Evaluation of a PEGylated Fibroblast Growth Factor 21 Variant Using Novel Preclinical Magnetic Resonance Imaging and Magnetic Resonance Elastography in a Mouse Model of Nonalcoholic Steatohepatitis

Animal Model

Two studies were conducted under the Institutional Animal Care and Use Committee approved Animal Care and Use Protocol procedures at Mayo Clinic (Rochester, MN) and Bristol Myers Squibb (BMS, Lawrenceville, NJ). The investigative procedures regarding diet and treatment groups were conducted in the same manner at both sites, as follows: 6-week-old, male C57BL/6J mice were purchased from Jackson Laboratories (Bar Harbor, ME) and acclimated to either the BMS (single-housed, N = 36) or Mayo Clinic animal facilities (group-housed, 3–4 per cage, N = 42) for 1 week on standard rodent chow (BMS: Teklad, 2018, Madison, WI; Mayo Clinic: Lab Diet 5053 PicoLab® Rodent Diet-20; St. Louis, MO). At 7 weeks of age, mice of normal body weight (~22–24 g) were randomized to receive either a matched control diet (CD) containing standard levels of choline, methionine, and fat (Research Diets, A13012807, New Brunswick, NJ) or a choline-deficient, amino acid-defined, (methionine, 0.1%), high-fat (60% of calories from fat) diet (CDAHFD; Research Diets, A06071302, New Brunswick, NJ).^{10-12, 33}

Study Design

Two studies were performed using a 7T animal MRI scanner (Bruker/Biospin, Billerica, MA) at BMS with 12 mice per treatment group and using a 3T human MR scanner (Signa HDxt, GE Healthcare, Milwaukee, WI) at Mayo Clinic with 14 mice per treatment group. In both studies, the CD (1/3 of the study mice) and CDAHFD (2/3 of the study mice) groups were maintained for 12 weeks. Half of the CDAHFD mice received subcutaneous injections of 0.6 mg/kg PEG-FGF21v BIW beginning 4 weeks after diet initiation, for a total duration of 8 weeks (CDAHFD + PEG-FGF21v). CD mice and the other half of the CDAHFD mice received subcutaneous injections of vehicle BIW for 8 weeks (CD + Vehicle or CDAHFD + Vehicle, respectively; 7T site) or nothing (3T site). PEG-FGF21v was diluted to 0.12 mg/mL with sterile-filtered vehicle, aliquoted, and stored at 4 °C. A fresh aliquot was used for each dose and was maintained in a wet ice bath until the time of injection. Vehicle was 250 mM sucrose and 20 mM Tris at pH 8.3.

MRI-HFF and MRE imaging were performed before introducing either the CD or CDAHFD diets (BD), after 4 weeks on diet and prior to dosing (BL), after 4 weeks of dosing (WK4), and after 8 weeks of dosing (WK8). Mice were weighed on the days that imaging was performed. Mice were sacrificed for necropsy after the WK8 imaging.

At the 7T site, in addition to the longitudinal study groups, three satellite terminal groups (N = 6 mice/group), each imaged and sacrificed at BD, BL, and WK4, were included to provide intermediate pathophysiologic data associated with disease progression with CDAHFD.

MRI and MRE

All mice were fasted for at least 2.5 hours before imaging. Mice were anesthetized and maintained with inhaled 0%–2% isoflurane delivered by 100% oxygen during the imaging. Respiration rate was monitored and isoflurane adjusted to maintain 30–80 breaths per minute. Warm water circulators (7T site) or heated rice bags (3T site) were used to maintain normal mouse body temperature. The mouse respiration rate and body temperature were monitored (SA Instruments Inc., Stony Brook, NY) during MRI/MRE scans at the 7T site.

