Multi-parametric MRI of the physiology and optics of the in-vivo mouse lens

Abstract

The optics of the ocular lens are determined by its geometry (shape and volume) and its inherent gradient of refractive index (water to protein ratio), which are in turn maintained by unique cellular physiology known as the lens internal microcirculation system. Previously, magnetic resonance imaging (MRI) has been used on ex vivo organ cultured bovine lenses to show that pharmacological perturbations to this microcirculation system disrupt ionic and fluid homeostasis and overall lens optics. In this study, we have optimised in vivo MRI protocols for use on wild-type and transgenic mouse models so that the effects of genetically perturbing the lens microcirculation system on lens properties can be studied. In vivo MRI protocols and post-analysis methods for studying the mouse lens were optimised and used to measure the lens geometry, diffusion, T1 and T2, as well as the refractive index (n) calculated from T2, in wild-type mice and the genetically modified Cx50KI46 mouse. In this animal line, gap junctional coupling in the lens is increased by knocking in the gap junction protein Cx46 into the Cx50 locus. Relative to wild-type mice, Cx50KI46 mice showed significantly reduced lens size and radius of curvature, increased T1 and T2 values, and decreased n in the lens nucleus, which was consistent with the developmental and functional changes characterised previously in this lens model. These proof of principle experiments show that in vivo MRI can be applied to transgenic mouse models to gain mechanistic insights into the relationship between lens physiology and optics, and in the future suggest that longitudinal studies can be performed to determine how this relationship is altered by age in mouse models of cataract.

1. Introduction

The optical properties of the crystallin lens can significantly impact the quality of vision. Lens optics are determined by its geometry and gradient of refractive index (GRIN), which are established by the cellular structure of the lens [1]. Being an avascular tissue, the lens operates an internal microcirculation to uptake nutrients from surrounding aqueous and vitreous chambers and to export waste [2]. This microcirculation is generated by a collection of ion and water channels, transporters and gap junctions, which together establish the necessary electrochemical and hydrostatic pressure gradients to drive the transport of ions, fluids and nutrients through the lens at rates faster than would occur by passive diffusion alone [3]. Pharmacological perturbation of these underlying ionic and fluid fluxes has been shown to alter the transparent and refractive properties of the lens, which has led to the suggestion that the lens microcirculation operates to actively maintain the optical properties of the lens [4,5].

Our current understanding of the link between lens physiology and optics has been facilitated by the application of magnetic resonance imaging (MRI) to lens research to measure the tissue-specific MRI properties, the longitudinal relaxation time, T1, and transverse relaxation time, T2, in the lens. T1 is generally proportional to the free water content, which is an essential component regulated by the lens microcirculation system [6]. T2 quantifies the interactions between water and proteins (water-bound protein ratios), and this ratio, which spatially varies over the lens, forms the basis of GRIN [7]. Furthermore, diffusion tensor imaging (DTI) measures the diffusivity and directionality of water diffusion in the lens, providing quantitative information about the underlying directionality of water transport in different regions of the lens, which can be correlated with the local tissue architecture in the different lens regions [8]. Finally, MRI can be used to measure lens geometry (the surface curvatures, conic constants, and thicknesses) non-invasively, without concerns of optical distortions.

The utility of these MRI modalities has been demonstrated in several studies: T1 and T2 mapping of organ-cultured bovine and rat lens under pharmacological perturbations or treatments [4,9,10], DTI to study water diffusion patterns within the lens [11,12], and calibration of T2 with the lens GRIN in human ex vivo lenses [12,13]. However, the majority of these studies were limited to ex-vivo lenses, with in vivo studies of the lens being limited to a few studies in large animals [14] and humans [15,16], which can suffer from issues associated with motion artefacts, magnetic susceptibility artefacts in the anterior eye, and the high protein content of the lens. However, with advancements in the pre-clinical ultra-high field (7 T or above) MRI that delivers a high signal to noise ratio and excellent spatial resolution, in vivo imaging of ocular tissues in small animals has made significant progress [17–20].

