Development of Non-proximate Probe Electrospray Ionization for Real-Time Analysis of Living Animal

Abstract

Ambient ionization mass spectrometry is one of the most challenging analytical tools in the field of biomedical research. We previously demonstrated that probe electrospray ionization mass spectrometry (PESI-MS) could potentially be used in the rapid diagnosis of cancer. Although this technique does not require a tedious sample pretreatment process, it was not possible for our previously reported setup to be applied to cases involving the direct sampling of tissues from living animal and large animal subjects, because there would not be enough room to accommodate the larger bodies juxtaposed to the ion inlet. To make PESI-MS more applicable for the real-time analysis of living animals, a long auxiliary ion sampling tube has been connected to the ion inlet of the mass spectrometer to allow for the collection of ions and charged droplets from the PESI source (hereafter, referred to as non-proximate PESI). Furthermore, an additional ion sampling tube was connected to a small diaphragm pump to increase the uptake rate of air carrying the ions and charged droplets to the ion inlet. This modification allows for the extended ion sampling orifice to be positioned closer to the specimens, even when they are too large to be placed inside the ionization chamber. In this study, we have demonstrated the use of non-proximate PESI-MS for the real-time analysis for biological molecules and pharmacokinetic parameters from living animals.

INTRODUCTION

The use of mass spectrometry (MS) in a clinical setting has increased significantly during the last decade, and the requirement for ambient ionization is growing rapidly. Ambient ionization is a MS technique that allows for the direct analysis of biological material from untreated samples and living animals. Representative examples of this technique include electrospray ionization (ESI),¹⁾ atmospheric pressure chemical ionization (APCI),²⁾ direct analysis in real time (DART),³⁾ rapid evaporative ionization mass spectrometry (REIMS),⁴⁾ desorption electrospray ionization (DESI),⁵⁾ and all five of these techniques have been applied to metabolomic and pharmacokinetics analyses, as well as medical diagnosis. Of note, Takáts’ group has demonstrated the applicability of REIMS based intelligent knife (iKnife) in operating theatre.⁶⁾

Probe electrospray ionization (PESI) is a variant of ESI that uses a solid needle, which behaves as an ESI emitter as well as a sampling probe.^7,8) In contrast to the capillary used in the conventional ESI, the solid metal probe used in PESI captures sample droplets on its surface that are picoliter in volume.⁹⁾ Furthermore, it is not possible for the solid needle used in PESI to become clogged-up. A high voltage (H.V.) is applied to the probe upon completion of the sampling process and electrospray ionization is initiated. The sampling and ionization processes are performed automatically by actuating the needle at a repetitive rate of 2–3 Hz. Notably, most of the captured samples are ionized and cleared away from the tip of needle during this process. Although some degree of cross-contamination between consecutive samples is not entirely avoidable, it may not be critical to the success of the analysis. Furthermore, the results of our previous studies have demonstrated that PESI is only mildly invasive as an analytical technique¹⁰⁾ and does not require the samples to be subjected to any special pretreatment processes.^10–12) This method was also found to be applicable to the analysis of biological samples containing high concentrations of salts¹³⁾ and detergents.¹⁴⁾ These properties therefore suggest that it would be possible to use probe electrospray ionization-mass spectrometry (PESI-MS) for the real-time analysis of molecular dynamics in living animals, as well as other living systems.¹⁵⁾

We have also previously demonstrated the applicability of PESI-MS as an analytical technique for the real-time analysis of living mice, where it was used to diagnose steatosis¹⁰⁾ and hepatocellular carcinoma.¹²⁾ Although the efficient uptake of ions into the mass spectrometer requires the ionization unit to be in close proximity to the ion inlet, especially in ambient MS, this requirement is largely dependent on the size of samples being placed within the MS system. Unfortunately, however, most commonly used mass spectrometers do not meet these criteria, and it was therefore necessary to modify our existing MS equipment during the initial stages of research by extending the sampling inlet using a long auxiliary tube made of stainless steel. This tube was connected to the original ion sampling inlet using a custom-made connector. As previously reported, appropriate gas flow can assist the ion transfer.^16,17) Then, to facilitate the effective uptake of ions into the original ion inlet of the mass spectrometer, it was necessary to attach a bespoke adaptor to an additional diaphragm pump. This modification made it possible for the ion sampling orifice to be placed immediately next to large living animals, which would otherwise be too large to be placed inside the ionization chamber of a conventional MS system. Based on the unique features of this system, we coined the term “non-proximate PESI” for this system. This system has successfully allowed us to directly monitor the kinetics of drug delivery and the metabolism of drug compounds in living animals on a real-time basis. Herein, we report the use of this newly developed non-proximate PESI system for the real-time in vivo analysis of large animals.

