Measures of kidney function by minimally invasive techniques correlate with histological glomerular damage in SCID mice with adriamycin-induced nephropathy

Maximising the use of preclinical murine models of progressive kidney disease as test beds for therapies ideally requires kidney function to be measured repeatedly in a safe, minimally invasive manner. To date, most studies of murine nephropathy depend on unreliable markers of renal physiological function, exemplified by measuring blood levels of creatinine and urea, and on various end points necessitating sacrifice of experimental animals to assess histological damage, thus counteracting the principles of Replacement, Refinement and Reduction. Here, we applied two novel minimally invasive techniques to measure kidney function in SCID mice with adriamycin-induced nephropathy. We employed i) a transcutaneous device that measures the half-life of intravenously administered FITC-sinistrin, a molecule cleared by glomerular filtration; and ii) multispectral optoacoustic tomography, a photoacoustic imaging device that directly visualises the clearance of the near infrared dye, IRDye 800CW carboxylate. Measurements with either technique showed a significant impairment of renal function in experimental animals versus controls, with significant correlations with the proportion of scarred glomeruli five weeks after induction of injury. These technologies provide clinically relevant functional data and should be widely adopted for testing the efficacies of novel therapies. Moreover, their use will also lead to a reduction in experimental animal numbers.


Results
The half-life of transcutaneously measured FITC-sinistrin becomes significantly increased over the course of ADR-induced nephropathy. Kidney injury was induced by intravenous (iv) ADR administration once in 6 SCID female mice while 5 control SCID females received saline injection. The weights of ADR-administered versus control mice reached their lowest point one week after nephrotoxin administration. Over the next four weeks, weights in the ADR group returned towards timematched control values but did not attain them (Fig. 1A, S1A). To determine damage to the glomerular macromolecular barrier, we measured total 24-hour urinary excretion and the albumin:creatinine ratio on a weekly basis (Fig. 1B,C, S1B,C). Using either measure, albuminuria was significantly elevated in the ADR group at weeks 2 to 4, with attenuation at week 4 ( Fig. 1B been previously reported in ADR-induced SCID mice 4,9,17 . A linear mixed-effects (LME) model showed statistically significant linear and quadratic changes of albuminuria over time within the ADR group (p < 0.001 for both the linear and second degree terms, Table 1).
To assess changes in glomerular filtration over time with the transcutaneous device, we undertook serial measurements of FITC-sinistrin half-life over 4 weeks. Representative examples of clearance curves from control and ADR animals are shown in Fig. 2A-D. The FITC-sinistrin half-life in the ADR group was increased with statistical significance from week 2 onwards (Fig. 2E, Table S1A, S3). Given that GFR is inversely correlated to the FITC-sinistrin half-life 10,18 , these results suggest that GFR is slightly impaired 2 weeks after ADR administration and deteriorates further between weeks 3 and 4, reflecting the progressive nature of the nephropathy 19 . An LME model revealed statistically significant linear changes in the FITC-sinistrin half-life within the ADR group over time, with an average increment of 2.45 (= (0.02 + 0.33) × 7) minutes per week (p < 0.001). These changes were statistically different when compared with controls, in which no significant changes in half-life were detected (p = 0.76, Table 1). The levels of BUN and SCr as indicators of kidney function revealed no significant differences between the experimental control groups at week 5 ( Figure S1D,E).
