Can Photoacoustic Imaging Provide a Sensitive Marker for Early Detection of Diabetic Nephropathy

ABSTRACT

Diabetic nephropathy (DN) is the leading cause of chronic kidney disease. Albuminuria or glomerular filtration rate (GFR) test is time‐consuming with low sensitivity. Here, we used a photoacoustic imaging (PAI) system to evaluate the pathophysiological change of DN in mice models in vivo. The results showed that compared with the control group, the average blood oxygen saturation (sO₂ Avr) significantly elevated at early‐stage group, then gradually decreased at the advanced stage. The average hemoglobin concentration (HbT Avr) and lipid PA signal intensity dramatically drop down at advanced stage of DN. The AUC of sO₂ Avr for diagnosis of early stage DN was 0.847 (95% CI: 0.636–1.0). The AUCs of HbT Avr and lipid PA signal intensity for diagnosis of advanced‐stage DN were 0.917 (95% CI: 0.795–1.0), 0.875 (95% CI: 0.734–1.0). In conclusion, PAI was potentially useful for the early evaluation of DN.

This study used a photoacoustic imaging (PAI) system to evaluate the pathophysiological change of diabetic nephropathy (DN) in mice models in vivo. The results showed that the PAI marker sO₂ Avr was altered in early murine DN and may act as an early sensitive biomarker for DN detection.

1. Introduction

Diabetes mellitus (DM) is a metabolic disease characterized by chronic hyperglycemia, dyslipidemia, abnormal insulin secretion or resistance. Overall, the prevalence of DM globally is growing rapidly, especially in developing countries, and it is estimated that the number of patients will reach 783 million in 2045 [1, 2]. Diabetic nephropathy (DN) is one of the most frequent, silent and severe complications of DM and is associated with increased morbidity and mortality in diabetic patients [3, 4]. Hyperglycemia has been implicated in the onset and progression of endothelial dysfunction and microcirculatory disorders, which can further lead to structural and functional damage to the kidney [5, 6]. It is known as the leading cause of end‐stage renal disease worldwide. DN is characterized by persistent albuminuria (or albuminuria excretion rate of > 300 mg/d or 200 μg/min) measured at least twice within a 3 to 6 months interval, progressive decreasing in glomerular filtration rate (GFR) [7, 8]. Currently, screening and diagnosis of DN is still based on the albuminuria assessment [9]. However, some diabetic patients develop decreased renal function and vascular complications without proteinuria, known as nonproteinuric (NP)‐DN [10]. The prevalence of NP‐DN in type 2 DM ranges from 45% to 70%, while based on the latest data, its prevalence in type 1 DM is from 50% to 60% [11]. Early detection and intervention of DN is essential; however, effective, efficient and economic methods for this detection are currently lacking. Therefore, the development of noninvasive and reliable diagnostic approaches is crucial for the effective management of these patients.

Photoacoustic imaging (PAI) has afforded a very rapid high‐radiant‐power interlock system in the near‐infrared region passing through centimeters of tissue to generate ultrasound signals, which substantially surpass light‐scattering interference of tissue in the formation of high‐contrast multiple spatial images. Through multispectral unmixing algorithms, this system can pinpoint some regions of interest in the target tissues and allow multiple detections of signals at various wavelengths simultaneously [12, 13]. Based on optical absorption of oxyhemoglobin and deoxyhemoglobin in the blood, PAI can clearly visualize the microvascular structure information and blood oxygen saturation of tissue, enabling further evaluation of the functional status of biological tissues and metabolism [14]. Previous studies have reported that hyperglycemia in diabetes affects kidney blood supply [15]. Therefore, PAI has the potential to provide enhanced insights into renal microvascular complications associated with DN.

Although previous studies have validated the feasibility of PAI in the potential use of renal imaging of experimental animals [16, 17, 18, 19], a reliable as well as standard examination method of using PAI to aid the diagnosis of DN has not been established yet. In this study, we investigated the impact factors on PAI measurement of the kidney, including the impact of the bilateral sides of the kidneys, longitudinal and transverse sections, and examiners, providing data support for the future application of this technology in kidney evaluation. Furthermore, we investigated the average blood oxygen saturation (sO₂ Avr), the average hemoglobin concentration (HbT Avr), the lipid photoacoustic (PA) signal intensity change with DN progression and analyzed the diagnostic efficacy of these PA markers, which has not been reported previously through literature review.