In this study, we have taken advantage of these advances in technology in order to develop the necessary protocols to allow in vivo MRI to be applied to a number of previously characterised genetically-modified mouse strains [21–25]. We have optimised MRI protocols to extract lens geometry, DTI, T1 and T2 from the wild-type mouse lens in vivo, and to show proof-of-principle we have applied these methods to a transgenic knock-in mouse model (Cx50KI46), in which the gap junction protein Cx50 has been replaced by Cx46 [22]. These mice have increased fiber cell gap junctional coupling, and reduced intracellular hydrostatic pressure, both of which alter the lens microcirculation [23]. We additionally fit our T2 data to recently reported GRIN data of the mouse lens [26]. Our results show that in vivo MRI can be utilised to detect changes induced by genetic modulation of gap junction communication in lens geometry, water transport, water content, and water-bound protein ratios, which can be converted to GRIN. These techniques could be applied to the growing catalogue of transgenic mouse models generated to understand how lens structure and function contribute to the refractive and transparent properties of the lens, and how changes in those properties initiate the onset of lens cataract.

2. Material and methods

2.1. Animal preparation

Animal use was approved by the Stony Brook University IACUC. All mice used in this pilot study were male. Five C57BL/6 mice (Taconic Biosciences) were studied as the wild-type group with normal lenses at ten weeks of age (weight: 26 ± 2 g), and four Cx50KI46 mice on a C57BL/6 genetic background were studied at 11 weeks of age (weight: 27 ± 1 g). Cx50KI46 mice, which have had the Cx46 coding region knocked into the Cx50 gene locus to eliminate Cx50 expression and increase Cx46 expression, were generated locally as previously described [22].

Animals were initially anaesthetised with 5% isoflurane, followed by a bolus of xylazine (6 mg/k, IP) to reduce eye motion and then maintained with 1–1.5% isoflurane [27,28]. To minimise anterior eye motion, topical 1% tropicamide and 0.5% tetracaine were given. Lubricant eye ointment was used to prevent drying of the cornea. Animals were laid into a cradle with a circulating warm water pad to maintain body temperature. The head was fixed with ear and tooth bars. Mice were continuously given 1.0–1.5% isoflurane from a nose cone under spontaneous breathing. The respiration rate and rectal temperature (~37 °C) were monitored and maintained to ensure physiological homeostasis.

2.2. MRI imaging protocol

MRI was performed on a 7 T horizontal preclinical scanner (Biospec, Brucker, Billerica, MA) equipped with a 650mT/m BGA12S-HP gradient. A customised circular surface coil transceiver (internal diameter = 6 mm) was placed on the left eye of the mouse. The imaging plane was located on the central axis of the lens by visually positioning the slice to bisect the eye into superior and inferior halves.

One slice with a thickness of 0.5 mm was obtained for all imaging series. T1, T2 and DTI were acquired with the same base resolution with a field of view (FOV) = 6.4 × 6.4 mm and matrix size = 64 × 64. Balanced steady-state free precession (bSSFP) images with higher resolution (FOV = 6.4 × 6.4 mm, matrix size = 128 × 128) were acquired for accurate ocular geometry measurements. The scan times for imaging sequences were 8 min for T2 mapping, 24 min for T1 mapping, 15 min for DTI, and 5 min for bSSFP for a total of 50mins. The detailed imaging protocols are as follows:

Ocular Geometry was measured from bSSFP images with TE = 2.5 ms, TR = 5 ms and acquired with four RF phase cycling angles (0, 90, 180, and 270°) used to generate a banding-free image [18]. DTI utilised a standard spin-echo readout with diffusion gradients (gradient duration = 2.2 ms, gradient separation = 7.49 ms, b-value = 900 s/mm²). A b0 image without diffusion weighting and six directions of diffusion-weighted images were acquired (TE = 14.08 ms, TR = 1000 ms, 2 repetitions). T1 mapping utilised the rapid acquisition with refocused echoes (RARE) sequence with variable TRs (TR = 200, 380, 620, 950, 1500 and 4000 ms) and with TE = 2.62 ms. The number of repetitions for each TR were 14, 10, 7, 4, 3 and 1, corresponding to TR order. T2 mapping utilised a multi-echo spin-echo sequence with twelve TEs (minimum TE = 2.78 ms and echo spacing = 2.78 ms), with TR = 1600 ms and five repetitions.