EXPERIMENTAL

Materials, animals, and sample preparation

Polypropylene Glycol, Triol Type (PPGT) mixed solution (PESI-MS standard solution) was kindly provided by Shimadzu Corporation (Kyoto, Japan). Eight-week-old male mice (C57BL/6J) were deeply anesthetized with 50 mg/kg pentobarbital sodium salt and subjected to laparotomy to expose their tissues. The liver, kidney, spleen, and brain of the animals were extirpated and washed with phosphate buffered serine (PBS). Specimens (ca. 3–4 mm²) punched out from each tissue and placed in plastic sample plates for their immediate analysis. For whole body analysis, anesthetized mice were held supine on a sample stage following surgery to perform the real-time analysis. If the surfaces of the tissues became dry during analysis, they were wetted with 2-μL portions of H₂O, which were supplied ad libitum via a micropipette. For the pharmacokinetics analysis, 5-fluoro-2′-deoxyuridine (5-FdU; SIGMA, St. Louis, MO, USA) was dissolved in Hank’s balanced salt solution (HBSS; Life Technologies, Carlsbad, CA, USA), and 100 μL of the resulting 100 mM 5-FdU solution was administered by a single intravenous injection to the tail vein of the mice. To evaluate the ion suppression effect, blood was collected from the left ventricle and treated with 1 unit/mL of heparin sodium (Mochida Pharmaceutical Company, Tokyo, Japan) to inhibit coagulation. All of the procedures conducted in the current study were performed in accordance with the policy of the University of Yamanashi Animal Care and Use Committee.

Probe electrospray ion source

A disposable acupuncture needle (J type No. 01, Seirin, Shizuoka, Japan) with a tip radius of less than 1 μm was used as both the sampling probe and an ESI emitter. The needle was fixed to a stainless steel tube that could be readily attached to the needle holder of the linear actuator with a precision crimper. The needle was moved up and down along a vertical axis using a linear motor actuator (Dyadic Systems, Kanazawa, Japan). When it was in the “down” position, the tip of the needle touched the surface of the sample. When it was in the “up” position, a H.V. in the range of 1.0–2.0 kV was applied to the needle to allow for the electrospray ionization of the sample attached to the tip of the needle. The motion of the needle and the application of H.V. was master-controlled using an in-house program, which was installed and operated from a personal computer (PC) (Fig. 1).

Fig. 1. (A) Panel showing the ion sampling tube and T-joint. The ion sampling tube is connected to the ion inlet of the mass spectrometer by custom-made metallic T-joint. (B) Schematic drawing of the non-proximate PESI-MS system. This system uses a stainless needle as a probe (right upper photograph). The panel on the right lower corner shows a scanning electron micrograph of the needle tip (bar=1 μm). When the needle is at its lowest position (Down; sampling phase), its tip touches the surface of the sample. At its highest position (Up; ionization phase), an electrospray plume is generated by application of HV. The charged droplets and ions are sucked through the ion sampling tube using the airflow generated by the vacuum pump. (C) Overview of the non-proximate PESI-MS system. Abbreviation: EVAC, evacuation; HV, high voltage; PC, personal computer.

Remote ion sampling

For remote ion sampling, an extended ion sampling tube (stainless steel) fitted with a caliber of 2 mm in diameter was installed at the ion inlet module of a single quadrupole mass spectrometer (LCMS-2020; Shimadzu Corporation, Kyoto, Japan) to collect the ions and charged droplets generated by the PESI, which was positioned 200 mm away from the MS. A custom-made metallic T-joint was used to mount the long ion sampling tube (Fig. 1A) (i.d.=1/8 inch and i.d.=2 mm, straight cut steel use stainless (SUS) tube from GL Science, Tokyo, Japan) to the mass spectrometer (Fig. 1B). The original air suction rate of the LCMS-2020 system was set at about 1 L/min. A small diaphragm pump (DAP-12S, Ulvac, Kanagawa, Japan) was connected to the ion inlet module to allow for the efficient uptake of charged droplets and ions directly into the ion inlet.

