Molecular MR Imaging of Renal Fibrogenesis in Mice

Significance Statement

Renal fibrosis is the final common injury pathway for nearly all CKDs. Direct quantification of renal fibrosis would help to accurately assess disease progression and facilitate the development of novel antifibrotic therapies. We describe the feasibility of molecular imaging using a magnetic resonance (MR) probe, Gd-oxyamine (OA), in two mouse models of renal fibrosis (Alport syndrome and nephrotoxic nephritis). Quantitative, noninvasive MR imaging shows that the Gd-OA probe concentration increases in the renal tissues of diseased animals and that the imaging measure is proportional to the extent of tissue fibrosis assessed biochemically. Gd-OA molecular MR imaging is a potentially useful method to detect and stage renal fibrosis.

Abstract

Background

In most CKDs, lysyl oxidase oxidation of collagen forms allysine side chains, which then form stable crosslinks. We hypothesized that MRI with the allysine-targeted probe Gd-oxyamine (OA) could be used to measure this process and noninvasively detect renal fibrosis.

Methods

Two mouse models were used: hereditary nephritis in Col4a3-deficient mice (Alport model) and a glomerulonephritis model, nephrotoxic nephritis (NTN). MRI measured the difference in kidney relaxation rate, ΔR1, after intravenous Gd-OA administration. Renal tissue was collected for biochemical and histological analysis.

Results

ΔR1 was increased in the renal cortex of NTN mice and in both the cortex and the medulla of Alport mice. Ex vivo tissue analyses showed increased collagen and Gd-OA levels in fibrotic renal tissues and a high correlation between tissue collagen and ΔR1.

Conclusions

Magnetic resonance imaging using Gd-OA is potentially a valuable tool for detecting and staging renal fibrogenesis.

Introduction

Renal fibrosis is the final common injury pathway for nearly all CKDs,¹ which afflict 500 million people worldwide² and include diabetes, hypertension, and many forms of chronic glomerulonephritis and tubulointerstitial diseases.^3–7 Many will develop ESKD, characterized by glomerulosclerosis, tubular atrophy, and interstitial fibrosis.⁸ There is currently no effective treatment for preventing the progression of renal fibrosis and CKD, resulting in expensive and resource-intensive health care management of declining renal function, including dialysis or kidney transplant.^9–12 Antihypertensive therapies such as renin–angiotensin modulators are commonly used to delay ESKD,^13–15 but identifying treatment strategies for CKD remains a major public health objective.¹⁶

There is an established clinical correlation between ESKD and the extent of renal fibrosis.¹⁷ Renal function progressively declines in response to the excessive accumulation of extracellular matrix proteins,^18,19 such as collagen types I and III,^20,21 along with lysyl oxidase (LOX) enzymes, which crosslink the collagen fibrils.^22,23 Accurate quantification of renal fibrosis may predict the long-term outcome of renal function before significant renal dysfunction occurs. It would also provide a surrogate end point to evaluate the response of novel antifibrotic therapies in clinical trials.²⁴

Biopsy is the gold standard for diagnosing renal fibrosis but is unsuitable for monitoring disease progression because it is invasive and subject to sampling error.²⁵ Nonspecific markers such as eGFR provide no insight into the underlying renal pathology, and there remains a major unmet medical need to develop noninvasive strategies to detect and monitor the progression of renal fibrosis.²⁴ Techniques to assess microvascular loss and increased renal stiffening during kidney fibrosis have been investigated, including diffusion magnetic resonance imaging (MRI), arterial spin labeling, blood oxygenation level dependent-MRI, and MR elastography.²⁶ While these noninvasive MR techniques provide whole kidney data, they are not specific for fibrosis because other molecular factors contribute to the values measured by these techniques.

Recently, molecular imaging approaches targeting extracellular matrix components such as elastin or collagen have been applied to renal fibrosis in animal models.^27,28 We have developed probes targeting the allysine residue formed by the action of LOX enzymes on collagens during fibrogenesis. The allysine residues on oxidized collagen are intermediates in crosslinking reactions. We have found this target to be useful for imaging fibrogenesis in animal lung and liver fibrosis models.^29–34 Here, we hypothesize that intravenous administration of the allysine-targeted probe Gd-oxyamine (OA) will result in extravasation from the vasculature to the interstitial space of the kidney followed by binding to oxidized collagen formed during fibrogenesis. Gd-OA enhanced MRI would thus provide a quantitative molecular readout of LOX-mediated collagen oxidation and renal fibrogenesis and enable disease detection in preclinical renal fibrosis models.

