Monitoring the Progression of Renal Fibrosis by T2-weighted Signal Intensity and Diffusion Weighted Magnetic Resonance Imaging in Cisplatin induced Rat Models

Abstract

Background:

Diffusion weighted imaging (DWI), with the applying of intravoxel incoherent motion model, has showed promising results in obtaining additional information about microperfusion and tubular flow associated with morphologic changes in chronic kidney diseases. The study aims to evaluate the potential of T2-weighted signal intensity (SI) and DWI with mono- and bi-exponential models to reflect the serial changes on cisplatin (CP) induced rat renal fibrosis models.

Methods:

Magnetic resonance exams were performed prior to and 2^nd day, 4^th day, 6^th day, 8^th day, 2^nd week, 3^rd week and 4^th week after CP injection at a 3.0T with an animal coil. Besides T2-weighted images (T2WI), DWI of 13 b values from 0 to 1500 s/mm² was acquired. Apparent diffusion coefficient (ADC), fluid fraction f, pure diffusivity D and pseudodiffusivity D* values were calculated. The regions of interest were placed on cortex (CO), outer stripe of the outer medulla (OM) and inner stripe of the outer medulla (OM), parameters were measured and compared among different time points. Five rats were scarified at each time point for pathological examination.

Results:

OM revealed remarkable hyperintense and broadened before it became an obscure thread, while CO demonstrated moderate hyperintense and IM didn’t show significant change on T2WI. On all three stripes, ADC values decreased firstly then kept increasing since the 4^th day; f values decreased on all stripes; D values had a tendency to increase with fluctuations but the changes didn’t achieve statistical significance; D* values increased at the 2^nd day then tended to be steady thereafter. Pathological findings revealed tubules epitheliums swelling followed by inflammation cells infiltration, interstitial fibrosis was observed since the 2^nd week.

Conclusions:

All of T2-weighted SI, ADC, and biexponential models parameters vary during fibrotic process; biexponential model is superior to monoexponential model in separating changes of microperfusion together with tubular flow from pure diffusion.

INTRODUCTION

Chronic kidney disease (CKD) is defined as a progressive loss in renal function over a time span of 3 months and longer. Approximately, 6.1 ~ 10% of the general population in China have CKD, with an even higher risk in people older than 40 years. CKD may be caused by a wide spectrum of diseases.[1] The common pathological consequences of CKD is renal fibrosis, assessment of the continuous deterioration of renal fibrosis is important.[2] Renal biopsy is unsurprisingly the golden standard to determine the degree of renal fibrosis in patients; however, it is an invasive method and inconformity for repeatedly monitoring the progression of fibrosis in clinical follow-up, and it remains challenging to establish a noninvasive biomarker to monitor the progression or regression of renal fibrosis in substitute of biopsy.

Magnetic resonance imaging (MRI) could provide both high spatial resolution and excellent tissue contrast in evaluating morphologic alteration in CKD. Moreover, diffusion weighted imaging (DWI) MRI has showed more promising results in assessing renal function.[3] The commonly used quantitative parameter to interpret DWI is apparent diffusion coefficient (ADC) value. ADC is calculated by fitting the signal intensities (SIs) from a series of DWI with different diffusion weightings (b values) using a monoexponential model. Previous reports revealed that the CKD patients have significantly lower ADC values in the cortex and medulla,[4] and found that ADC was able to assess the degree of renal fibrosis in unilateral ureteral obstruction murine models.[5]

In 1986, Le Bihan et al. introduced the principles of intravoxel incoherent motion (IVIM).[6] The concept leads to a more dedicated approach to enable tissue diffusivity and microperfusion to be quantitatively estimated separately. It could be speculated that additional information about microperfusion and tubular flow obtained with IVIM DWI may be especially helpful to evaluate the progress of renal fibrosis.[7]

The study aims to evaluate the potential of T2-weighted SI and DWI with mono-and bi-exponential models to reflect the serial changes of both morphology and function on cisplatin (CP) induced rodent renal fibrosis models with reference to the corresponding pathological observations.

