In vivo expression of VCAM1 precedes nephron loss following kidney tubular necrosis

Nephron loss is a key event during the onset and progression of chronic kidney disease, yet the mechanisms determining whether tubules undergo successful repair or progress to atrophy remain poorly understood. While fibrosis has been proposed to drive progressive organ damage, antifibrotic therapies have failed in clinical trials. Here, we reveal that tubular VCAM1-expression precedes nephron loss, fibrosis, and long-term kidney dysfunction. Using serial intravital microscopy in transgenic mice, we track tubulointerstitial remodeling between injured and intact tissue over 3 weeks. VCAM1 is rapidly induced in a distinct subset of injured tubules, preceding atrophy with sustained fibroblast recruitment. However, fibroblasts remain confined to injury sites and do not cause secondary damage in uninjured tubules. Finally, in human kidney transplant biopsies, tubular VCAM1 expression - but not KIM1 - correlates negatively with early and 12-month graft function, underscoring its potential as a biomarker of adverse outcomes. These findings position VCAM1 as an early indicator of tubular fate and nephron loss.

Methods
Study Approval

All experimental procedures involving animals were approved by the local authorities (Animal Experiments Inspectorate, Denmark, permit numbers: 2020-15-0201-00443 and 2024-15-0201-01839) and reported according to ARRIVE guidelines. Processing of human samples and data was approved by the National Committee on Health Research Ethics (M-20100269). Informed and written content was obtained from the participants. The trial was conducted according to the Helsinki Declaration.

Animals

A total of 60 (26 male and 34 female) mice with a mean weight of 22.3 ± 0.5 g and age of 12.6 ± 0.6 weeks (mean ± SEM) were included in the study (Tables S1 and 2). Transgenic mice were obtained from Jax Laboratory and bred in the animal facilities at the Department of Biomedicine, Aarhus University: PDGFRβCre-ERT2 – Salsa6F reporter mice were generated by crossing B6.Cg-Tg(PDGFrβ-cre/ERT2)6096Rha/J (Strain #:030201) 27 and B6(129S4)-Gt(ROSA)26Sortm1.1(CAG-tdTomato/GCaMP6f)Mdcah/J)] (Strain #: 031968) 28. After induction with tamoxifen, this strain identifies  PDGFRβcells by Salsa6F, a fusion protein of GCaMP6 and tdTomato. Additionally, CycB1-GFP reporter mice (Tg(Pgk1-Ccnb1/EGFP)1Aklo/J) (Strain #: 023345) 36 were used to detect renal cell proliferation through transient nuclear/cytosolic expression of  green fluorescent protein (GFP) expression stages S-, G2- and M of the cell cycle.

Mice were randomly allocated to one of three groups; i) in vivo imaging, ii) GFR and urine collection and iii) ex vivo imaging, as summarized in Table S2. Group i) was further divided into mice undergoing partial IRI surgery, sham surgery or laser injury, respectively. Groups ii) and iii) were divided into mice undergoing partial IRI surgery and sham surgery. Additionally, some mice from group i) were also used for ex vivo correlative microscopy upon sacrificing. Our study design precluded blinding of the experimental treatment, as the validation of surgical outcomes required close microscopic examination during the procedure. Serial intravital microscopy of partial IRI animals was challenged by strong tissue remodeling that sometimes caused missing data points from late acquisition time points. To compensate for missing data points, we included higher n-numbers in the partial IRI group. For statistical analysis all available serial imaging data was included. No data was excluded.

Surgery

Prior to surgery mice were accommodated in groups of 2 to 5 within ventilated cages (Technoplast model GM500 or GM9000) at 21 ± 2°C and 55 ± 10% relative humidity. Littermates were weaned at 3 weeks of age. Mice were maintained on a standardized 12-hour light-dark diurnal cycle, with unrestricted access to standard diet and water. The bedding material consisted of Aspen wood chips measuring 2-3 mm in size (Abedd, bedding midi), complemented by compressed cotton tubes for nesting purposes. Cages were enriched with Mouse Igloo with spinning wheels, wooden chew sticks, cardboard tubes, and Shepherd Shacks. Upon surgery, Mouse Igloo and cardboard tubes were replaced by cardbox houses to prevent the abdominal image window from getting stuck. Cages were routinely cleaned once a week. To investigate whether (myo)fibroblasts engage in progressive and self-perpetuating tubulointerstitial fibrosis, two individual surgery procedures were performed. First, an unilateral partial ischemia-reperfusion injury (IRI) was done by occluding one branch of the left renal artery for 21 min, as previously described 7. Second, a winged abdominal image window (wAIW) was placed above the postischemic and non-ischemic parts of the partial IRI kidney and implanted into the left flank of the mouse as described previously 29, 63. Thus, it was possible to image the border between the ischemic and non-ischemic regions.