Methods

Probe

Gd-OA is a water-soluble, low molecular weight, extracellular gadolinium-based imaging agent functionalized with an OA moiety for targeting allysine and was synthesized as previously reported.³⁰ It binds to aldehyde-rich proteins and allysine-rich porcine aorta (K_d=360 μM). The relaxivity of Gd-OA increases from 4.3 mM⁻¹s⁻¹ when unbound to 16.9 mM⁻¹s⁻¹ when bound to an allysine-bearing protein.

Animal Models

All experiments were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee. A total of 72 mice were used. The nephrotoxic nephritis (NTN) model was induced as previously reported with some modifications.³⁵ In brief, 6-week-old male “129 Sc/Ev” mice (Charles River Labs, Wilmington, MA) were dosed at day 0 with 250 µg of Sheep immunoglobulin G (Lampire Biological Laboratory, Pipersville, PA) and then dosed with 125 µl of sheep anti-rat glomeruli antibody (glomerular basement membrane [GBM]) serum (PTX-001S; Probetex, San Antonio, TX) at day 5. Control animals were dosed with 125 µl of PBS on day 5 (NTN: n=21, Control: n=9). For the mouse model of Alport syndrome, male and female Col4a3-deficient (Col4a3KO) mice with 129/SvJ background (129-Col4a3tm1Dec/J) were purchased from the Jackson Laboratory (Bar Harbor, ME) and maintained as a heterozygous colony (knockout [KO] mice: n=16; heterozygous: n=26).

MR Imaging and Analysis

NTN mice were imaged on day 12 (1 week after serum injection). Col4a3KO mice and heterozygous littermates were imaged when they were 6–9 weeks old. Animals were anesthetized with isoflurane (1%–2%) and placed in a specially designed cradle inside a custom-built volume coil of a small-bore 9.4 T MRI scanner. Body temperature was maintained at 37°C with warm air. Anesthesia was adjusted to maintain a respiration rate of 40±5 breaths per minute using an MR-compatible small animal respiratory monitoring system (Model 1025, SA Instruments, Inc., NY). Mice were imaged twice: An inversion recovery sequence (fast spin echo sequence with nine inversion times ranging from 7 to 5000 ms, repetition time 5000 ms, echo time 7.26 ms, acceleration factor 16, respiratory gated, field of view 30×30 mm, matrix 128×128, single slices 0.75 mm thick, acquisition time approximately 10 minutes) was used to measure R1 maps of the kidneys; mice were removed from the scanner, injected with 100 µmol/kg Gd-OA (≤100 µl; intravenous) and then allowed to wake up before returned to the scanner 4 or 24 hours later and another R1 map was obtained. Gd-OA was coinjected with an equimolar amount of Eu-DOTA (nonbinding, MRI silent control probe). After the second imaging session, the mice were euthanized, and the kidneys were removed for further analysis. The renal cortex and medulla were segmented in the MR images based on the contrast in the baseline inversion recovery images. T1 was fit voxel-by-voxel, converted to R1 voxelwise, and then averaged to obtain the R1 value of each region of interest.

Gene Expression

RNA was isolated from kidney tissue using TRIzol (Life Technologies, Grand Island, NY) according to the manufacturer's instructions and subsequently treated with DNAse I (Promega, Madison, WI). Total RNA (1 µg) from each sample was used to synthesize cDNA by single-strand reverse transcription (SuperScript III First-Strand Synthesis SuperMix; Life Technologies). Gene expression was analyzed by quantitative real-time PCR using TaqMan gene expression assays (Life Technologies) on Applied Bioscience 7900HT Fast Real-Time PCR system using 96-well plates with a reaction volume of 20 μl. The 2^−∆C.T. method was used for the relative quantification of mRNA with normalization to 18S.³⁶ The Taqman assays were as follows: 18S (Hs03003631_g1), Col1a1 (Mm00801666_g1), Acta2 (Mm00725412_g1), Timp1 (Mm), Lox (Mm00495386_m1), Loxl1 (Mm01145738_m1), Loxl2 (Mm00804740_m1), Loxl3 (Mm01184865_m1), and Loxl4 (Mm00446385_m1).