METHODS

Animal model

This study was approved by Animal Care and Use Committee, Tianjin Medical University Hospital. A total of 45 male Sprague-Dawley rats (Beijing University Health Science Center) weighted 226 ± 25 g were used. Renal fibrosis was induced by intraperitoneal injection of CP solution (QILU Pharmaceutical Company Shandong, China, Batch Number:WA2A1210071) 4.5 mg kg^{− 1} day^{− 1} in 2 consecutive days. After injection, rats were marked and raised in clean plastic cages in a temperature- and humidity-controlled facility with a constant 12 h light/dark cycle. All rats showed intoxication symptoms including fatigued, loss of appetite, diarrhea, and weight losing especially during the 1^st week after CP injection. Nine rats died from serious diarrhea (n = 6) or anesthetic accident (n = 3), finally 36 rats enrolled in the present study. The rats were randomly selected to undergo MR scanning immediately before (n = 21) and 2 days (n = 24), 4 days (n = 24), 6 days (n = 23), 8 days (n = 18), 2 weeks (n = 11), 3 weeks (n = 8) and 4 weeks (n = 6) after CP injection.

Magnetic resonance imaging protocol

MR examination was performed on a 3.0T scanner (MR750, GE Company, US) using an 5 cm inner diameter animal coil (Magtron, China) under a stable anesthetized condition by intramuscular injection of 2% pentobarbital sodium solution (0.15 ml/100 g weight). Rats were fasting but free of drinking 2 h before MR exam and respiratory movement was restricted by proper compression on the abdominal region. Coronal fat suppressed fast recovery fast spin echo T2-weighted images were acquired with following parameters: Repetition time/echo time (TR/TE) 3182.6 ms/94.1 ms, matrix 352 × 352, field of view 6.0 cm, thickness 2.0 mm, spacing 0.5 mm, bandwidth ± 83.3 kHz, two signal averaged with a total acquisition time of 4 min 44 s. Axial multisection single shot-echo planar imaging (EPI) DWI was acquired with the following parameters: TR/TE 2075/111.9 ms; 10 sections coving both kidney regions, section thickness 2 mm, intersection spacing 0.5 mm, field of view 6.0 cm, matrix 192 × 96, bandwidth ± 100 kHz. The following 13 diffusion gradient b values were used: 0, 20, 40, 60, 80, 100, 140, 180, 240, 300, 500, 1000 and 1500 s/mm², the gradients were applied in three orthogonal directions and subsequently averaged to minimize the effects of diffusion anisotropy. The numbers of signal average gradually rose from 2 to 8 along with the increasing b value; the acquisition time was 2 min 45 s.

Image analysis

The MRI data series was transferred to an offline workstation (ADW4.5, GE Company, US). Both morphologic changes and functional changes were evaluated on the workstation singly by two radiologists with 15-year and 3-year experience in urogenital imaging respectively.

The morphologic changes of the renal parenchyma were evaluated on coronal T2WI. Besides visual inspection, the regions of interest (ROI) were placed free-hand on entire regions of renal cortex (CO), the outer stripe of the outer medulla (OM) and the inner stripe of the inner medulla (IM) at the middle coronal section of bilateral kidneys and average SI were measured. To reflect the signal changes on T2WI, T2-weighted standardized SIs were obtained by calculating the ratios between the average SIs of each of the three stripes and SI of homolateral psoas major.

Functional data of DWI were analyzed using Microsoft Data Access Components (MADC) software in the functiontools package installed in the workstation. Both monoexponential model (equation 1) and biexponential model (equation 2) were used on the same multiple b value DWI data series.[8]

S (b)/S (0)=exp(-b×ADC)      (1)

S (b)/S (0)=(1-f).e^–bD+f.e^–b.(D*+D)      (2)

S (0) is the SI in DWI for b = 0 (b₀). S (b) is the SI in a serial of DWI acquired by different b values. ADC maps were generated on renal parenchyma pixel-by-pixel by fitting the SIs with corresponding b values to a monoexponential decay model. Biexponential model yielded three parameters: f refers to the ratio of flow content in the entire volume, it reflects tissue microvasculature as well as tubular flow; D reflects pure diffusion coefficient and is sensitive to microstructural barriers, which limit slow water molecular movement; D* indicates pseudodiffusivity of fast water molecular movement which is sensitive to flow velocity.

