1.

Introduction

Acute renal failure (ARF) occurs in 5 to 7% of hospitalized patients¹ and results in a mortality rate of about 50%.² A particularly disturbing feature has been the inability since the mid-1960s to substantially lower mortality rates in acute renal failure.³ The financial costs of acute renal failure are estimated to be 8 billion dollars per year, or about $130,000 per life-year saved.³ It is unlikely that this high mortality and associated cost will be reduced until we have better tools for the early diagnosis of renal injury. Accordingly, a critical need exists to develop clinically useful markers of early kidney injury and clinically efficient means for their detection. Several authors have pointed out the deficiencies of standard measures of kidney function, e.g., serum creatinine, for the early detection of acute renal failure.^{4, 5} Early detection of ARF is essential to reduce the incidence, progression, and associated morbidity and mortality of these common problems. Currently, point-of-care testing for ARF is limited due to the lack of sensitivity of common markers, such as serum creatinine, and the complexity and expense of measuring more sensitive and specific biomarkers. The lack of real-time detection capabilities for ARF biomarkers also severely impedes progress in clinical research, which relies on the early identification of affected individuals. The use of detection technologies in doctor offices or at the bedside will require a sensitive, selective, and portable sensor. Recent advances in the fields of nanotechnology have enabled the development of a new class of detection based on surface-enhanced Raman spectroscopy (SERS), which provides highly accurate, portable, real-time detection technology for urine biomarkers. Here, we present a highly sensitive and robust SERS substrate for urine analysis.

Urine analysis is one of the most commonly used approaches to receive information of renal function and metabolic condition of the body. The feasibility of Raman spectroscopy for urine analysis has been investigated by several researchers. Dou ⁴ and McMurdy and Berger⁵ studied the quantitative Raman analysis of creatinine in human urine. Premasiri ⁶ studied the potential of the SERS method for analyzing creatinine in urine using gold colloids as the SERS active substrate. Wang reported the semiquantitative measurement of creatinine concentration in human urine samples also using gold colloids as the SERS active substrate.⁷ Recently, urine samples have been analyzed on a SERS-based microfluidics device⁸ using an array of polystyrene nanospheres as the SERS substrate. Furthermore, Qi and Berger⁹ described the quantitative measurement of creatinine by Raman spectroscopy and optical fiber technique. Most of the SERS analyses of urine have been done on a SERS active metal colloid-based system. SERS has remarkable analytical sensitivity, but the difficulty in preparing robust and uniform SERS substrates with surface morphologies that can deliver maximum SERS enhancement with high reproducibility is a major problem for detection.¹⁰

We demonstrate the use of a nanostructured SERS substrate in urine analysis of healthy and diabetic patients. We showed that the SERS spectra of urine samples are highly reproducible, resistant to differences in ionic strength, and the concentration of chemical components can be measured quantitatively with high sensitivity (i.e., creatinine concentration as low as $0.5\mu \mathrm{g}∕\mathrm{mL}$ ). Preparation of these SERS active substrates are inexpensive and easy to prepare.^{11, 12, 13} The creatinine levels of urine samples from diabetic patients and healthy persons were measured quantitatively using metalized nanostructured parylene substrates. As a result, we showed the suitability of using our SERS substrates in urine analysis.

2.

Results and Discussion

We recently demonstrated that nanostructured poly(chloro-p-xylylene) (PPX-Cl, also known as parylene) films can be fabricated by the oblique angle polymerization (OAP) method.^{11, 12, 13} The nanostructured PPX films are deposited on a substrate from a directional vapor source. Vapor deposition and polymerization at an oblique angle relative to the substrate surface permits the fabrication of films possessing nanostructured morphology. The nanostructured PPX film comprises free-standing, slanted, parallel columns containing nanowires. Subsequently, we deposited a thin layer (i.e., $\sim 60\mathrm{nm}$ ) of Ag metal film onto the nanostructured PPX template. Figure 1 shows the metalized nanostructured PPX film coated on a glass slide. Figure 1 shows the schematics of nanostructured PPX film metallization. Scanning electron microscope (SEM) images show the nanostructured morphology of the metalized PPX film at two different magnifications [Figs. 1 and 1]. Ag nanoparticles $(\sim 60\mathrm{nm})$ cover the nanostructured PPX films uniformly. The SERS substrates have high uniformity ( $\sim 5\mathrm{\%}$ signal variation in a $1{\mathrm{mm}}^2$ area) and sample-to-sample reproducibility.¹⁴ The enhancement factor (EF) was calculated as ${10}^5$ for the Ag metalized substrates.^{14, 15}

Fig. 1

(a) Surface-enhanced Raman substrate and (b) schematic of SERS substrate preparation are shown. The surface area of the SERS-active region is approximately $1{\mathrm{cm}}^2$ . SEM image of the Ag-PPX-Cl SERS substrate at (c) low magnification $(5500\times )$ and (d) high magnification $(20,000\times )$ . Scale bar for SEM images is $1\mu \mathrm{m}$ .