Tissue Analysis

Fibrosis was assessed in trichrome and Sirius red stains and collagen III immunohistochemistry (1:200 dilution of polyclonal rabbit anti-human collagen III, LS-B693; Lifespan Biosciences, Seattle, WA).³⁷ Whole kidney (NTN model) or renal cortex and medulla (Alport model) were acid-digested to measure hydroxyproline (by HPLC)³⁸ and gadolinium/europium (by inductively coupled plasma mass spectrometry). The results are expressed as amounts per wet weight of tissue.

Statistics

All data are shown as mean±SEM. Normality was confirmed with the Shapiro-Wilk test. Differences between the groups were tested with an unpaired t test or the Mann–Whitney test with P < 0.05 considered significant.

Results and Discussion

In the NTN model, MRI was performed 1 week after anti-GBM serum administration, where there was glomerular injury and fibrosis as evidenced by Trichrome and collagen III stainings (Figure 1, A and B). Gene expression of LOX and its paralogs (LOXL1–4) were also significantly upregulated in the NTN mice at this time point compared with vehicle-treated mice (Figure 1C). Kidney hydroxyproline, a quantitative measure of collagen content, was increased by 50% in the NTN group compared with controls (Figure 1D). The hydrophilic Gd-OA species is eliminated by renal filtration and has a blood elimination half-life of 5 minutes in mice with normal renal function. The half-life for hydrolysis of the oxime bond formed between Gd-OA and a model aliphatic aldehyde was 1120±72 hours at pH 7.4°C and 25°C. To ensure minimal signal from unbound Gd-OA probe, mice were imaged 24 hours postinjection. Representative coronal T1-weighted MR images of control and NTN mice are shown in Figure 2A, along with kidney R1 maps overlaid in color (Figure 2B) that were measured before and 24 hours post-Gd-OA injection. The kidneys of the NTN mice appear different on T1-weighted MRI, with a notable punctate appearance in the cortex (Figure 2A). Histogram analysis of these imaging data showed a shift to the right (i.e., contrast enhancement) in the mouse with NTN but not in the control animal (Figure 2C). The change in R1, ΔR1, observed post-Gd-OA is directly proportional to the amount of probe present. There was a three-fold increase in ΔR1 of the cortex of NTN mice compared with control mice (Figure 2D). This result was in good agreement with the amount of Gd measured in the kidneys by elemental analysis in which the Gd level was four-fold higher in the NTN mice compared with the control mice (Figure 2E). To confirm that Gd-OA uptake in NTN kidney was due to specific probe binding and not a result of impaired renal function, the uptake of Gd-OA was compared with the nonspecific and MRI silent probe Eu-DOTA, which was coinjected at an equal concentration. The Gd:Eu ratio was significantly higher in the NTN group than in the sham animals (Figure 2F), confirming the specific uptake of the targeted Gd-OA probe; if the higher Gd-OA level was a result of differences in pharmacokinetics, then the Gd:Eu ratio would be unchanged between NTN and sham mice. We also measured serum creatinine and BUN levels in control and NTN animals as a surrogate for renal function. No differences were observed between groups indicating that in this model of early-stage renal fibrosis, Gd-OA reflects changes in fibrogenesis before renal function has declined.

Figure 1.

Characterization of NTN mouse model. (A) Representative trichrome stains showing diffuse tubular injury with intratubular casts (*) and glomerulosclerosis (arrow) in the mouse with NTN. (B) Representative immunohistochemical staining for collagen III showing mild fibrosis in the animals receiving the nephrotoxic serum. (C) LOX gene expression expressed as relative quantification of the measured value normalized to the mean of the control was significantly upregulated in NTN mice (orange) compared with the control (blue). P values on the graph were calculated using the Mann–Whitney test for LOX, unpaired t test for LOXL1–4. Error bars show SEM. (D) Hydroxyproline (µg per gram kidney) as a biochemical surrogate for total collagen in kidneys showing >50% increase in the NTN animals. P value on the graph was calculated using the Mann–Whitney test. Error bars show SEM.

Figure 2.