After ADC, f, D and D* maps were generated, ROIs of entire CO, OM and IM stripes were plotted on the renal helium section according on the b = 0 images and transferred to the maps. The average value of ADC, f, D and D* of the three stripes were measured on bilateral kidneys respectively.

Histopathological examination

Five rats were randomly scarified after MRI examination accomplished at each time point after CP administration. Both kidneys were harvested for histopathological examination. The specimens were fixed in 5% buffered formalin and processed, then cut serially into 5 mm thickness section, and stained with hematoxylin and eosin pathological findings including tubules swelling and necrosis, inflammation cells infiltration in tubulointerstitium and the progression of renal fibrosis was assessed visually by an experienced pathologist.

Statistical analysis

Results are presented as mean ± standard deviation. The measurements of bilateral kidneys in each subject were analyzed for discrepancies by paired student t-test. The Intraclass correlation coefficient (ICC) was calculated to evaluate the reliability between the observers and the average values. The measured parameters on bilateral kidneys by the two observers were then averaged. The data were analyzed by the way of independent sample t-test among different time points. The discrepancies among each time point were statically analyzed on CO, OM and IM stripes respectively. For all statistical tests, a P < 0.05 was considered to be statistically significant. Statistical analyses were performed with Statistical Product and Service Solutions(SPSS) software 12.0.1 and Microsoft Office Excel 2003 workbook (Microsoft Company).

RESULTS

There was no significant difference in all the measurement between bilateral kidneys on all the three stripes (n = 675, P = 0.94). Interobserver agreement ICCs were ranged from 0.66 to 0.99 for OM, 0.73–0.99 for CO and 0.76–0.98 for IM.

Morphologic evaluation on magnetic resonance imaging

Observation of standardized signal intensity on renal fibrosis rats

Cortex (2.35 ± 0.40) and OM (2.35 ± 0.40) couldn’t be distinguished between each other, and IM showed relative hyperintense (3.13 ± 0.61) prior to CP injection. Two days after CP injection, OM began to appear as a hyperintense stripe (3.78 ± 0.72, P < 0.05) with distinct boundaries with CO and IM; then it gradually broadened and became less intense with transit downward fluctuation but beyond normal level during 4^th day and 4^th week (P < 0.05), consequently it shrink to be an obscure thread 4 weeks after CP injection. CO revealed slightly increased intensity (2.60 ± 0.41, P < 0.05) at the 2^nd days after injection, then decreased to normal level (2.28 ± 0.51, P > 0.05) 2 days later then appeared as steady hyperintense thereafter (P < 0.05). IM showed the transit and moderate hypointense (2.52 ± 0.72, P < 0.05) at only 4^th day [Figures 1 and 2].

Figure 1.

Serial morphologic changes of rat bilateral kidneys prior to and after CP injection on coronal fat suppressed fast recovery fast spin echo T2-weighted images. (a) Healthy kidneys; (b-h) 2^nd day, 4^th day, 6^th day, 8^th day, 2^nd week 3^rd and 4^th week after CP injection. On healthy kidneys prior to CP injection, CO and OM showed similar moderate signal intensity, and IM showed relative hyperintense. After CP injection, OM revealed a remarkable hyperintense stripe with distinct boundaries with CO and IM; it gradually broadened and became obscure and less hyperintense till the 4^th week after CP injection. CO revealed slightly increased intensity at the 2^nd days, then decreased to the normal level 2 days later and kept increasing thereafter. IM changed slightly during the process. CP: Cisplatin; OM: Outer stripe of outer medulla; IM: Inner stripe of outer medulla; CO: Renal cortex.