2.3. Data analysis

Images were processed by custom-written routines in Matlab (Math-Works, Natick, MA). Repetitions of images were co-registered using the imregister function in the Matlab Image Processing Toolbox and then averaged. bSSFP images from the phase-cycled images were combined using nonlinear-averaging (Fig. 1A) to remove banding artefacts [18,29]. Canny edge detection was used to find the boundary of the cornea, lens, and vitreoretinal interface. The lens anterior and poster surfaces were individually fitted with an ellipse using constrained least-squares (Fig. 1B) [4,30]. The corneal and vitreoretinal edges were fit with a 4th order polynomial. The central axis was taken as the line through the vertices of the anterior and posterior lens ellipses. Along the central axis, the aqueous chamber depth (ACD), lens thickness (LT), and vitreous chamber depth (VCD) were measured. The radius of curvatures (R_A & R_P) and conic constants (Q_A & Q_P) of both lens surfaces, with A and P subscripts denoting anterior and posterior, respectively, were calculated from the elliptic geometry as [31], R_A, _P = a²/b and Q_{A, P} = (a/b)²−1, where a and b are major and minor axes of the ellipse.

Fig. 1.

Analysis of MRI images of the mouse lens. (A) bSSFP images were acquired at four different phase cycling angles, which shifts inherent banding artefacts to different locations. (B) bSSFP images were combined to correct for banding artefacts and were then processed to extract the boundaries of the cornea (blue), anterior lens (red), posterior lens (green) and vitreoretinal interface (yellow). The central axis (white) was taken as the line through the anterior and posterior lens surfaces. T1/T2 maps were analysed by two approaches – (C) Line profiles were taken from both equatorial and sagittal directions. (D) Based on the lens anatomy, two ROIs representing lens nucleus (N) and lens cortex (C) were generated. The nucleus mask occupied the inner 70% of the lens, while the remaining was considered the lens cortex.

DTI images were used to fit the diffusion tensor, and the mean diffusivity (MD) and fractional anisotropy (FA) were obtained from the eigenvalues of the tensor [8]. Pixel-wise T1 values were fit from the signal intensity, S, at an array of TRs as:

$$\text{S}\left(\text{TR}\right){\text{=S}}_{\text{0}}\left(1-e^{-\frac{TR}{T1}}\right)$$

where S₀ is the signal at infinite TR. Pixel-wise T2 values were fit from the signal intensity at an array of TEs as:

$$\text{S}\left(\text{TE}\right){\text{=S}}_{\text{0}}{e}^{-\frac{TE}{T2};}\text{S}\left(\text{TE}\right)>\sigma$$

where S₀ is the signal at TE = 0 ms, and σ is the noise threshold obtained from the image background [32] so that data at TEs below the threshold are discarded from the fitting to avoid biasing by noise [16].

T2 measurements were fit to refractive index (n) measurements from three-month old C57BL/6 mice (Charles River) from a recent study by Cheng et al. [26]. n was measured with X-ray Talbot interferometry and was fit to our T2 data (in s) to obtain a calibration:

$$n=\mathrm{1.38+2}{\text{.97×10}}^{\text{-4}}\left(\frac{\text{1}}{\text{T2}}\right)\text{+3}{\text{.14×10}}^{\text{-6}}{\left(\frac{\text{1}}{\text{T2}}\right)}^{\text{2}}$$

T1 and T2 maps were analysed by 1D trend analysis (Fig. 1C), in which values were extracted from the equatorial and sagittal axis (averaging a band of three pixels). The T1 and T2 profiles were fit with a power function over distance, as power functions have previously been used to model various parameters across the lens as [16,33,34]:

$$T1,T2=m+\delta {\chi }^p$$

where x is the distance from the lens centre, m is the offset, δ is the difference between maximum and minimum values and p is the exponent describing the rate of change across the lens. Only p, which characterizes the spatial variation transitions within the lens, was tabulated. Data from all mice were populated to perform the fitting.