Gd-OA detects fibrogenesis in the NTN mouse model. Representative T1-weighted coronal MR images (A) and R1 maps (B) of the kidneys from control and NTN mice before and 24 hours postinjection of Gd-OA. Note the more punctate enhancement in the cortex of the NTN kidneys in T1-weighted images (A). R1 maps of the kidney at baseline and 24 hours post-Gd-OA injection showing higher R1 values in the cortex of the NTN mouse after injection but very little R1 change in the sham mouse (B). Associated histograms of cortical R1 values (C) pre-Gd-OA and post-Gd-OA showing a shift to the right in the NTN mouse. (D) MR signal change computed as ΔR1 showed a three-fold increase in the NTN mice compared with control mice. P value on the graph was calculated using an unpaired t test. Error bars show SEM. (E) Ex vivo analysis of Gd in the kidneys showing a four-fold increase in probe in the NTN kidneys compared with control mice. P value on the graph was calculated using the Mann–Whitney test. Error bars show SEM. (F) Ex vivo analysis of Gd-OA and Eu-DOTA (nontargeted control) in the kidneys of control and NTN mice showing specific binding to NTN kidney. P value on the graph was calculated using the Mann–Whitney test. Error bars show SEM.

We next evaluated Gd-OA in a progressive model of chronic hereditary glomerular disease using Col4a3KO mice (Alport model). Col4a3KO mice showed histopathological changes associated with kidney damage including glomerulosclerosis and interstitial fibrosis, tubular changes, and hyaline casts (Figure 3A). Gene expression analysis showed an upregulation only in LOX, LOXL1, and LOXL2 (Figure 3B). Overall, there was >50% increase in hydroxyproline in the cortex (470±3 versus 718±27 µg/g, P < 0.0001) and the medulla (650±5 versus 1040±34 µg/g, P < 0.0001) of the Alport mice compared with heterozygous controls.

Figure 3.

Characterization of the Alport mouse model. (A): Representative hematoxylin and eosin staining of the kidneys of a heterozygous mouse (Col4a3 Het) and Alport mouse (Col4a3KO) show sclerotic glomeruli (circle) compared with the normal-appearing glomeruli (arrow) as well as tubular dilatation and hyaline casts (*). (B) LOX gene expression was significantly upregulated in Alport mice (orange) compared with the heterozygous animals (blue). P values on the graph were calculated using the Mann–Whitney test for LOX, an unpaired t test for LOXL1–4. Error bars show SEM.

The fibrosis varied from mild to severe in Alport mice, as detected by Sirius red staining (Figure 4A. Pilot studies established that nonspecific renal contrast with Gd-DOTA (non-targeted control MR probe) returned to baseline levels by 4 hours postinjection, and this time point was used in the MRI assessments of the Alport model. Figure 4, B and C shows T1-weighted images and R1 histograms before and 4 hours post-Gd-OA for a heterozygous mouse with no disease and two mice with increasing fibrosis, all having normal renal function as assessed by creatinine. With the early onset of fibrosis, there are no obvious differences in the baseline T1-weighted images, but Gd-OA enhanced signal is apparent in these mice and increases with increasing fibrosis. This signal change is quantified by ΔR1, which correlated linearly with hydroxyproline in the cortex (r²=0.83) and medulla (r²=0.82) (Figure 4D). There were no significant differences in ΔR1 values between female and male mice in either the heterozygous or knockout mice for either cortex or medulla.

Figure 4.

Gd-OA detects and quantifies fibrogenesis in the Alport mouse model. Representative Sirius red staining (A) T1-weighted coronal MR images (B) and histograms (C) of the kidneys of a heterozygous mouse (top row), Alport mouse with mild fibrosis (middle row), and Alport mouse with severe fibrosis (bottom row) 4 hours postinjection of Gd-OA. Cortical R1 values were higher in the Alport than in the heterozygous animals, depending on the severity of the disease, as assessed by the Sirius Red staining. The degree of fibrosis in stained sections was quantified by calculating the CPA. Linear regression analysis showed a significant correlation between ΔR1 and hydroxyproline in the renal cortex and medulla of Alport mice (D). Simple linear regression R² values for Col4a3KO (KO) are displayed on the graphs (P < 0.0001 for cortex, P = 0.0002 for medulla). Col4a3Het (Het) cortex R²=0.05135 (P = 0.3659), medulla R²=0.05135 (P = 0.5190). CPA, collagen proportionate area.