Figure 2.

Serial changes of SSI on three stripes of the renal parenchyma in rats prior to and at different time points after CP administration. CO: Renal cortex; CP: Cisplatin; SSI: Standardized signal intensity; OM: Outer stripe of the outer medulla; IM: Inner stripe of the outer medulla. SSI of OM was higher than the other two. *P < 0.05 compared with normal kidney; ^†P < 0.05 compared with the 2^nd day after kidney CP administration; ^‡P < 0.05 compared with the 4^th day after kidney CP administration; ^§P < 0.05 compared with the 6^th day after kidney CP administration; ^||P < 0.05 compared with the 8^th day after kidney CP administration; ^¶P < 0.05 compared with the 2^nd week after kidney CP administration; **P < 0.05 compared with the 3^nd week after kidney CP administration.

Functional evaluation on magnetic resonance imaging

Observation of apparent diffusion coefficient values on renal fibrosis rats

Mean ADC values of all the CO, OM and IM decreased firstly and then kept increasing since the 4^th day. The mean ADC values of OM were higher than normal subject (4.38 ± 0.63 × 10⁻⁴ mm²/s vs. 4.91 ± 0.56 × 10⁻⁴ mm²/s, P < 0.05), while the ADC values of CO were similar to normal level (5.7 ± 0.79 × 10⁻⁴ mm²/s, P > 0.05) since the 6^th day. The changes of ADC values for OM and CO tended to be stable during the 6^th and 4^th week (P > 0.05). In IM, excepted for the 4^th week (4.33 ± 0.93 × 10⁻⁴ mm²/s, P = 0.07), mean ADC values were significant lower than normal subjects (5.23 ± 1.05 × 10⁻⁴ mm²/s) at all time points (P < 0.05) [Figure 3a].

Figure 3.

Serial changes of mono-exponential and biexponential models parameters on renal parenchyma in rats prior to and at different time points after CP administration. (a) Serial changes of ADC value. ADC values decreased firstly then kept increasing since the 4^th day; (b) Serial changes of f fractions. The f values of all three stripes were below the normal level after CP injection, and the changes on OM were more remarkable; (c) Serial changes of D values. D values had a tendency to increase with downward fluctuation; (d) Serial changes of D* values. D* values of CO, OM and IM stripes increased at the 2^nd day, and then tended to be steady thereafter till the 3^rd week. CO: Renal cortex; CP: Cisplatin; ADC: Apparent diffusion coefficient; OM: Outer stripe of the outer medulla; IM: Inner stripe of the outer medulla. *P < 0.05 compared with normal kidney; ^†P < 0.05 compared with the 2^nd day after kidney CP administration; ^‡P < 0.05 compared with the 4^th day after kidney CP administration; ^§P < 0.05 compared with the 6^th day after kidney CP administration; ^||P < 0.05 compared with the 8^th day after kidney CP administration; ^¶P < 0.05 compared with the 2^nd week after kidney CP administration; **P < 0.05 compared with the 3^nd week after kidney CP administration.

Observation of bi-exponential models on renal fibrosis rats

The f values of all the CO, OM and IM stripes decreased after CP injection and were below normal level all the time (P < 0.05), and the changes in OM were more remarkable [Figure 3b]. The slow recovery of f values were seen as early as 4^th day in IM, then 6^th in CO, but f values in OM kept decreasing till 8^th day before going steadily upward (P < 0.05).

D values of CO, OM and IM stripes, on the contrary to f values, had a tendency to increase in spite of some downward fluctuations [Figure 3c]; however, changes on CO and OM were not significant at all time points when compared with healthy kidneys (CO: mean 0.31 ± 0.28 × 10⁻³ mm²/s; OM: 1.04 ± 0.80 × 10⁻³ mm²/s) (P > 0.05). D values of IM kept increasing till the 2^nd week as in Figure 3c, then turned back to normal level; however, the difference between each time point and healthy kidney (1.12 ± 1.51 × 10⁻³ mm²/s) wasn’t significant (P > 0.05), resulting from the standard deviations of D values on IM were large at all the time points.