2. Materials and Methods

The study was approved by the Institutional Animal Care and Use Committee (IACUC number 20240507002) at West China Hospital of Sichuan University. Given that male mice are more susceptible to diabetes than female mice, male C57BL/6 mice were chosen as the experimental subjects for this study. C57BL/6 mice (7–8 weeks, 20–25 g) were housed under controlled conditions (25°C, 12 h light/dark) with ad libitum access to food/water.

2.1. Photoacoustic Imaging System

During the imaging process, mice were anesthetized via inhalation (2%–4% isoflurane in oxygen) with vital signs monitored (respiratory rate: 80–100 beats per minute; heart rate: 400–450 beats per minute) and were maintained at 37°C in a supine position. MSOT used the Vevo LAZR‐X system (Fujifilm VisualSonics) integrating tunable lasers (680–970 nm and 1200–2000 nm) with a 40 MHz linear ultrasound transducer (MS550D). Flashlamp pumped Q‐switched Nd:YAG laser with optical parametric oscillator (OPO) for wavelength tuning with second harmonic generator was used to generate pulses with a repetition rate of 20 Hz. The energy supplied by each pulse of the tunable laser was 1.3 mJ/cm², well below the standard set by American National Standard Institute across the wavelength range. The following parameters remained consistent throughout the entire imaging process: B‐mode gain: 20 dB; PA gain: 40 dB; depth/width: 6.00/12.00 mm; pulse repetition rate: 20 Hz. Four to six scans per section were acquired, each comprising 10 frames. The value of photoacoustic signals per frame was calculated within the region of interest (ROI), and these were averaged to yield a value per scan. Since PAI is depth‐dependent, only the renal parenchyma located in the near field of the image was analyzed when delineating ROI. The size of ROI was consistent throughout the analysis process. Quantitative analysis of the photoacoustic signals was performed using VisualSonics VevoLAB software (Toronto, ON, Canada).

2.2. Experimental Procedure

2.2.1. PAI Methodology of Kidney in Healthy Mice

In the field of photoacoustic research, the average blood oxygen saturation (sO₂ Avr) is widely applied [20, 21, 22]. Therefore, in the research of photoacoustic methodology of the kidney, sO₂ Avr was selected to carry out relevant methodological exploration. sO₂ Avr was measured and quantified using the 750/850 nm (“Oxy‐Hemo mode” of the Vevo LAZR‐X system). Specifically, when in “Oxy‐Hemo mode”, the system switches between 750 nm and 850 nm laser pulses at a millisecond‐level rapid rate. The impact of left kidney and right kidney, longitudinal section and transverse section, and consistency of examiner (including the interclass consistency, the intraclass consistency of the same examiner at 24‐h intervals) were assessed. Two operators (Experienced Examiner 1, Novice Examiner 2) independently performed PAI acquisitions.

2.2.2. Establishment of STZ‐Induced T1DM Model and Photoacoustic Imaging

The T1DM model was induced by administering a single high‐dose streptozotocin injection (STZ, 180 mg/kg) [23]. After modeling, the fasting blood glucose (FBG) of the mice was monitored. Only when the mice exhibited a sustained hyperglycemia state with FBG ≥ 16.7 mmol/L for two consecutive weeks, they were included in the subsequent study. Finally, a total of 24 mice were included in the T1DM model group. The control group (n = 8) received an equal volume of sodium citrate buffer.

PAI was performed on the kidneys in both T1DM model and control groups. For in vivo imaging, PAI signals in the kidney were acquired at 750/850 nm (Oxy‐Hemo mode to obtain sO₂ Avr and the average hemoglobin concentration (HbT Avr)), and 750/850/930 nm (lipid), respectively [24]. sO₂ Avr represents the average blood oxygen saturation, which is based on the ratio of oxyhemoglobin to total hemoglobin. HbT Avr represents the average total hemoglobin concentration within the ROI, which is calculated from pixels exhibiting hemoglobin signals and comprises both oxygenated and deoxygenated hemoglobin. At 750/850/930 nm, the photoacoustic signal intensity can reflect the status of lipid deposition. PAI system and data processing was presented in Figure 1.