The passage of IRDye through the kidney measured using MSOT is delayed in ADR mice. To investigate the potential of MSOT for evaluating renal function in ADR mice, IRDye clearance was Animals of the ADR group had a mean body weight loss of approximately 20% by week 1, while control animals had a mean body weight gain of 5%. (B) Mean albuminuria over 24 hours, measured weekly. ADRadministered animals had a mean maximum of albuminuria of approximately 90 mg/24 h at 3 weeks, while in control animals urinary albumin stayed constant throughout the study at around 0.03 mg/24 h. (C) Mean albumin:creatinine ratio (alb:cr) measured weekly, showing similar temporal dynamics to 24 h albuminuria. Data points (circles = control, n = 5; squares = ADR-administered, n = 6) and bars show mean ± standard error. Asterisks indicate significance of mixed-design ANOVA models: p ≤ 0.01 (**), p ≤ 0.001 (***), see also   monitored in real-time at week 5, before killing for necropsy studies. Movies and snap-shot images taken prior to, and 15 s, 45 s, 1 min 30 s and 3 min after administration of IRDye showed that its passage through the kidney appeared to be delayed in ADR versus control mice (Fig. 3A, supplementary movie 1, 2). More specifically, temporal colour maps of the clearance kinetics confirmed that in nephropathic mice, the dye took longer to transit from the kidney cortex to the papilla/pelvis region than in controls (Fig. 3B). Visualisation of the T MAX allows discrimination between the ADR and the control group, highlighted by the peak signal intensity in the cortex of the control animals appearing at ~30 s and visualised in yellow. By contrast, the yellow tone indicating 30 s peak signal intensity is absent in the ADR treated mice; instead, green colour tones visualise peak signal intensities that appear with a delay at around 1 min. Furthermore the papilla/pelvis region of control animals show peak concentrations from 1.5-3 minutes (green, blue, violet, red), whereas the ADR treated mice show peak concentrations in this region only after 3 minutes (violet, red, Fig. 3B). We determined the pharmacokinetics of IRDye clearance by calculating the T MAX delay between the renal cortex and pelvis (Fig. 3C,D, Table S1B), and the clearance half-life in the cortex, using exponential decay fitting (Fig. 3C,E, Table S1B). T MAX delay was significantly increased in the ADR-administered group versus controls (p = 0.015), and there was a non-significant increase in clearance half-life between these two groups (Fig. 3D,E; Table S2). However, equilibrium dialysis revealed that over 40% of IRDye bound to plasma proteins (Table S4), suggesting that the exponential decay might be shorter than expected in ADR mice due to leakage of plasma proteins through the glomerular filtration barrier, thus leading to a potential underestimation of the true clearance half-life in the cortex.

Minimally invasively measured clearance kinetics of sinistrin and IRDye show a strong correlation with glomerular histological damage.
To visualise and measure structural damage, we made use of Picro-Sirius Red (PSR) staining which binds specifically to fibrillar collagen type I and III under polarising microscopy. Analysing PSR staining on parasagittal kidney sections from mice necropsied at week 5, we determined that ADR kidneys contained a subset (12-29%, across the whole experimental group) of glomeruli with lesions ( Fig. 4B,C; Tables S1B, S2). In these glomeruli, there was a loss of capillary loops and extension of PSR staining into the glomerular tuft beyond the normal tree-like mesangial pattern. In control kidneys, such glomerular lesions were rarely detected (Fig. 4A). Quantification of total PSR staining (i.e. not confined to glomeruli) under bright field microscopy, and of fibrillar collagen under polarising microscopy, failed to show significant differences in ADR versus control kidneys (Fig. 4D,E, S2A,B; Tables S1B, S2). However, we noted the occurrence of proteinaceous casts together with flattened tubular epithelia in both PSR-and Masson's Trichrome-stained ADR kidneys viewed under bright field microscopy ( Figure S2B,D), but not in control kidneys ( Figure S2A, C). These observations show that in this model, at this particular time point, glomerular damage is the predominant lesion, while tubulo-interstitial changes are much less prominent. Next, we evaluated whether the two novel minimally invasive dye clearance methods we have used here to detect kidney filtration function, are suitable for long-term regenerative medicine therapy studies through their statistical association with detected glomerular damage. Importantly, since we had observed a wide range in the proportions of scarred glomeruli (as assessed by the above criteria) in the ADR kidneys, we questioned whether specific glomerular damage in individual mice was associated with specific measures of kidney function. Therefore, we examined the relationship in individual animals between glomerular histological lesions and (i) albuminuria, (ii) FITC-sinistrin half-life and (iii) IRDye excretion kinetics in individual animals (Fig. 5). In ADR mice, there was no significant association between glomerular histological damage (week 5) and the amount of albuminuria measured at the maximum (weeks 2-3) or at the final point of urine collection in week 4 (Table 2). Therefore, albuminuria is not a reliable measure of glomerular histological damage in this model. By contrast, assessment of the relationship between FITC-sinistrin half-life and the proportion of histologically damaged glomeruli revealed a significant positive correlation between FITC-sinistrin half-life values at week 4 and glomerular damage at week 5 (coefficient = 0.94, p = 0.001; see Table 2).