3T Imaging

MRI/MRE was performed in a supine position on a 3T/60-cm whole-body MRI scanner (Signa HDxt, GE Healthcare, Milwaukee, WI). A custom-made 4-cm inner diameter, 8-channel birdcage receive-only imaging coil was used. The silver needle was inserted vertically through the anterior body wall into the ventral abdomen approximately 2 mm caudal to the xiphoid process and approximately 2 mm right of the ventral midline. The needle handle was then mounted on the pneumatic passive driver (Fig. S3 in the Supplemental Material) to produce shear waves in the liver. Pressure waves generated at 200 Hz by the active pneumatic driver outside the scanner room were delivered to the passive needle driver via a long plastic tube. Wave images were acquired using a flow-compensated, phase-contrast, multislice SE-EPI MRE sequence, with repetition time/echo time (TR/TE) = 400/43.5 msec, four phase offsets, eight slices of 2-mm thickness, and three orthogonal motion-sensitizing directions at multiple excitation frequencies. Two-point GRE-based, Dixon MRI with TR/TE = 140/2.7 msec (water fat in-phase), 140/3.2 msec (water fat opposed-phase), and the same slice geometry settings as MRE was performed to quantify MRI-HFF. Detailed imaging parameters are shown in Table 1.

7T Imaging

MRI and MRE were performed on a 7T/20-cm Bruker Biospec system using an 86-mm inner diameter rodent body radiofrequency (RF) coil (Bruker/Biospin) for signal transmitting and a planar surface coil (20-mm diameter loop coil) with a low-noise preamplifier (MT0105) for receiving. Similar to the 3T site, mice were positioned supine for needle insertion, and the planar surface coil was then placed on the abdomen with the needle positioned through the center of the loop coil. The pneumatic passive needle driver was customized and embedded with the mouse cradle to fit within the rodent body coil to generate longitudinal needle vibrations in the mouse liver during MRE scans.

The position of the passive needle driver was adjustable with respect to the needle placement to ensure that the longitudinal needle vibrations were perpendicular to the coronal imaging planes. Wave images were acquired using an SE-MRE sequence with a pair of sinusoidal-shaped motion-encoding gradients (MEGs) on both sides of the 180° refocusing pulse (eg, 5-msec duration and 2.5-msec separation for 200 Hz): TR/TE = 400/16 msec, four phase offsets, and 12 slices of 1-mm thickness. The SE-MRE sequence contained two sets of acquisitions, with and without reversal of the MEGs. The phase-contrast subtraction of the two acquisitions removed the static phase component and doubled the wave amplitude. The TR was set as an integer multiple of the wave period to allow for continuous shear wave motion during the MRE scan. Two-point SE-based Dixon MRI was performed with TR/TE = 2000/8 msec, water and fat in-phase, water and fat opposed-phase with the refocusing RF shifted by 240 μsec (π/2 at 7T), and respiration gating to quantify MRI-HFF.¹⁰

Pneumatic Needle Driver

A silver needle (Asahi Medical Instrument Co., Japan) was used in the mouse MRE studies to generate shear waves inside the liver tissue. Longitudinally oriented needle vibrations were introduced into the mouse liver through a pneumatic passive needle driver developed for the mouse liver MRE studies.^{34, 35} The passive needle driver was excited by an active pneumatic driver (Resoundant Inc., Rochester, MN) located outside of the scanning room. Phantom MRE tests³⁶ were performed to verify the MRE settings on 7T and to compare the MRE-SS measurements generated on both 3 and 7T systems using the identical cylindrical phantoms made by the tissue-mimicking polyvinyl chloride (LureCraft Inc., Orland, IN) mixtures to model the mouse abdomen, with various stiffness representing the valid range of mouse liver SS. The representative shear wave images and elastograms are demonstrated in Figs S1 and S2 in the Supplemental Material.

MRI/MRE Analysis

The 3T MRE data were analyzed using a customized program in MATLAB version 2019a (MathWorks, Inc., Natick, MA). The magnitude of the complex shear modulus (SS reported in units of kilopascals [kPa]) was calculated from the measured MRE shear wave displacement data using a 3D direct inversion of the Helmholtz wave equation using the curl of the measured wave field.³⁴ The 7T MRE data were analyzed using customized MRE-Lab software (Mayo Clinic). SS was calculated using the local frequency estimation (LFE) inversion algorithm. Two-point Dixon MR images were used to define liver regions of interest (ROIs) for MRI-HFF and MRE-SS measurements. Increased fat content in the CDAHFD mouse liver resulted in enhanced contrast in the opposed-phase images, which significantly facilitated the liver tissue segmentation. The liver ROIs were delineated by J.L. (3T site) and H.T. (7T site), both of whom have over 5 years of experience with this type of analysis. The pixel values within liver ROIs for six slices for the 3T data and 10 slices for the 7T data (excluding the first and last slices to avoid edge artifacts) were pooled to calculate a mean SS for the entire volume. Wave amplitude and share wave propagation were reviewed using the dynamic phase-contrast images, shear wave maps, and the LFE certainty estimate maps for quality control. Large vessels, the gall bladder fossa, edges, and the area surrounding the needle were excluded from the liver ROIs.