Region of interest (ROI) analysis was performed using lens nucleus and cortex masks to evaluate the average T1 and T2 values within both regions (Fig. 1D). Based on the lens anatomy, the lens nucleus composed of mature fiber cells occupies the central 70% of the lens [35]. The lens nucleus mask was defined as an ellipse with the semi-axis lengths 70% those of the total lens, with the remaining outer regions defined as the cortex mask. For DTI data, the reliable signal could only be detected in the lens cortex due to the extremely short T2 of the lens nucleus. As such, Canny edge detection was applied to the diffusion-weighted images to differentiate the lens cortex ROI with adequate signal. FA and MD were only measured from within the lens cortex.

Group data were tabulated as mean ± standard deviation (SD). Statistical comparisons were performed by two-sided t-tests with p < 0.05 considered as significant.

3. Results

To study the physiological optics of the mouse lens, we employed and optimised four different in vivo MRI protocols to extract lens geometry (bSSFP), water diffusion and directionality (DTI), water content (T1 mapping), and water-bound protein ratio (T2 mapping). All of these are essential determinants of lens physiology and optics, and we have applied them to both wild-type (C57BL/6) and transgenic (Cx50KI46) mice.

3.1. Geometry

The bSSFP sequence provides a high signal to noise ratio with fast image acquisition, allowing for higher resolution in shorter scan times, from which lens and ocular geometry were measured (Fig. 1B). Representative geometrical analysis from a wild-type C57BL/6 and a Cx50KI46 mouse are shown in Fig. 2. The average lens geometrical properties and other ocular biometry obtained from bSSFP images are presented in Table 1. The Cx50KI46 mice had significantly reduced lens thickness (p < 0.05) and decreased radius of curvatures for anterior (p < 0.05) and posterior (p < 0.05) lens surfaces compared to wild-type C57BL/6 mice.

Fig. 2.

bSSFP images showing lens anatomy and geometrical analysis from a representative wild-type C57BL/6 (A) and Cx50KI46 mouse (B).

Table 1.

Lens and ocular biometry measurements for C57BL/6 and Cx50KI46 mice.

	C57BL/6 (n = 5)	Cx50KI46 (n = 4)
Lens geometric parameters
Lens thickness (LT) (mm)	1.90 (0.01)	1.63 (0.04)†
Anterior lens radius of curvature (RA) (mm)	1.30 (0.27)	0.91 (0.23)†
Posterior lens radius of curvature (RP) (mm)	1.28 (0.02)	0.75 (0.17)†
Anterior lens conic constant (QA)	0.60 (0.50)	0.52 (0.34)
Posterior lens conic constant (QP)	0.40 (0.10)	0.38 (0.44)
Ocular geometric parameters
Aqueous chamber depth (ACD) (mm)	0.35 (0.02)	0.31 (0.04)
Vitreous chamber depth (VCD) (mm)	0.60 (0.01)	0.62 (0.02)

Values expressed as mean (SD).

^†p < 0.05 compared to C57BL/6.

3.2. DTI

FA and MD maps of the lens cortex from a wild-type C57BL/6 and a Cx50KI46 mouse are presented in Fig. 3A–D, and representative plots of the diffusion-weighted and nondiffusion-weighted signal intensities from the lens nucleus and cortex ROIs are shown in Fig. 3E–F. All diffusion-weighted data in the cortex were well above the background noise threshold, while the signal intensities in the nucleus at all b-values were at the noise level. The mean MD of the lens cortex in C57BL/6 mice was 0.78 ± 0.02 μm²/ms, while the mean FA was 0.63 ± 0.01. No significant differences were observed for MD and FA of the lens cortex in Cx50KI46 mice (MD: 0.75 ± 0.03 μm²/ms, p = 0.54; FA:0.64 ± 0.01, p = 0.67) compared to C57BL/6.