D* values of CO, OM and IM stripes has a similar upward trend at the 2^nd day, and then tended to be steady thereafter till the 3^rd week [Figure 3d]. At all the time points except for 4^th week, D* values of three stripes were significant higher than healthy kidneys (P < 0.05). Mean D* values of IM (7.33 ± 3.99 × 10⁻² mm²/s) were higher than CO (2.50 ± 2.13 × 10⁻² mm²/s and OM (3.77 ± 2.47 × 10⁻² mm²/s) in healthy kidneys (P < 0.05).

Serial histopathology findings on renal fibrosis rats

At the 2^nd and 4^th day after CP injection, the epitheliums of renal tubules showed obviously cell swelling and tubules narrowing, and OM stripes which was consisting of proximal convoluted tubules demonstrated the most remarkable changes [refer to the acute stage, Figure 4a]. During the 6^th and 8^th day, cell swelling relieved while cell turbidity and renal tubular mold could be observed, massive inflammation cells began to infiltrate in the interstitium [refer to the chronic stage, Figure 4b and c]. Renal interstitial fibrosis was initially observed at 2^nd week; tubular atrophy and necrosis, broadened interstitial space with proliferation of myofibroblasts and cellular density increasing could be demonstrated throughout the fibrotic process after the 3^rd week [refer to the fibrosis stage, Figure 4d].

Figure 4.

Serial pathological changes of the renal parenchyma in rats after CP administration (H and E, ×200). (a) At the 2^nd day after CP injection, the epitheliums of renal tubules showed obviously cell swelling; (b) At the 6^th day, cell turbidity (yellow arrow) and renal tubular mold could be observed (orange arrow); (c) At the 8^th day, inflammation cells infiltrated in renal interstitium (arrow) resulting in cellularity increasing; (d) At the 4^th week, massive fibrotic collagen (arrow) within broadened renal interstitial space was demonstrated together with tubular atrophy. CP: Cisplatin.

DISCUSSION

From the findings of this longitudinal study, all of T2-weighted SI, ADC value, and biexponential parameters showed variations that agree with pathologic observations. Unlike human and swine kidneys, which are multilubular in anatomic structure, rodent kidneys are unilobular, this is ideal to delineating among different anatomic regions. Rodent kidney parenchyma is divided into four stripes: CO contains all glomeruli and their corresponding proximal convoluted tubules, OM is mainly occupied by the tortuous proximal convoluted tubules, IM is consisting of descending thin limbs and thick ascending limbs, and the inner medulla stripe is occupied by renal collecting ducts.[9] Rodent renal fibrosis models have been successfully induced by administration of CP in a large number of histopathologic studies. CP is a widely used antineoplastic agent, but associated with nephrotoxicity, and prone to result in long-term renal fibrosis after administration, the damages involve both glomeruli and tubules, while the major damage happens in proximal convoluted tubules.[10] The pathologic changes of inner medulla in CKD is minor, true but without shown in the present study, the inner medulla is isointense to renal pelvis which makes it difficult to be delineated; hence the analysis of inner medulla is neglected in the study.

The standardized T2-weighted SIs that are sensitive to tissue edema and necrosis were introduced in this study to evaluate morphological changes of renal fibrosis. The elevated T2-weighted SI, predominately seen on OM in the early stage implied the happening of acute cytotoxic edema and tubular necrosis that was proved histopathologically. However, T2-weighted SIs were then going in a steady state that was inconsistent with fibrotic progression. Similar observation was described by Hueper et al. on acute kidney injury mice,[11] suggesting T2-weighted SI could not stand alone by itself to be a biomarker of renal fibrosis.