Necropsy

Mice were sacrificed approximately 72 hours after the final PEG-FGF21v dose, 1 week after the final MRE imaging at the 3T site (Fig. 1, dashed dark gray arrow) and 72 hours after the final PEG-FGF21v dose and on the same day following terminal MRE imaging at the 7T site (Fig. 1, solid dark gray arrow). Mice were anesthetized with isoflurane and then euthanized by cervical dislocation.

Blood samples were collected via cardiac puncture and placed in Sarstedt Microvette tubes coated with K2-ethylenediaminetetraacetic acid (EDTA) (Aktiengssellschaft and Co., Germany) at the 7T site or sodium citrate at the 3T site. The blood samples were then centrifuged at 4 °C at approximately 4000–6000 RPMs for 20 minutes, and plasma was separated into aliquots and stored at −80 °C for later processing. Blood samples were analyzed for glucose, lipid panels, total bile acids, alkaline phosphatase (ALP), alanine transaminase, and aspartate transaminase. The liver of each mouse was isolated and weighed. A piece of the peripheral left lateral lobe was harvested and stored at −80 °C for biochemical analysis of liver triglyceride (TG) content. The medial and left lateral lobes of the liver were collected with a fresh blade, fixed for 24–48 hours at 4 °C in 10% neutral buffered formalin and then stored in 75% ethanol prior to paraffin embedding for histology analysis. Histological slices were stained with picrosirius red (PSR) for quantification of liver fibrosis content (area %) and stained with hematoxylin & eosin (H&E) for quantification of liver steatosis (vacuole area %) using HALO software (version 2.3, Indica Labs, Inc.). Biochemical assays and histology for all samples from both the 3 and 7T studies were completed by the 7T site (Bristol Myers Squibb, Lawrenceville, NJ), although they were analyzed by the same reader (S.B.) at different times based on when each study was concluded.

Statistical Analyses

The primary metric of interest was the attenuation ratio (R_DI) of the decrease (D) in true mean response from CDAHFD (μ_CDAHFD) to PEG-FGF21v (μ_{CDAHFD+PEG-FGF21v}), relative to the increase (I) in true mean response from CD (μ_CD) to CDAHFD (μ_CDAHFD):

$R_{\mathrm{DI}}=\left[{\mu }_{\text{CDAHFD}+\mathrm{PEG}-\mathrm{FGF}21\mathrm{v}}-{\mu }_{\text{CDAHFD}}\left]/\right[{\mu }_{\text{CDAHFD}}-{\mu }_{\mathrm{CD}}\right]$

A Bayesian posterior distribution for R_DI was constructed from each of the 3 and 7T SS data. Posterior means and the corresponding 95% credible intervals are presented. Weakly informative prior distributions for population means and vague prior distributions for population precisions (1/population variances) were constructed based on the information presented by Yin et al.³⁵ A two-tailed Dunnett's multiple comparison procedure was used to compare each of the CD and CDAHFD + PEG-FGF21v groups with the CDAHFD group at WK8. Statistical significance was defined as P < 0.05.

Dunnett's multiple comparison procedure was similarly performed for SS at other time points and for other endpoints. All values are presented as mean ± SE in the results. The corresponding values of R_DI were estimated simply by ratios of the differences of the arithmetic means of the data. Associations between imaging measurements, MRI-HFF and MRE-SS, and histological measurements were determined using Spearman correlation. The main analysis population consisted of all mice with evaluable imaging and histological assessments in the three study groups at WK8 and in the satellite groups at BD, BL, and WK4 (7T). A Bayesian variance components analysis was performed on the WK8 data from the 3T site to evaluate the proportion of total variability due to cages within each of the three treatment groups (Table S1 in the Supplemental Material).