When we assessed the relationship at 5 weeks between glomerular histological damage and IRDye clearance half-life, we found a significant association in the ADR group (coefficient = 0.12, p = 0.02; Table 2). Furthermore, the T MAX delay was significantly associated with glomerular histological damage in the ADR group (coefficient = 0.16, p = 0.002; Table 2). The coefficient suggests that an increment of 60 seconds in T MAX delay is associated with an increment of glomerular histological damage of about 10%.

Discussion
Here, we report for the first time the use of transcutaneous measurement of FITC-sinistrin decay and MSOT detection of IRDye clearance as measures of glomerular filtration function in SCID mice with ADR-induced nephropathy. Our albuminuria measurements strongly suggest that ADR-induced nephropathy featured a loss of glomerular macromolecular barrier integrity in SCID mice from week 2, peaking at week 3, similarly to recently published results 4,9,17 . In ADR mice, the FITC-sinistrin half-life was significantly prolonged from week 2. The fact that the half-life continued to rise thereafter in the ADR group is a strong indicator of a progressive loss in the ability of the kidneys to excrete small molecules from the circulation, i.e. a progressive decline in GFR. Previous studies have demonstrated  Table S2.
the accuracy and reliability of the transcutaneous measurement of FITC-sinistrin half-life as a measure of GFR in various strains of rats and mice 10,11,18,[20][21][22][23] . Of note, although remaining elevated, the degree of albuminuria fell between weeks 3 and 4. This again might be explained by a loss in total filtration surface associated with glomerular damage as observed in week 5, consistent with the observed rise in FITC-sinistrin half-life during this period.
Using MSOT imaging, we showed here that IRDye is cleared in a dynamic fashion as it first appears in the kidney cortex and then transits to the renal pelvis, confirming a previous report 14 . The simplest  Table S2. interpretation of the T MAX delay of IRDye clearance we measured in the ADR versus control group, is that the glomerular clearance of this small molecule is impaired. The observation that some IRDye binds to plasma proteins, and that the ADR model features proteinuria, may have led to an underestimation of the true 'T MAX delay' . In addition, it is possible that the passage of IRDye along the lumen of kidney tubules may be compromised in the ADR model but this will require further study.
Our BUN and SCr measurements revealed that these are not useful indicators of ADR-induced nephropathy at a time point when sinistrin clearance deviated most from normal. Both BUN and SCr measurements have been questioned as biomarkers for the early detection of human and rodent kidney disease at the histological level since they frequently identify abnormal kidney function only in the later stages of the diseases 24,25 . Previously, an increase in SCr in ADR-induced nephropathy has only been

Figure 5. Correlation graphs showing the correlation of both control and ADR group data between the percentage of abnormal glomeruli and maximum observed albuminuria (A), FITC-Sinistrin half-life at week 4 (B), IRDye excretion half-life in the cortex at week 5 (C) and IRDye T MAX delay (D) at week 5.
Data points represent individual animals (circles = control, n = 5; squares = ADR-administered, n = 6). Trendlines for control (dashed line) and ADR-administered (solid line) animals are displayed.
observed in male rodents, while in female rodents, SCr levels remained unchanged when compared to control animals, even though pathohistology clearly indicated the full spectrum of renal damage 3,26-30 .
