduction (Eq. 2) to the a-waves (gray lines with dots). Fitting was restricted to the leading edge of the a-wave to a maximum of 10 ms after the flash. S_ROD and R_ROD are indicated in each panel. Adapted from Akula et al., 2007a; copyright Association for Research in Vision and Ophthalmology.

Fig. 26

Integrated curvature of retinal arterioles (IC_A) and rod photoreceptor sensitivity (S_ROD) as proportion of the control mean in 50/10 model, 75 model, and control rats. S_ROD at P20 is plotted on the abscissa and IC_A at P60 on the ordinate. The dotted horizontal and vertical lines indicating the control means intersect at IC_A = 1.0 and S = 1.0. The solid line is a linear regression through the data. Rod sensitivity at P20 predicted vascular abnormality at P60. Re-plotted from Akula et al., 2007a.

Fig. 27

Mean (±SEM) rates of change in post-receptor b-wave sensitivity (log σ) and integrated curvature of the retinal arterioles (IC_A) for 50/10 model, 75 model, and control rats. Reprinted from Akula et al., 2007a; copyright Association for Research in Vision and Ophthalmology.