Fig. 3.

DTI images showing representative MD maps from a wild-type C57BL/6 (A) and a Cx50KI46 mouse (B), and FA maps from a wild-type C57BL/6 (C) and a Cx50KI46 mouse (D). FA and MD maps were masked to show data only from the lens cortex ROI and overlaid on the non-diffusion-weighted (b0) image for reference. The signal intensities with b = 0 and 900 s/mm² (six diffusion directions) from an ROI in the lens nucleus and cortex are plotted, as well as the noise level, for a wild-type C57BL/6 mouse (E) and a Cx50KI46 mouse (F).

3.3. T1, T2 & GRIN

Fig. 4 shows images and trend profiles of T1, T2 and GRIN in Cx50KI46 mice compared to wild-type C57BL/6 mice. Representative T2-weighted signal intensities and curve fits from the lens nucleus, and cortex ROIs are plotted in Fig. 5A–B from a C57BL/6 and a Cx50KI46 mouse. Table 2 summarises the results from the trend and ROI analyses. The lens had the lowest T1 and T2 values in the nucleus, which gradually increased towards the cortex. The rate of change, exponent p, of both directions of T2 were substantially larger compared to T1. The lens nucleus T1 of Cx50KI46 mice was significantly higher than C57BL/6 wild-type controls (p < 0.05), while the lens cortex T1 of Cx50KI46 mice was similar to C57BL/6 (p = 0.44). The exponents of T1 trends in both directions were also significantly higher for Cx50KI46 mice (p < 0.05). The lens nucleus T2 of Cx50KI46 mice was significantly higher than C57BL/6 controls (p < 0.05), while the lens cortex T2 of Cx50KI46 mice was slightly higher than the C57BL/6, although the difference did not reach the significance level (p = 0.054). There were no significant differences in the exponents of T2 trends obtained in either anatomical direction. However, by converting T2 of the lens into GRIN (Fig. 4E–F), it became more apparent that the refractive index, n in the nucleus of Cx50KI46 lenses was significantly lower than C57BL/6 controls (p < 0.05).

Fig. 4.

Comparisons of lens T1 and T2 of wild-type C57BL/6 and Cx50KI46 mice. (A) T1 maps and (B) T1 equatorial trend profiles showed that Cx50KI46 mice had higher T1 values at the lens nucleus compared to wild-type C57BL/6 mice, and smaller lens size. (C) T2 maps and (D) T2 trend profiles for both groups of mice. (E) n maps and (F) profiles calculated from T2 showed that Cx50KI46 mice had reduced GRIN. Note that n is only valid in the lens as the refractive index only applies to the lens and that T1 and T2 of the vitreous and aqueous fluids could not be accurately fit as their values were too high relative to the longest TR and TE used herein. In all trend profiles, the boundaries between the lens nucleus (N) and cortex defined by the 70% of the lens thickness are shown by the dashed lines for wild-type C57BL/6 and Cx50KI46 mice.

Fig. 5.

Signal intensities with exponential curves for T2 calculations, as well as the noise level, in the lens nucleus and cortex from a wild-type C57BL/6 mouse (A) and a Cx50KI46 mouse (B). The echo signals that were below the noise level were discarded for fitting.

Table 2.

Mean T1, T2 and n values obtained from the lenses of C57BL/6 and Cx50KI46 mice using ROI and profile analysis.

Lens T1 and T2 parameters	C57BL/6 (n = 5)	CX50KI46 (n = 4)
ROI analysis
Lens nucleus T1 (ms)	847.9 (38.8)	964.35 (18.87)†
Lens cortex T1 (ms)	1427.51 (53.21)	1436.33 (63.22)
Lens nucleus T2 (ms)	8.06 (0.40)	10.40 (1.75)†
Lens cortex T2 (ms)	43.88 (2.26)	49.43 (5.11)
Lens nucleus n	1.47 (0.007)	1.43 (0.02)†
Lens cortex n	1.38 (0.001)	1.38 (0.001)
Trend analysis
T1 sagittal exponent	3.53 (0.52)	4.53(0.63)†
Tl equatorial exponent	3.03 (0.26)	5.78 (0.94)†
T2 sagittal exponent	7.06 (1.35)	7.50 (1.30)
T2 equatorial exponent	7.36 (0.35)	7.66 (1.74)

Values expressed as mean (SD).