Limitations

The treatment efficacy of PEG-FGF21v was reproducible across the clinical (3T) and preclinical (7T) imaging platforms, as similar proportional reductions of elevated SS were detected in both systems. Given that the dynamic range of SS for the 3T system was smaller compared with the 7T system, the lower SS values at the higher end (stiffer tissue) for the 3T-based findings may be due to the limited resolution and SNR at 3T and/or potential distortion caused by the SE-EPI sequence. For a given shear wave frequency (eg, 200 Hz), a longer wavelength and fewer waves pass through a stiffer object, which results in an underestimation of SS in a small and/or stiffer region in the mouse liver. Using higher frequency shear waves can improve the depiction of these smaller and stiffer regions, but high-frequency shear waves attenuate exponentially with increasing distance from the needle, which restricts the useful area of shear wave visualization and the quality of the MRE-SS measurement in the liver. SE MRE at 7T with stronger gradient hardware has advantages over the 3T clinical scanner by providing improved spatial resolution and greater SNR for better wave depiction that would allow for higher frequency shear waves to be visualized throughout the mouse liver. However, as tissue mechanical properties are frequency-dependent, this study was designed to use the same mechanical frequency for MRE at both sites, which required the study to use the highest vibration frequency that was practical for the 3T site.

It was noted that MRI-HFF measurements were lower at the 3T site compared with that of the 7T site. First, there were intersite variations in the histology and biochemistry measurements. Assuming that MRI-HFF is proportional to H&E vacuole area or TG, the lower H&E or TG values would verify that the MRI-HFF should be lower at the 3T site. Although mouse age and diet were the same at both sites and the samples were analyzed at the same laboratory, there were intersite variations in housing the mice (group housing at 3T vs. single housing at 7T). Second, there were several differences in the scanner and sequence setups. The SE-based Dixon technique and the use of a 7T small-bore scanner allows for much smaller in-phase and out-of-phase time of echos than can be achieved using a GRE-based Dixon technique, particularly on a less powerful 3T clinical scanner. Because of the T₂* effects in the GRE-based two-point Dixon technique, there was more signal dropout in the fat data than that of the SE-based two-point Dixon method. Two- or multipoint Dixon-based water–fat separation methods are limited in their ability to quantify fat as a result of T₂* decay, T₁-related bias, and the lack of accurate separation of fat, which has multiple spectral peaks, especially at lower fields, leading to inaccurate quantification of fat signals and fat fraction at 3T.²¹ The MRI-HFF is not equivalent to the MRI-PDFF. MRI methods that employ independent T₂* corrections for water and fat, together with spectral modeling, should be considered to enable more accurate quantification of liver fat with the 3T scanner.²¹ Nonetheless, MRI-HFF measured at both the 3 and 7T sites were highly correlated with the H&E vacuole areas and TG measurements, although MRI-HFF examines the entire liver, while the histology and TG utilizes cross sections and very small portions from two lobes. The treatment efficacy of PEG-FGF21v, as assessed by the relative changes of MRI-HFF, was reproducible across the 3 and 7T imaging platforms.

In general, the pattern and percent changes for the measurements between the two sites were in line with each other. Differences in the absolute values of the biochemical assays were assessed and were found to fall within the normal variation seen for this model and could be due to the slight differences in the model (housing and handling of different vivarium staff and scientists) and sample collections (sodium citrate at 3T vs. EDTA at 7T) between the two sites. The two sites used the same needle oscillator for MRE scans. The acupuncture needles were inserted with a fixed depth, and the needle trajectory was checked using the anatomical MRI to ensure avoidance of major blood vessels, gall bladder fossa, and ribs. No injuries caused by the needle were indicated in the histological samples collected from the mice in the study. In addition, the plasma chemistries and histology examinations supported the liver imaging results in this study. Histology and biochemical assays (TG, PSR, ALP, and so on) of the study group of CDAHFD control mice (four repeated MREs, vehicle administrations) at WK8 followed the time course trend of the three satellite groups (single MRE, no vehicle administration), each sacrificed at BD, BL, and WK4. The cross-sectional MRI-HFF and MRE-SS of the satellite groups are consistent with the longitudinal MRI/MRE measurements of the study group at BD, BL, and WK4. These data suggest that repeated vehicle administration and the needle oscillator did not cause notable adverse effects during the longitudinal study.