Raman spectra of pure creatinine, artificial urine, and urine from a diabetic patient were measured on the nanostructured PPX-Cl substrate. Significant peaks at 700, 840, 900, and $1420{\mathrm{cm}}^{-1}$ were observed in the Raman spectra of creatinine [Fig. 2 ]. However, in urine measurements, we found that 840- and $900\text{-}{\mathrm{cm}}^{-1}$ peaks are stronger than the $700\text{-}{\mathrm{cm}}^{-1}$ peak.

Fig. 2

(a) SERS spectra of: a creatinine solution $\left(0.5\mathrm{mM}\right)$ (top), artificial urine solution composed of urea, creatinine, and uric acid (middle), and a urine sample (bottom). Artificial urine concentrations are 1639, 104, and $34\mathrm{mg}∕\mathrm{dl}$ , respectively, which is approximately the same level of healthy human urine.¹⁸ All SERS measurements are colleted at $10\text{-}\mathrm{s}$ acquisition time with four scans. (b) SERS spectra taken with different amounts of KCl added to a urine sample. Concentrations of KCl (from top to bottom) are $0\mathrm{M}$ , $0.5\mathrm{mM}$ , $5\mathrm{mM}$ , $50\mathrm{mM}$ , and $0.5\mathrm{M}$ , respectively ( $10\text{-}\mathrm{s}$ acquisition time and four scans).

A significant advantage of our nanostructured substrate over traditional metal colloids is the enhanced stability of Raman signals when detecting samples with different ionic strengths. The variation of the ionic strength of the analyte can affect the aggregation process of metal colloids. It has been reported that the reproducibility of Raman signal drops drastically if the salt content of the urine varies.⁸ In contrast, our nanostructured SERS substrates show superior stability due to the immobilized metal nanoparticles on the surface of the SERS surface. Figure 2 shows robustness of SERS signal with various potassium chloride (KCl) concentrations. The urea peak at $1000{\mathrm{cm}}^{-1}$ shows an intensity variation of less than 5%. Each spectrum is shifted for clarity. A concentration profile for varying creatinine concentration in artificial urine is measured.

The reproducibility of our SERS substrate is demonstrated using $1\text{to}3\mu \mathrm{l}$ of urine sample from a diabetic patient. 12 spots were chosen randomly on the surface of the nanostructured substrate to collect Raman spectra [Figure 3 ]. From spot to spot, the urea signature peak at $1000{\mathrm{cm}}^{-1}$ (symmetrical C-N stretch) shows an intensity variation of less than 5%. This major peak can be used as an internal reference for urine analysis, considering the fact that urea is the dominant organic component of human urine, and the excretion rate of urea in urine is relatively stable. Figure 3 shows SERS spectra of creatinine at 0.5-, 5.1-, and $10.2\text{-}\mu \mathrm{g}∕\mathrm{mL}$ concentrations, respectively. SERS intensity at 840 and $900{\mathrm{cm}}^{-1}$ are measured from artificial urine samples. These results also show that creatinine concentrations as low as $0.5\mu \mathrm{g}∕\mathrm{mL}$ can be detected easily by SERS method using our nanostructured substrate. Mass spectroscopy technique, commonly used in clinical chemistry, can detect creatinine concentrations of $0.5\mu \mathrm{g}∕\mathrm{mL}$ , with an uncertainty of approximately 3%.¹⁶ Our substrate provides a similar sensitivity and reproducibility; hence it may be used as a quantitative technique in clinical measurements in the future.

Fig. 3

(a) Reproducibility of SERS spectra of a urine sample. (b) SERS spectra from artificial urine samples. Concentrations of creatinine (from bottom to top) are 0.5, 5.1, and $10.2\mu \mathrm{g}∕\mathrm{mL}$ respectively.