FIGURE 1.

Schematic diagram of photoacoustic imaging system and data processing. (A) Schematic of the imaging system. (B) Workflow of signal and image processing. HbT Avr, the average hemoglobin concentration; sO₂ Avr, the average blood oxygen saturation.

After the imaging was completed, the mice were euthanized by CO₂ asphyxiation. Kidney sampling was performed immediately after euthanasia. Then, the histology was examined after hematoxylin and eosin (H&E).

2.3. Data Analysis

Statistical analysis was performed using SPSS 24.0 software (IBM, Armonk, NY, USA) and GraphPad Software (San Diego, California USA). Descriptive statistics were presented as mean, standard deviation (SD), and coefficient of variation (CV). Paired‐samples t‐tests were conducted to assess the significance of factors influencing average oxygen saturation (sO₂ Avr) in healthy mice. When comparing three groups, the median (25th percentile to 75th percentile) was used for the continuous variables. The Kruskal–Wallis test was adopted, followed by Bonferroni's multiple comparison test for post hoc analyses if necessary. The areas under the receiver operating characteristic (ROC) curves (AUCs) with 95% confidence intervals (CIs) were used to assess diagnostic values for detecting renal injury. The cutoff value was determined by the highest Youden index. The corresponding sensitivity and specificity were calculated. All tests were two‐tailed, and a p‐value < 0.05 was considered significant.

3. Results

3.1. PAI Methodology of Kidney

The results of the PAI methodology were presented in Table 1. The impact of the left and right kidneys on sO₂ Avr measurement was analyzed. The results indicated that there was no significant difference between bilateral kidneys (p = 0.895). The left kidney had a smaller coefficient of variation (CV) (left 9% vs. right 11%). The results of the transverse and longitudinal section showed there was no statistical difference between the two sections (p = 0.639), with the longitudinal section having a smaller CV (longitudinal section 10% vs. transverse section 12%). The interclass consistency of sO₂ Avr measured by the different examiners was 0.733 (95% CI: 0.639–0.871), indicating a good consistency between the different examiners in kidney (p = 0.064). The intraclass consisitency by Examiner I of two measurements 24 h apart were 0.738 (95% CI: 0.645–0.878). The intraclass consisitency by Examiner II were 0.787 (95% CI: 0.615–0.871). For the two examiners, their respective two measurements all yielded no statistical difference, where the p‐values were 0.147 and 0.412.

TABLE 1.

The impact of photoacoustic measurement methods.

Variable	Median (%)	Mean (%)	Range (%)	SD	CV (%)	p
Bilateral kidneys
Left	66.44	62.28	52.86–72.05	6.26	9	0.895
Right	61.13	62.51	48.61–73.07	6.72	11
Different cross‐sections
Transverse	62.56	62.33	48.61–73.27	7.48	12	0.639
Longitudinal	65.89	62.69	53.58–70.53	6.02	10
Interclass consistency
Examiner I	64.83	64.22	52.86–73.27	5.97	9	0.064
Examiner II	63.51	63.13	48.61–72.05	6.56	10
Intraclass consistency
Examiner I	64.47	64.45	52.45–77.44	8.47	9	0.147
	63.47	63.01	53.58–74.98	9.58	10
Examiner II	66.45	64.04	52.78–74.48	7.85	10	0.412
	64.98	63.05	50.98–78.04	8.95	9

Abbreviations: CV, coefficient of variation; SD, standard deviation.

3.2. Pathological Examination of Kidney in STZ‐Induced T1DM Model

Figure 2 showed the renal pathological staining results. Normal glomerular structure and well‐arranged renal tubules were seen in the control group, without inflammatory cell infiltration. Based on the morphological changes of renal tubules, interstitium, and glomerulus, the T1DM model group was divided into the early‐stage group and advanced‐stage group. H&E staining of the early‐stage group showed necrosis and shedding of some renal tubular epithelial cells and an increase in inflammatory cells. In the advanced‐stage group, H&E staining showed extensive atrophy of renal tubules and glomerular sclerosis.

FIGURE 2.