There are some limitations to the study. Contrast agents are regulated as drugs, and clinical translation of this molecular probe will first require extensive safety and toxicological studies in animal models before human studies can begin, and this precluded us from testing the probe in human subjects. There remains a concern about the safety of a gadolinium chelate in patients with poor renal function. To minimize this risk, we used the Gd-DOTA core, which is the most chemically inert contrast agent and for which there have been no unconfounded reports of nephrogenic systemic fibrosis. Alternately, the gadolinium chelate could be replaced by chelated manganese for MRI signal generation.³⁹ In this study, we used an accurate, but slow method, to measure T1. However, clinical MRI scanners have sequences that can map T1 throughout both kidneys in 1–2 minutes.⁴⁰

Several approaches have been used for imaging fibrosis, often using labeled peptides or antibodies that are not amenable to kidney imaging.⁴¹ Here, we show a small molecule designed to target allysine residues formed on collagen during crosslinking. This probe shows promise for detecting fibrogenesis in patients with CKD before the renal function is lost and for monitoring response to antifibrotic therapies in clinical trials.

Disclosures

I. Ay reports employer: Takeda Pharmaceuticals (spouse). P. Caravan has equity in and is a consultant to Collagen Medical LLC, has equity in Reveal Pharmaceuticals Inc., and has research support from Indalo Therapeutics, Pfizer, Janssen, Pliant Therapeutics, Takeda, and Transcode Therapeutics. P. Caravan also reports research funding: Mariana Oncology; patents or royalties: Factor 1A LLC and Reveal Pharmaceuticals; and advisory or leadership role: Board member Reveal Pharmaceuticals, unpaid. R.B. Colvin reports consultancy: eGenesism Sangamo and NephoSant; research funding: Egenesis; and advisory or leadership role: NephroSant. B.C. Fuchs was a consultant for Gilead and had research support from Blade Therapeutics, Collagen Medical, and Enanta. B.C. Fuchs also reports employer: Ferring Pharmaceuticals and advisory or leadership role: Mediar Therapeutics. R.V.P. Martinez is employee of Pfizer. R.V.P. Martinez also reports ownership interest: Pfizer and research funding: Pfizer. I.A. Rosales reports advisory or leadership role: Frontiers in Transplantation (Associate Editor, unpaid) and Philippine Journal of Pathology (associate editor, unpaid). F.J. Schlerman is employee of Pfizer. F.J. Schlerman also reports ownership interest: Pfizer. J. Wellen reports employer: Bristol Myers Squibb, Janssen Research and Development, and Pfizer Global Research and Development and ownership interest: Bristol Myers Squibb, Johnson & Johnson, and Pfizer. All remaining authors have nothing to disclose.

Funding

This work was supported by a research grant from Pfizer to P. Caravan and the following National Institutes of Health grants: DK104956, DK104302, DK121789, OD025234, OD010650, and OD032138.

Author Contributions

Conceptualization: Peter Caravan, Robert B. Colvin, Bryan C. Fuchs, Robert V.P. Martinez, Franklin J. Schlerman, Jeremy Wellen.

Formal analysis: Ilknur Ay, Gunisha Arora, Yinching Iris Chen, Derek J. Erstad, Diego S. Ferreira, Ivy A. Rosales, Philip A. Waghorn.

Funding acquisition: Peter Caravan, Bryan C. Fuchs, Robert V.P. Martinez.

Investigation: Gunisha Arora, Yinching Iris Chen, Derek J. Erstad, Diego S. Ferreira, Chloe M. Jones, Ivy A. Rosales, Nicholas J. Rotile, Philip A. Waghorn.

Methodology: Gunisha Arora, Derek J. Erstad, Bryan C. Fuchs, Ivy A. Rosales, Franklin J. Schlerman, Lan Wei.

Project administration: Peter Caravan, Bryan C. Fuchs.

Resources: Bryan C. Fuchs.

Supervision: Peter Caravan, Robert B. Colvin, Bryan C. Fuchs.

Writing – original draft: Ilknur Ay, Philip A. Waghorn, Bryan C. Fuchs.

Writing – review & editing: Ilknur Ay, Peter Caravan, Robert B. Colvin, Robert V.P. Martinez, Franklin J. Schlerman, Jeremy Wellen.

Data Sharing Statement

All data used in this study are available on request to the authors. Limited amounts of the Gd-OA probe are available and can be obtained via a material transfer agreement.