Statistical analysis

Mass spectra (m/z: 10–1200; Δm/z=1) from each tissue were plotted with the relative intensity (0–100%) or normalized intensity as a function of their m/z values. The normalized intensity was obtained by dividing the absolute intensity at each m/z value by the median intensity of the entire spectrum, and the resulting value was used for the following statistical analyses. Welch’s two-sample t-test was performed on each m/z value to allow for the spectra of living mouse livers to be compared with those of extirpated livers. The partial least squares (PLS) technique was used for supervised dimensionality reduction. Based on a data matrix X composed of n rows (observations) and p columns (m/z), and an n×m response matrix Y, the PLS algorithm can be used to calculate the d number (d≤p) of latent variables of X (PLS scores) as a product of the iterative decomposition of X and Y.¹⁸⁾ To allow the PLS technique to be used for classification purposes, we coded Y according to the following dummy matrix:   

where m is the number of classes, n_k is the number of observations in the k-th class and n=∑_kn_k. Spectral data for the extirpated liver, kidney, spleen, and brain were transformed into PLS scores. The first two PLS scores were used for plotting and the first four PLS scores were subjected to logistic regression (LR), which uses a one-vs.-all strategy for multi-class discrimination.¹⁹⁾ The following regularization function was used for LR:   

where w_k is the k-th weight vector. Parameters α∈{0, 1} and C∈{0.1, 1, 10, 100, 1000} were optimized by grid search using a 3-fold cross-validation process. Probabilistic classification of the spectra was conducted using a two-step method involving a combination of the PLS and LR methods (PLS-LR), and the predictive accuracy of this model was evaluated by leave-one-out cross validation (LOOCV). The PLS-LR algorithm was implemented in Version 2.7.6 of the Python software (Python Software Foundation, Wolfeboro Falls, NH, USA) using the Scikit-learn library.²⁰⁾ All of the other analyses were conducted in MATLAB (R2014a, The MathWorks, Inc., Natick, MA, USA).

RESULTS AND DISCUSSION

Performance evaluation of non-proximate PESI

It was not possible to separate the ionization chamber module of the LCMS-2020 system from the body of mass spectrometer, and it was therefore not possible to place large samples such as whole living animals into the chamber. However, the use of an ion sampling tube and sampling stage allowed for the PESI module to be placed in close proximity to the original ion sampling orifice of the mass spectrometer (Fig. 1C). This unique modification therefore allowed us to overcome the limitations imposed by the original design of the LCMS-2020 system and analyze much larger living samples. To evaluate the detectability of ions by the non-proximate PESI system for detecting ions, we initially used this system to analyze a PPGT mixed solution. However, the overall ion intensity of the sample analyzed using the non-proximate PESI system was significantly weaker than that of the original PESI system (Fig. 2A). It was envisaged that the flow rate of the air driven by the intrinsic vacuum system of the mass spectrometer, which was about 1 L/min, was too low to carry the charged droplets and ions over the long distance to the ion inlet. To address this issue, we employed an additional vacuum pump, which successfully recovered the ion intensity by 50% to that of the original PESI system (Fig. 2B). The optimal evacuation rate of the auxiliary pump was evaluated and determined to be 6 L/min for the 200-mm long ion sampling tube. The optimal voltage applied to the probe needle was also investigated and found to be 1.5–1.6 kV (Fig. 2B). Using these optimized parameters, it was possible for this system to detect a variety of different biological molecules directly from several different mouse tissues (Fig. 2C). Furthermore, the ion signals were strong enough to allow for the different tissue types to be readily discriminated from each other using the PLS-LR classifier (Fig. 2D and Table 1). The spectral data acquired from the liver, kidney, spleen, and brain samples of the mice were subjected to LOOCV analysis to evaluate the diagnostic potential of non-proximate PESI-MS combined with PLS-LR. Although we only had access to a limited amount of training data, this system correctly discriminated 41 out of 42 spectra (97.6%), and therefore highlighted the potential applicability of this system to the analysis of biological samples (Table 1).

Fig. 2. (A) Mass spectra (black line) and total ion chromatogram (TIC, blue line) from the analysis of the PPGT mixed solution by original PESI and non-proximate PESI. (B) TIC from the analysis of non-proximate PESI without (black line) or with (red line) evacuation. The numbered labels on each spectra indicate the applied voltage. (C) Representative mass spectra of the extirpated liver, kidney, spleen, and brain. The blue dots in the insets indicate the regions of interest for each tissue. (D) Plot of the first two PLS scores for the mass spectra from each tissue. The arrow indicates the misclassified spectral data (Brain 3) in Table 1.