Crucially, our observations that BUN and SCr fail to detect significant differences in kidney function at a time point when histological changes are clearly quantifiable, emphasises the need to develop novel molecular biomarkers that are better able to monitor renal health and assess the efficacy of therapeutic interventions, including regenerative medicine therapies 24,31,32 .
We conclude that transcutaneous measurements of FITC-sinistrin half-life provide a minimally invasive method to repeatedly assess GFR in conscious mice, thus allowing for longitudinal evaluation of kidney function. Furthermore, our analysis suggests that MSOT imaging has the potential to be employed for the repeated and minimally invasive measurement of kidney function in mice by assessing renal clearance of injected small near infrared dyes. The use of near infrared dyes with negligible plasma protein binding properties will further improve MSOT imaging performance for the measurement of renal clearance kinetics. Optoacoustic imaging can also be employed to track administered stem cells labelled with gold nanorods, expressing tyrosinase or near infrared fluorophores in whole animals [33][34][35][36][37][38][39][40] . Therefore, this imaging technology will be of high importance for preclinical studies in regenerative approaches to nephropathies as it will allow the detection of labelled cells in parallel with functional measurements of renal clearance kinetics.
In order to assess whether functional kidney data correlated with glomerular scarring, we performed histological analysis of sections by staining with PSR and Masson's Trichrome. While SCID animals with ADR-induced nephropathy showed a strong elevation in the number of abnormal glomeruli at 5 weeks, at this time point we only observed occasional pathohistological changes in the tubulo-interstitial zone of ADR mice. Cortical glomerular and tubulo-interstitial damage have been reported previously in mice with ADR-induced nephropathy 19,26,28,41,42 .
Importantly, we aimed to statistically analyse whether any associations could be detected between glomerular histological damage and albuminuria, FITC-sinistrin half-life or IRDye clearance, respectively, in mice with ADR-induced nephropathy on the SCID background. Our evaluations provide evidence that FITC-sinistrin half-life and T MAX of IRDye clearance were strongly correlated with glomerular scarring in SCID mice with ADR-induced nephropathy. We therefore conclude that FITC-sinistrin half-life and IRDye clearance (T MAX ) measurements may be good predictors of glomerular histological damage. In contrast, neither peak albuminuria nor albuminuria at 4 weeks after ADR-induced nephropathy were significantly correlated with glomerular scarring, suggesting that urinary albumin measurements fail to provide an accurate reflection of the histological glomerular damage in this model at this time point.
In a recent report, rats with ADR-induced nephropathy were analysed with magnetic resonance imaging to monitor changes in kidney function and histopathology during disease progression. This study is another example, similar to the one we report here, to demonstrate the feasibility of a longitudinal, minimally-invasive imaging technology as an approach to assessment of changes in the kidney over time 43 .
In conclusion, our results indicate that the minimally invasive technologies for detecting FITC-sinistrin half-life and IRDye clearance kinetics could be widely adopted for in vivo monitoring of models of RMT in progressive kidney disease. Each technology potentially allows recurrent testing in individual animals, maximising the information obtained as nephropathy and/or regenerative effects progress. The observations that sinistrin clearance and IRDye kinetics significantly correlate with the proportions of damaged glomeruli in individual animals, indicates that these minimally invasive measurements could markedly reduce animal numbers in preclinical models of nephropathy.