During surgery mice received analgesia through intraperitoneal administration of buprenorphine (0.1 mg/ kg BW, Temgesic), and anesthetized with isoflurane (3.5% for induction, 1.5–1.75% for maintenance, 1.2-1.8 L/min flow rate, 50% oxygen in medical air). Surgery was conducted on a heating plate set to 37 ± 0.5°C, while eye ointment prevented cornea dehydration. The mouse’s left flank was shaved, and the skin was disinfected using chlorhexidine (0.5 % in 70% ethanol). Hereafter an incision into the left flank exposed the kidney. The kidney was gently repositioned, and the right artery branch was occluded with a clamp for 21 minutes. During occlusion, kidney temperature, and hydration were maintained with a nonwoven swab and occasional flushing with 37°C saline 29. Upon reperfusion, a purse-string suture connected the muscle layer and the skin around the incision, and the kidney glued to the wAIW using 50 μl cyanoacrylate glue. Lastly, the skin was secured around the groove of the wAIW by tightening of the purse-string suture 68. 10 μl/g BW sterile saline was injected i.p. for fluid supplementation. Following the image session at day 0, meloxicam (1 mg/kg BW meloxicam, Metacam) was administrated s.c. as postoperative analgesia as well as at days 1 and 2 post surgery. Additionally, mice had restricted access to normal drinking water but had to drink water containing buprenorphine (1 ml buprenorphine 0.3 mg/ml to 35 ml water) until 3 days post-surgery.

Image acquisition protocol

In vivo imaging was conducted using an upright Olympus FVMPE-RS 2-photon microscope (Olympus, Japan) operated with Fluoview FV31S software (Olympus, Japan), and an upright Ultima Investigator Plus multiphoton setup (Bruker Corporation, Billerica, MA, USA) with PraireView IV software as described previously 7.

The Olympus microscope was equipped with a MaiTaiHP DS-OL excitation laser (Spectra Physics, United States), a 25X XLPLN25xWMP2 objective, water immersion, (Olympus, Japan, NA 1.05; WD 2.00mm), with an IR cut filter of 690 nm, and the following detection cubes: Ch1: λet = 705/45 nm (multialkali PMT), Ch2: λet = 610/35 nm (multialkali PMT), Ch3: λet = 540/40 (GaAsP), Ch4: λet = 480/40 (GaAsP).

The Bruker microscope was equipped with a 20X Olympus XLUMPLFLN Objective, water immersion (Olympus, Japan, NA 1.00; WD 2.00mm), a 720 sp filter, and the following detection cubes: Ch1: λet=525/50 nm (GaAsP), Ch2: λet= 525/50 nm (GaAsP), Ch3: λet= 460/50 nm (GaAsP).

Prior to imaging, 1 μl/g body weight of a 2.5 mg/ml conjugated Albumin-Alexa Fluor 594 dye solution (Alexa594-albumin, Invitrogen) and 10 μl of propidium iodide (PI, 0.25 mg/ml, Thermo Fisher) was administered via retro-orbital injection. For imaging, the wAIW implant was slotted into a custom 3D-printed frame ensuring image stabilization on an upright microscope setup 64. During imaging, anesthesia and temperature were maintained with a low-flow (30–50 ml/min, 1.0–1.5%) isoflurane vaporizer (SomnoSuite, Kent Scientific), which was equipped with a heating blanket placed beneath the animal.

We identified ≈ 4 FOV, which were imaged using dual-track excitation using 750 nm and 940 nm, respectively (512 x 512, dwell time: 4 μs, line averaging x 2, Galvano laser). Z-series, with a step size of 1-2 μm and depth of ≈ 60 μm, were acquired from each FOV, starting from the capsule. Upon the end of the image session, mice recovered from anesthesia and placed back in their cage. This process was repeated for each imaging day as detailed in Table S2.