H&E of the kidneys in the normal and model groups at the different stages. (A) control group; (B) early‐stage of T1DM model; (C) advanced‐stage of T1DM model. H&E, hematoxylin and eosin; T1DM, type 1 diabetes mellitus.

3.3. General Condition of STZ‐Induced T1DM Model

Compared with the control group, T1DM model presented typical diabetic symptoms, specifically manifested as obvious polydipsia and polyuria behaviors, accompanied by drowsiness and dull fur. Meanwhile, FBG level of the model mice remained at a high level consistently. With the progression of the disease course, the body weight of the advanced‐stage group decreased significantly compared with that of the control group and early‐stage group. In the advanced‐stage group, plasma lipid disorders occurred; the levels of total cholesterol, total triglycerides, and low‐density lipoprotein cholesterol (LDL‐C) changed compared with the control group. In addition, blood urea nitrogen (BUN) and blood creatinine (CREA) in the advanced‐stage group were significantly higher than those in the control group, suggesting that the renal function of the advanced‐stage group has been impaired (Table 2).

TABLE 2.

General condition of the control and T1DM model group.

	Body weight (g)	FBG (mmol/L)	TC (mmol/L)	TG (mmol/L)	LDL‐C (mmol/L)	BUN (mmol/L)	CREA (μmol/L)
Control group	25.83 ± 0.27	8.7 ± 1.67	2.32 ± 0.23	1.00 ± 0.60	0.49 ± 0.25	27.12 ± 3.14	14.33 ± 3.23
Early‐stage group	23.58 ± 1.25	24.47 ± 2.45*	2.58 ± 0.45	1.04 ± 0.56	0.51 ± 0.28	28.25 ± 7.28	15.04 ± 4.98
Advanced‐stage group	17.04 ± 2.33*, #	27.10 ± 4.25*	3.40 ± 0.54*, #	1.23 ± 0.52*	0.62 ± 0.30*	31.01 ± 10.07*	17.77 ± 4.87*

Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; BUN, blood urea nitrogen; CREA, creatinine; FBG, fasting blood glucose; LDL‐C, low‐density lipoprotein cholesterol; TC, total cholesterol; TG, triglycerides.

^# p < 0.05 vs. early‐stage group.

^* p < 0.05 vs. control group.

3.4. PAI of Kidney in STZ‐Induced T1DM Model

The results of the PAI Oxy‐Hemo mode were presented in Figure 3. At 750/850 nm, the vascular structure within the kidney was continuous. The quantitative analysis results revealed that compared with the control group, sO₂ Avr in the early‐stage group exhibited an upward trend. As DN progressed to the advanced‐stage, sO₂ Avr gradually decreased and eventually approached the level observed in the control group (control group vs. early‐stage group vs. advanced‐stage group = 61.11 (59.71–69.23) % vs. 71.63 (68.45–72.54) % vs. 67.26 (63.51–69.97) %). There was a significant difference in sO₂ Avr among the three groups. Post hoc tests further demonstrated that there was a significant difference in sO₂ Avr between the control group and the early‐stage group. However, no significant differences were found between the control group and the advanced‐stage group.

FIGURE 3.

Ultrasound and photoacoustic oxygenated hemoglobin, total hemoglobin imaging of mice kidney at 750/850 nm. (A) control group; (B) early‐stage of T1DM model; (C) advanced‐stage of T1DM model; (D) quantification of sO₂ Avr; E, quantification of HbT Avr. *p < 0.05, **p < 0.01, ***p < 0.001. HbT Avr, the average hemoglobin concentration; ns, not significant; sO₂ Avr, the average blood oxygen saturation; T1DM, type 1 diabetes mellitus.

Quantitative analysis results of HbT Avr showed that with the progression of DN, HbT Avr exhibited a gradual decreasing trend. Kruskal–Wallis test indicated there was a significant difference among the three groups (39 175 (37448–42 008) counts vs. 38 295 (38111–39 466) counts vs. 36 028 (35039–37 371) counts). Further post hoc analysis revealed that there were significant differences in HbT Avr between the control group and the advanced‐stage group, as well as between the early‐stage group and the advanced‐stage group. However, there was no significant difference between the control group and the early‐stage group.