Materials and Methods
Animals. Female BALB/c severe combined immunodeficient (SCID) mice (Charles River, Margate, UK) were housed in individually ventilated cages at a 12 hour light/dark cycle, with ad libitum access to food and water. At age seven to eight weeks, ADR-induced nephropathy was induced in six mice  Table 2. Table summarising the associations in the ADR group between abnormal glomeruli (5w) and the four biomarkers: albuminuria at the maximum observed and at four weeks, FITC-Sinistrin halflife, IRDye clearance half-life, IRDye T MAX delay. The results included in this table were derived from the statistical models described in Table S5 by using contrast analysis. Each coefficient estimate indicates the change in percentage of abnormal glomeruli per unit change of the corresponding biomarker.
by injecting once intravenously (iv) adriamycin (ADR, doxorubicin hydrochloride, Tocris, Bristol, UK) at 6.3 mg/kg body weight (BW) in 0.9% saline (Braun, Melsungen, Germany), while five control mice received saline. The optimal dose had been previously determined in a dose-finding study, and is similar to previously reported adriamycin dose given to BALB/c SCID mice 4,9,17 . Mortality in the ADRadministered group during the five-week study period was zero. Experimental animal protocols were performed in accordance with the approved guidelines under a licence granted under the Animals (Scientific Procedures) Act 1986 and approved by the University of Liverpool Animal Ethics Committee. BALB/c SCID mice were used in order to determine parameters for future preclinical regenerative medicine studies in mice with ADR-induced nephropathy.
Albuminuria, serum creatinine and BUN. To collect urine, mice were housed individually in metabolic cages (Tecniplast, Buguggiate, Italy) once a week for 24 hours (h). Despite being gradually acclimated to the metabolic cages, 24 h urine collection is stressful for mice, leading to weight loss of up to 1.5 g during the time spent in the cages. Because the mice on average weigh less than 20 g, it was necessary to omit urine collection at week five, since the long term anaesthesia for MSOT imaging is another stressful procedure for the mice. Total urine volume was measured and albumin levels were quantified using a Mouse Albumin ELISA Quantification Kit (Bethyl Laboratories, Montgomery, TX, USA) according to manufacturer's instructions. Urinary creatinine (UCr) was quantified using a plate based colourimetric assay. Blood was collected via cardiac puncture after sacrifice, and separated into serum. Serum creatinine (SCr) and blood urea nitrogen (BUN) were quantified according to manufacturer's instructions (Detect X Serum Creatinine Detection Kit, Arbor Assays, Ann Arbor, MI, USA; QuantiChrom Urea Assay Kit, BioAssay Systems, Hayward, CA, USA, respectively).

Histopathology.
Kidneys were dissected at week 5, cut along the sagittal plane, and processed for histology using standard methods. Sections (4 μ m) were stained with Picro-Sirius Red (PSR, Sigma-Aldrich, Dorset, UK) or Masson's Trichrome following standard protocols. All glomeruli (on average 116 glomeruli) in one PSR-stained kidney section per animal were blindly scored for glomerular histological damage and the percentage of abnormal glomeruli calculated. Glomeruli were categorised as abnormal if the PSR staining extended beyond the confines of the mesangium into the glomerular tufts. These scarred glomeruli also featured obliteration of capillary spaces and adhesions of tufts to the Bowman capsule. Using one panorama kidney section per animal, fibrillar collagen was detected using polarised microscopy, while PSR-staining in the cortex was captured by bright field microscopy. Images were stitched together 44 and image analysis software (Fiji) was used for quantification.
In short, the transcutaneous device (Mannheim Pharma & Diagnostics GmbH, Mannheim, Germany) was fixed to the depilated skin on the back of mice using a double-sided adhesive patch (Lohmann, Neuwied, Germany). Transcutaneous measurement started with background reading one to three min before 0.3 mg/g BW FITC-Sinistrin (Fresenius Kabi, Linz, Austria; diluted in 0.9% saline; Braun, Melsungen, Germany) was administered iv. Animals were allowed to fully recover and move freely until transcutaneous measurement was stopped after 90 minutes (min). Using a 1-compartment model, the half-life of FITC-Sinistrin was calculated from the transcutaneously measured kinetics ( Figure S3) 11 .