To induce selective laser-injury in a few nephrons within the otherwise healthy PDGFRβ-CreERT2-tdTomato kidneys (n = 4), we used the 2-photon laser as a micromanipulator and focused high laser power on a kidney region of approximately 60 um2. Tissue remodeling at the injury site was then monitored via serial in vivo imaging for up to 4 weeks as detailed in Tables S1 and 2. At some image sessions, a tile scan approach was utilized if the damaged area exceeded the normal FOV. These tiles were later stitched using the cell sense software (Olympus, Japan).

After the last imaging session, mice were anesthetized with isoflurane and perfusion-fixated with 4% PFA. For correlative microscopy the cortical renal area attached to the wAIW was separated as a ≈ 2 mm thick tissue section, post-fixated with 4% PFA for 1 hour, and then washed in PBS (3 x 5 min). Whereafter they were stained ex vivo.

Image analysis

During the analysis, all tubular segments from each FOV were sequentially numbered, re-identified on consecutive imaging time points, and qualitatively assessed for categorization into one of the following classifications: 'undamaged,' 'damaged,' 'recovered,' or 'atrophic'. Tubular segments were classified as 'damaged' if they exhibited any of the following criteria: positive staining for propidium iodide, dilated or collapsed tubule lumen, luminal granular cast accumulation, flattened tubule epithelium, visually decreased NADH autofluorescence and/or, in the case of PT-S1, visually reduced albumin endocytosis. 'Undamaged' tubule segments were defined by the absence of all characteristics mentioned above. Tubule segments were classified as 'atrophic' if they adopted a strongly collapsed, non-reversible state with decreased NADH autofluorescence. ‘Recovered’ tubules were identified with any of the damage characteristics outlined above and later adopted an ‘undamaged’ state. Additionally, atrophic’ tubule segments were further sub-classified as 'degraded' if they completely disappeared. The classification of individual tubule segments across consecutive imaging time points was performed and confirmed by two experimenters.

Denoising of images was conducted using the BM3D algorithms ^70 ^as previously described 69. In short, a denoising pipeline was created utilizing both FIJI 75 and MATLAB. FIJI facilitated batch data input/output, handled multichannel images, and merged denoised outputs. MATLAB executed BM3D algorithms, processing one channel at a time by reading temporary files written on disk by FIJI and producing the denoised data in separate temporary files.

Registration of serial imaging data was performed using a combination of a manual landmark-based volumetric registration within the FIJI plugin BigWarp using rigid rotation transforms 76, and intensity-based medical image registration using Elastix 38.

To quantify PDGFRβ-cell accumulation, two different approaches were utilized. In the first method, PDGFRβ-cells were segmented using Ilastik 77 (v5.1.0). Here, tdTomato-labeled PDGFRβ-cells were segmented from the registered 940 tracks using pixel classification. To train Ilastik, labels were assigned to Z-stack images to identify structures/cells as either tdTomato-positive or negative cells. Following exportation of the 2D Ilastik predictions, a threshold was applied using the Otsu method 78, whereafter the percentage of the of tdTomato cell area was assessed relative to the total FOV area (Figures S2B, C). This approach provided an unbiased quantification of PDGFRβ-cell abundance on a per FOV basis.

In the second method, the perimeter of each individual tubule was measured for each day equivalating 6903individual measurements. Afterward, it was measured how much of the tubule in question was in touch with surrounding PDGFRβ-cells and PDGFRβ-cell enclosure was expressed in percent of the total perimeter of the tubule (Figure 2E). This approach provided insights into PDGFRβ-cell dynamics with individual tubules.

Second harmonic generation (SHG) was quantified following image segmentation using Ilastik. The software was trained to recognize SHG signals from 940 tracks. Following the export of the 2D Ilastik predictions, a threshold was applied using the Otsu method, after which the percentage of the SHG area was evaluated relative to the total FOV area. All SHG analysis was done on decapsulated, ex vivo samples to avoid bias when separating interstitial SHG signal from SHG signal deriving from the kidney capsule. 

NADH signal quantification was only performed on images acquired on the Olympus system to maintain consistent detection optics for the quantification. In 750 nm excitation track data, a ROI was manually drawn around the epithelium of each tubule, and the average intensity of channel 4 was measured.