Several investigators have reported decreased ADC in both cortex and medulla in CKD patients or renal fibrosis animal model compared with normal subjects.[5,12,13] The ADC values of cortex and medulla after CP administration were initially decreased as expected, however, both CO and OM had a tendency to increase and were higher than that of normal kidney since 6 days after CP injection, at times inflammation and fibrosis were observed. This phenomenon is controversial with most of the published data; the elevation of ADC values in diseased tissues was described in few reports with increased microperfusion.[14] The ADC values behaved as a weight of the pure diffusion and pseudodiffusion, pure diffusion restriction was mainly depending on the cellularity which was increasing during fibrogenic process,[5] hence the increasing of pseudodiffusion should be responsible for the ADC upward alteration.

Intravoxel incoherent motion is capable of separating pure diffusion and microperfusion theoretically, which is beneficial beyond simple ADC in oncology imaging and functional imaging. The model hypothesizes that the tissue signal decay in vivo might be divided into flow-dependent fast decay and flow independent slow decay, which makes it particularly meaningful in evaluating renal diseases.[15] The water movement within extracellular space is modeled as pure diffusivity when DWI are acquired with higher b value (over 600–800 s/mm²), while the microapillary perfusion and tubular flow are classified as pseudodiffusion, which is predominant with low b values (under 100–200 s/mm²). Recent studies demonstrated the capability of IVIM in differencing subtypes of renal tumors.[16] Zhang et al. showed that the biexponential parameters were more accurate than monoexponential ADC in healthy volunteers.[4]

Among of three parameters in biexponential model, f fraction of all stripes showed early decreasing and subsequent recovering process. F fraction referred to the proportion of fast flow contents in overall diffusion effect. Therefore, the decrease of f value indicated both the lessened capillary vasculature and narrowed renal tubular which is coincident with pathology. Meanwhile, the D values of CO, OM and IM had a tendency to increase after CP injection but failed to demonstrate significant alterations during the period of observation. The steady state of D values were also observed in some but disagree with the majority of published data, and it agreed with our previous study on unilateral ureteral obstruction rat model, which demonstrating increased D value along with time.[17] Pure diffusivity was also affected by a number of imaging factors including signal to noise ratios, registration artifacts, field strength, imaging sequence, and parameters, as well as the b value setting.[4,18] However, it wouldn’t interfere with a longitudinal study design like ours, the reason of D value increasing instead of decreasing during renal fibrosis was unclear and under investigation. D* values of all stripes increased after CP injection, and the mean D* values of IM were higher than the others. The pseudodiffusion D* of the renal parenchyma depends on both microperfusion and tubular flow, the elevated D* value suggested either increased microcirculation or faster renal tubular flow. Ebrahimi et al. suggested that D* mainly depended on the velocity of tubular fluid instead of microperfusion in the kidney by correlated IVIM parameters with computed tomography perfusion parameters on swine model.[19] According to the pathological observation, with the progressing of renal fibrosis, destroy, and atrophy of renal tubular was common seen, then tubular flow velocities would increase as compensating and repairing mechanisms, which were intermingled together. A higher D* in IM implied that there existed faster tubular fluid flow in descending and ascending limbs than proximal convoluted tubules, which was rational on the basis of renal physiology.

The study had some limitations. Because of intrinsic distortion of EPI, separately analysis of the stripes prone to error because of small ROI size; large variations might have prevented the achievement of statistical significance. The ridicules that makeup of proximal convoluted tubules extends from OM to CO had blurred the boundary between them, hence the arbitrarily ROI delineating between OM And CO leading the large variations of parameters. A more robust Gaussian distribution model and auto segmentation techniques might be helpful in further investigation. Although the present data rendering the feasibility that the microperfusion together with tubular dynamics might be evaluated using biexponential models, they couldn’t be separated from each other by IVIM method. Finally, the study is lack of evidence of specific pathologic finding for renal fibrosis such as cellularity counting and assessment of microperfusion and tubular dynamics alterations, a correlation study with IVIM parameters would be more convincing in future study.