Lipid PA signal intensity at 750/850/930 nm was presented in Figure 4. Quantitative analysis results showed the lipid photoacoustic signal intensity declined as DN progressed (1.16 (1.04–1.28) a.u. vs. 1.19 (1.07–1.39) a.u. vs. 0.95 (0.86–1.06) a.u.). Kruskal–Wallis test indicated there was a significant difference among the three groups. Further post hoc analysis revealed that there were significant differences between the control group and the advanced‐stage group, as well as between the early‐stage group and the advanced‐stage group. Similar to HbT Avr, there was no significant difference between the control group and the early‐stage group.

FIGURE 4.

Ultrasound and photoacoustic imaging of mice kidney at 750/850/930 nm. (A) control group; (B) early‐stage of T1DM model; (C) advanced‐stage of T1DM model; (D) quantification of lipid photoacoustic signal intensity at 750/850/930 nm. **p < 0.01, ***p < 0.001. ns, not significant.

3.5. Diagnostic Performance of Different Indicators for Renal Injury

The diagnostic efficacy of these PA markers for renal injury was evaluated as showed in Table 3. Among them, sO₂ Avr was used for the diagnosis of early renal injury, HbT Avr and lipid PA signal intensity were used for the diagnosis of advanced‐stage renal injury.

TABLE 3.

Diagnostic performance of different indicators for renal injury between the control group and those at different stages.

Renal injury stage	Parameter	AUC 95% CI	Cutoff value	Sensitivity (%)	Specificity (%)
Control vs. early renal injury	sO2 Avr	0.847 (0.636–0.976)	0.6685	88.9	75.0
Control vs. advanced‐stage renal injury	HbT Avr	0.917 (0.795–0.984)	38 565 counts	93.3	93.3
	Lipid	0.875 (0.734–0.978)	1.02 a.u.	73.3	87.5

Abbreviations: AUC, area under the receiver operating characteristics curve; CI, confidence interval.

As shown in Figure 5, sO₂ Avr exhibited favorable diagnostic efficacy for early renal injury. The AUC was 0.847 (95% CI: 0.636–1.0), with a cutoff value of 66.85%. At this cutoff value, sO₂ Avr yielded a sensitivity of 88.9% and specificity of 75%, suggesting that sO₂ Avr has good diagnostic discriminative ability for early‐stage renal injury.

FIGURE 5.

The receiver operating characteristic (ROC) curve of different indicators for renal injury. sO2 Avr for the diagnosis of early renal injury (A), HbT Avr and lipid photoacoustic imaging signal intensity for the diagnosis of advanced‐stage renal injury (B). HbT Avr, the average hemoglobin concentration; sO2 Avr, the average blood oxygen saturation.

HbT Avr showed excellent efficacy in the diagnosis of advanced‐stage renal injury. Its AUC value reached 0.917 (95% CI: 0.795–1.0), indicating high diagnostic accuracy; with a cutoff value of 38 565 counts, sensitivity of 93.3%, and specificity of 93.3%. The lipid photoacoustic signal intensity was also applicable for the diagnosis of advanced‐stage renal injury. Its AUC value was 0.875 (95% CI: 0.734–1.0), reflecting a high level of diagnostic accuracy, with a cutoff value of 1.02 a.u., sensitivity of 73.3%, and specificity of 87.5%.

4. Discussion

DN is a type of “silent”, chronic and multifactorial pathological change which includes hemodynamic abnormalities, metabolic disorders, and immune dysregulation [25]. Conventionally, diagnosis of DN mainly depends on albuminuria test and/or renal function assessment on two measurements with at least a 3‐month difference in the timing of measurements. However, this is a time‐consuming approach with low sensitivity and specificity, and the albuminuria test is not applicable for nonproteinuric (NP)‐DN [15, 26]. In this study, our first step was to investigate PAI methodology of kidney in healthy mice to ensure a standard measurement protocol. Compared to previously reported photoacoustic exam settings for monitoring kidney oxygen saturation [27], our results further optimize this imaging modality protocol. Although we found that the bilateral sides of the kidneys, longitudinal and transverse sections, and examiners were not significant impact factors for the PAI measurement of kidneys, the left kidney with longitudinal section was selected as a better screening target for the following experiment due to its relatively good measurement reproducibility. In addition, the inter‐ and intra‐class consistency of operators was also evaluated and indicated a low operator dependence for PAI exam of kidney.