^†p < 0.05 compared to C57BL/6 wild-type mice.

4. Discussion

In this study, we have shown that MRI can be applied to the mouse lens in vivo to study the effects of manipulating lens gap junction channels on lens geometry (Fig. 2, Table 1), water diffusion and directionality (Fig. 3), water content (Fig. 4), and the water to protein ratio and its effect on the GRIN (Fig. 4). Lens geometric measurements from bSSFP MRI were broadly in line with other histological and MRI studies (Table 3). Furthermore, relative to other in vivo imaging modalities such as optical coherence tomography [36], or optical low coherence interferometry [37], MRI has the unique ability to image the shape of the entire lens clearly without optical distortions. We found that Cx50KI46 mice had rounder lenses and reduced lens thickness, results consistent with previous reports of reduced lens size due to deficiency in epithelial cell proliferation in the postnatal period [38,39].

Table 3.

Comparison of measurements of the lens and ocular geometry obtained for C57BL/6 mice of the same age as in the current study with previously published studies.

Lens parameters	Study	Values	Method
LT	This study	1.90 (0.01)	In-vivo, MRI
	Schmucker & Schaeffela [43]	1.95	Ex-vivo, Microscopy
	Tkatchenko et al. [44]	1.94 (0.24)	In-vivo, MRI
Ra	This study	1.30 (0.27)	In-vivo, MRI
	Schmucker & Schaeffela [43]	1.1	Ex-vivo, Microscopy
Rp	This study	−1.28 (0.02)	In-vivo, MRI
	Schmucker & SchaeffeT [43]	−1.02	Ex-vivo, Microscopy
Other biometry
ACD	This study	0.35 (0.02)	In-vivo, MRI
	Schmucker & Schaeffela [43]	0.24	Ex-vivo, Microscopy
	Tkatchenko et al. [44]	0.38 (0.03)	In-vivo, MRI
VCD	This study	0.60 (0.01)	In-vivo, MRI
	Schmucker & Schaeffela [43]	0.70	Ex-vivo, Microscopy
	Tkatchenko et al. [44]	0.91 (0.36)	In-vivo, MRI

^aData were predicted from published age trends. Values were expressed as mean (SD).

Several previous ex-vivo investigations have suggested that diffusion in the outer cortex of the lens is highly anisotropic, and then transits to isotropic diffusion in the lens nucleus, due to regional differences of fiber cell structures and their cell to cell coupling [9,11,12]. The high FA values we observed in the cortex of the mouse lens agree with these previous studies, but in the current study, the diffusion parameters of the nucleus could not be measured, due to the inherently short T2 and the relatively long TE time required to apply the diffusion gradients. In theory, TE could have been shortened by using stronger imaging gradients or by reducing the b-value. However, achieving a sufficiently short TE to perform DTI on the nucleus of the mouse lens would be difficult, since decreasing the b-value to achieve adequately short TE would likely provide insufficient diffusion weighting. Interestingly, the FA values in the lens cortex were not significantly different between wild-type and Cx50KI46 lenses, indicating that although the two lenses had different overall geometries their underlying cell architecture and cell to cell diffusion pathways were similar in this region.

Measurements of T1 in wild-type and Cx50KI46 lenses showed that T1 values were lowest in the lens nucleus and gradually increased towards the outer cortex. T1 generally reflects the free water content in the lens, with lower water content in the lens nucleus. This observation is due to a combination of increased binding of water to the higher concentration of crystallin proteins found in the lens nucleus, and the active removal of water from the nucleus that is driven by the hydrostatic pressure gradient generated by the microcirculation system [40]. The magnitude of this pressure gradient is inversely related to gap junction conductance [23]. Hence, since increasing gap junction conductance by knocking Cx46 into the Cx50 gene locus decreases the hydrostatic pressure gradient [23], we would expect to see an increase in free water content in the lens nucleus of Cx50KI46 lenses. This was indeed the observation seen in Fig. 4B.