Table 1 provides details of Raman measurement parameters and creatinine concentrations measured by enzymatic assays. The creatinine concentrations of clinical urine samples are measured by Diazyme’s enzymatic creatinine reagent kit. This assay is known to have no interference from ascorbic acid, bilirubin, hemoglobin, or triglycerol. In this assay, creatinine is converted into creatine, and then transferred into sarcosine. The sarcosine is oxidized to hydrogen peroxide, which is measured by a Trinder reaction.

Table 1

Creatinine concentration of urine samples from diabetic patients and healthy samples, and Raman data collection parameters are shown.

Figure 4 shows the quantitative measurement of creatinine from diabetic and healthy patients. Two peaks at 840 and $900{\mathrm{cm}}^{-1}$ are chosen in the SERS spectra for the quantitative analysis of creatinine. Creatinine has two major peaks at 840 and $900{\mathrm{cm}}^{-1}$ , but there is spectral overlap from other components of urine in the same region (e.g., urea and creatinine at 840 and $900{\mathrm{cm}}^{-1}$ ). Therefore, the SERS data are analyzed by an ordinary least-squares method to generate linear predictive models of creatinine concentration of urine samples. There is a linear relationship between creatinine concentrations and SERS peak areas at 840 and $900{\mathrm{cm}}^{-1}$ . Percentage of variance $R^2$ equals 0.907 and 0.967, for 840 and $900{\mathrm{cm}}^{-1}$ data, respectively. The linear model is also analyzed with residual plots,¹⁷ and the $p$ -values were calculated by the Anderson-Darling test. We conclude that the residuals are normally distributed, since the $p$ -values (0.514 and 0.245 for SERS data at 840 and $900{\mathrm{cm}}^{-1}$ , respectively) are greater than 0.05 (level of significance). Figure 4 shows the creatinine concentrations measured by SERS and the enzymatic method. The SERS data are in good agreement with the enzymatic data. The correlation coefficients are 0.907 (based on the SERS peak at $840{\mathrm{cm}}^{-1}$ ) and 0.968 (based on the SERS peak at $900{\mathrm{cm}}^{-1}$ ), respectively.

Fig. 4

(a) Creatinine concentration of 13 clinical urine samples and corresponding SERS peak area at 900 and at $840{\mathrm{cm}}^{-1}$ are plotted. (b) Creatinine concentration of 13 clinical urine samples from SERS and as enzyme-based method are plotted. Open and closed circles denote SERS peak at $840{\mathrm{cm}}^{-1}$ (correlation coefficient equals 0.907) and at $900{\mathrm{cm}}^{-1}$ (correlation coefficient equals to 0.968), respectively.

3.

Experimental

4.

Conclusion

We use SERS to quantitatively measure the creatinine level of urine samples from diabetic patients and healthy persons using our metalized nanostructure parylene (PPX-Cl) film as a SERS substrate. The reproducibility and robustness of our SERS substrate are demonstrated at various ion concentrations. The advantage of these new types of substrates is that no template or lithography is involved, thus providing a simple, inexpensive, and quick production method to achieve highly sensitive and spatially uniform signals. We demonstrate a highly sensitive (i.e., concentration as low as $0.5\mu \mathrm{g}∕\mathrm{mL}$ ) and reproducible (i.e., signal variation less than 5%) substrate that can be used in clinical measurements. Rapid data analyses (i.e., $10\text{-}\mathrm{s}$ integration time) at low laser power (i.e., $2.5\text{-}\mathrm{mW}$ laser at $632\mathrm{nm}$ ) can be performed to determine creatinine concentration in urine. These advantages may improve the feasibility of using SERS as a fast, effective, and inexpensive tool for urine analysis in hospitals. We should also note that these data were collected at $2.5\text{-}\mathrm{mW}$ laser power and $10\text{-}\mathrm{s}$ integration time, which can be preferable for integration of these substrates into a low-power handheld diagnostic device.

Our method is also applicable to blood creatinine level measurements. The normal creatinine level in blood is about $0.5\text{to}1.2\mathrm{mg}∕\mathrm{dl}$ ,¹⁶ which can be detected by our SERS substrate. However, a preseparation (e.g., centrifugation and filtration) may be required to remove the interfering constituents. We will focus on the feasibility of selective biomarkers and metabolites in blood in the future.