Then we applied PAI to assess oxygen saturation, hemoglobin, and lipid change in the development and progression of DN. Interestingly, we found that sO₂ Avr alteration occurred in early‐stage of DN, which was prior to HbT Avr and lipid PA signal intensity. This may suggest that sO₂ Avr was a sensitive marker to reflect the development of DN. Pan et al. found that PAI combined with PA gold nanoparticle (GNP)‐based bioprobes was a faster, more sensitive, and accurate technique for the detection of renal injury, compared with conventional methods [28]. Another study used PAI and ultrasound localization microscopy (ULM) with microbubbles to obtain a mouse kidney microangiography with tissue oxygenation [19]. Also, PAI has been used for pre‐transplant kidney quality evaluation [17]. These emerging studies demonstrated the fact that PAI has a high potential for renal vasculature, function, and oxygenation assessment. Our study focused on DN‐related pathophysiological conditions by means of PAI. With reference to pathological results of DN, it is noted that sO₂ Avr significantly elevated at early‐stage group, while HbT Avr and lipid PA signal intensity dramatically drop down at advanced‐stage of DN. This phenomenon may be explained by the early buffer response of kidney vasculature to hyperglycemia‐induced oxidative stress and the persistent progressive tubulointerstitial fibrosis [29, 30]. Furthermore, we evaluated the diagnostic efficacy of the three PA markers, all of which exhibited good, even excellent diagnostic performance with optimal sensitivity and specificity for differentiation of various stages of DN. And these preliminary data provided a valuable prospect for the application of PAI in assessment of DN development and progression.

There were several limits of this study. First, the sample size of different stages of DN in this study is small, and further expansion of sample size is needed to improve the test power. Second, there are other influencing factors to be explored in the methodology section such as imaging depth. To translate this PAI technique into human applications, limited penetration depth represents a considerable obstacle. Human kidneys lie 3–7 cm below the abdominal surface, requiring a minimum penetration depth of approximately 4 cm for parenchymal evaluation [31]. Traditional preclinical PAI systems only achieve effective penetration < 2 cm, while currently available clinical photoacoustic devices attain around 3–4 cm depth [32, 33]. Further hardware innovation is necessary to satisfy the imaging demands of human diabetic nephropathy. Third, renal fibrosis condition of DN was not assessed as PA signal intensity of collagen fibers was not obtained. Fourth, this is a preliminary study to explore new sensitive PA markers for early DN detection based on pathologic stages of DN. The widely adopted diabetic mouse model was utilized in our research. Early renal injury in mice manifests as partial tubular epithelial necrosis and shedding, inflammatory infiltration, mild glomerular hypertrophy and initial tubular dysfunction, recapitulating early human DN with glomerular hyperfiltration and microvascular damage prior to microalbuminuria [34]. No prominent glomerulosclerosis or severe tubulointerstitial fibrosis can be observed at this stage in either species.

Advanced injury in mice exhibits progressive mesangial expansion, glomerulosclerosis, tubular atrophy and severe tubulointerstitial fibrosis, resembling overt human DN accompanied by sustained albuminuria, decreased eGFR and irreversible renal injury. Notably, diabetic mice rarely form Kimmelstiel–Wilson nodules typical of human DN, and fibrosis progression rates vary markedly across species [35]. Importantly, sO₂ Avr was altered in early murine DN before albuminuria. If analogous early microvascular hypoxia exists in pre‐proteinuric human DN, PAI may act as an early sensitive biomarker. Further human studies are warranted considering interspecies differences in disease course and renal photoacoustic characteristics.

5. Conclusion

Photoacoustic imaging was effective to reflect oxygenation, hemoglobin, and lipid change in diabetic nephropathy models of mice in vivo. Multi‐wavelength PAI of kidney was potentially useful for the evaluation of diabetic nephropathy.

Funding

This work was supported by Sichuan Province Science and Technology Program, 2025ZNSFSC1760, 2024YFFK0228 and Chengdu Municipal Science and Technology Program, 2024‐YF05‐00110‐SN.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.