T2 measurement in the lens of some species can be difficult due to the short T2 in the lens nucleus. In the current study, T2 values for the lens nucleus were 8.0 ms, which is comparable to the values recorded of 7.7 ms at 1.5 T [14], and 2.1 ms at 9.4 T [10] for in vivo rabbits and ex vivo rats, respectively. In some studies, the echo spacing may be inadequate to quantify accurately the extremely short T2 times in the lens nucleus, while using echoes at long TE to fit the exponential curve could be biased by noise [41]. In our study, echo signals that fell below the noise threshold were discarded during the fitting routine [16]. This method ensures the lens nucleus T2 measurements are unbiased by the noise while preserving the fitting accuracy for the lens cortex, where longer TEs are needed for the fitting. In the lens nucleus, generally three or four echoes had signal above the noise threshold, which should be adequate for T2 fitting. However, it is possible that varying numbers of echoes fitted could also introduce some bias, which could be further investigated in future studies. In the lens cortex, with long T2, all echo signals were well above the noise threshold. Another consideration for lens T2 measurement is that T2 in fibrous tissue, such as tendon or sclera, is dependent on the orientation of the tissue fibers relative to the main magnetic field (B₀) direction [42]. B₀ in our studies was parallel to the imaging plane, so the radial profiles of the lens will vary over all orientations over 360° relative to B₀. As such, the T2 profiles of the lens may vary depending on orientation, but we did not observe any orientation-dependence on T2 maps. The orientation-dependence is also altered by loading of the tissue [42], so it is possible that at normal ocular pressures, the lens only has a weak dependence on B₀ orientation.

The resultant T2 gradient reflects the water to protein ratio in the lens, and T2 values are inversely related to the refractive index $\left(n\alpha \frac{1}{T_2^2}\right)$ and can be converted to absolute values of refractive index, n using calibration curves [4]. Here, we have established a calibration to convert MRI measured T2 values to n for the mouse lens. In the nucleus, n decreased significantly for the Cx50KI46 lenses, which would be predicted from the elevated free water content, detected in the Cx50KI46 lens using T1 mapping. The increased free water would alter the water to protein ratio in the lens nucleus detected by T2, and thus the GRIN in Cx50KI46 lenses. T2 was significantly higher in the nucleus of the Cx50KI46 lens, consistent with decreased n in this region by the genetically induced increase in gap junction coupling.

In summary, we have optimised in vivo MRI protocols to measure lens geometry, diffusivity, water content, and water to protein ratios in wild-type and genetically modified mouse lenses. These proof of principle experiments serve to demonstrate the feasibility of applying in vivo MRI to genetically engineered mouse models of lens cataract to gain mechanistic insight into the roles played by different lens proteins in the maintenance of the optical properties of the lens. The protocols developed here can be applied to additional mouse models of cataract to gain better mechanistic insight into how lens homeostasis fails with age. Many of these models have already been extensively characterised using invasive in vitro approaches, so the ability to apply non-invasive in vivo MRI imaging will allow for longitudinal studies to be conducted that can follow changes over time in lenses where a single gene has been manipulated. More importantly, such preclinical MRI studies on animals could then be used to guide clinical MRI studies on humans to determine how the properties of the lens change with age and the onset of presbyopia and cataract.

Abbreviations:

ACD

aqueous chamber depth

bSSFP

balanced steady state free precession

DTI

diffusion tensor imaging

FA

fractional anisotropy

FOV

field of view

GRIN

gradient of refractive index

LT

lens thickness

MD

mean diffusivity

MSE

multi-spin echo

RARE

rapid acquisition with refocused echoes

ROI

region of interest

SD

standard deviation

SE

spin echo

TE

echo time

TR

repetition time

VCD

vitreous chamber depth