Measuring the clearance of IRDye 800CW carboxylate using MSOT. Five weeks after adriamycin or saline administration, anaesthetised (isoflurane) mice had hair removed from the abdominal region and were imaged in the inVision 256-TF MSOT imaging system (iThera Medical, Munich, Germany) using a multispectral protocol for 30 min (rate of 10 frames per second using wavelengths: 700, 730, 760, 775, 785, 800 and 850 nm, and averaging 20 consecutive frames to minimise influences of motion). Five min into the imaging mice received 200 μ l (20 nmol) IRDye 800CW carboxylate (LI-COR, USA) in 0.9% saline through a tail vein cannula over a period of 10 seconds (s). Data was reconstructed and multispectral processing performed to resolve signals for the IRDye, including gradient scaling for each animal at a time prior to the injection of the IRDye. Regions of interest (ROIs) drawn around renal cortex and the renal papilla-pelvis region (in short: pelvis) of the right kidney of each mouse were used to determine the time between the mean peak pixel intensity (T MAX ) in the cortex and the pelvis (T MAX delay). Exponential decay was fitted for mean cortex pixel intensities to determine the characteristic excretion half-life.
To depict the temporal dynamics of IRDye clearance in colour code, a Matlab routine was developed to process a stack of images and calculate a composite image on a pixel-by-pixel basis of T MAX , defined as the time (or frame number in the stack) at which the highest intensity was recorded at that pixel. A pseudocolor was applied to display this time (or frame number) per pixel as different colours on the spectrum colour scale. Image stacks were typically composed of 15 images. With a temporal resolution from in vivo imaging of approximately 16 s this equated to approximately 4 min of imaging data after injection that were used for temporal analysis. A C MAX image as a maximum intensity projection was computed of the same image stack. ImageJ was used to amplitude modulate the pseudocoloured T MAX image using the C MAX in order to dim areas with little contrast and therefore to reduce background noise in the T MAX caused by an absence of substantial temporal change in a certain pixel.
Statistical analyses. Linear mixed-effects (LME) models were fitted to characterise changes of albuminuria and of FITC-sinistrin half-life over time. The advantage of using a linear mixed-effects model is that the correlation between measurements across time points within mice is taken into account in a term called the random term. Differences in albuminuria and FITC-sinistrin half-life between the treatment and control group (e.g., differences in linear or quadratic changes over time) were subsequently tested using these models.
Multiple linear regression models were applied to test whether glomerular histological damage was associated with transcutaneous measurements of albuminuria (maximum observed values and week 4 values), FITC-sinistrin half-life (week 4 values), T MAX delay and excretion half-life of IRDye (the latter measured at 5 weeks).
Independent t-tests were applied to compare measures of kidney function at 4 weeks and of post-mortem histology measurements at 5 weeks between the control and the treatment group. The assumptions of normality and homogeneity of variances were checked.
Plasma Protein Binding of IRDye. A dye-protein stock solution was prepared by incubating of 2 ml IRDye (10 μ mol) with 8 ml Sprague Dawley rat plasma Li Heparin (Innovative Research, Novi, MI, USA) in phosphate buffered saline (PBS) at 37 °C, while 2 ml PBS was incubated with 8 ml rat plasma as control. Plasma protein-binding measurements were performed by equilibrium dialysis of PBS against dye-protein stock solution (or control stock solution) using a two-chamber dialysis set-up (Equilibrium Dialyzer, Havard Apparatus, Holliston, MA, USA) 45,46 . After 24 h the absorption of IRDye in PBS and plasma were determined in three independent measurements by absorption spectroscopy in a microplate reader (Tecan Infinite M200) and the concentrations of IRDye was calculated on the basis of the corresponding molar absorption coefficients ( Figure S4, Table S3).
Plasma protein binding (PPB) of IRDye in percent was determined by averaging three independent measurements and following the equation of Beer-Lambert law: where A is denoted as corresponding to UV absorption.