Table 1. LOOCV of mass spectra from each tissue by PLS-LR.

Tissue	#	Probability*1				Correct/failure
		Liver	Kidney	Spleen	Brain
Liver	1	0.603	0.254	0.109	0.034	Correct
Liver	2	0.537	0.001	0.070	0.392	Correct
Liver	3	0.661	0.204	0.113	0.022	Correct
Liver	4	0.602	0.129	0.235	0.034	Correct
Liver	5	0.606	0.091	0.258	0.044	Correct
Liver	6	0.520	0.034	0.246	0.200	Correct
Liver	7	0.565	0.138	0.185	0.113	Correct
Liver	8	0.560	0.008	0.103	0.330	Correct
Liver	9	0.484	0.008	0.249	0.259	Correct
Liver	10	0.480	0.002	0.172	0.346	Correct
Liver	11	0.605	0.067	0.164	0.164	Correct
Liver	12	0.667	0.042	0.196	0.095	Correct
Liver	13	0.465	0.108	0.426	0.001	Correct
Liver	14	0.562	0.023	0.334	0.081	Correct
Liver	15	0.587	0.134	0.174	0.105	Correct
Liver	16	0.606	0.082	0.240	0.072	Correct
Liver	17	0.682	0.092	0.127	0.099	Correct
Liver	18	0.683	0.189	0.033	0.095	Correct
Liver	19	0.630	0.267	0.038	0.065	Correct
Liver	20	0.592	0.029	0.209	0.170	Correct
Kidney	1	0.017	0.485	0.476	0.023	Correct
Kidney	2	0.032	0.507	0.243	0.219	Correct
Kidney	3	0.100	0.476	0.149	0.275	Correct
Kidney	4	0.157	0.556	0.088	0.200	Correct
Kidney	5	0.030	0.605	0.033	0.332	Correct
Kidney	6	0.145	0.630	0.069	0.156	Correct
Kidney	7	0.123	0.626	0.119	0.132	Correct
Kidney	8	0.169	0.788	0.015	0.028	Correct
Kidney	9	0.037	0.801	0.025	0.138	Correct
Kidney	10	0.006	0.838	0.140	0.016	Correct
Kidney	11	0.001	0.486	0.448	0.066	Correct
Spleen	1	0.282	0.006	0.379	0.332	Correct
Spleen	2	0.020	0.271	0.607	0.102	Correct
Spleen	3	0.003	0.222	0.424	0.351	Correct
Spleen	4	0.148	0.121	0.579	0.153	Correct
Spleen	5	0.347	0.105	0.507	0.040	Correct
Brain	1	0.038	0.000	0.480	0.483	Correct
Brain	2	0.040	0.232	0.053	0.675	Correct
Brain	3	0.001	0.355	0.316	0.328	Failure
Brain	4	0.129	0.014	0.338	0.519	Correct
Brain	5	0.401	0.003	0.071	0.525	Correct
Brain	6	0.257	0.047	0.139	0.557	Correct

Accuracy*²: 97.62% (41/42)

^*1 A test spectrum was assigned to the class with the highest posterior probability and highest posterior probability is highlighted in red.

^*2 Accuracy is the percentage of the correctly classified spectrum samples to the total number of samples.

Real-time analysis of living animal

To determine whether our newly developed non-proximate PESI-MS system was suitable for analyzing in vivo samples, we evaluated its performance in mice, which were anesthetized and laparotomized to expose their liver. The mice were then fixed in the supine position on a X, Y, Z-axis movable sampling stage (Fig. 3A). Under these conditions, the non-proximate PESI system generated good mass spectra data from the living mouse liver (Fig. 3B) by simply touching the surface of the liver with the tip of the probe. Upon completion of the MS analysis, the liver was extirpated from the mouse and analyzed a second time at the same position on its surface (Fig. 3C). The spectral patterns obtained from the livers ex vivo were found to be significantly different from those of the livers analyzed in situ, as revealed by Welch’s t-test (Fig. 3D). This difference in the spectral patterns of the livers was attributed to exsanguination during the extirpation of the liver. To determine whether these differences in the spectral patterns resulted from changes in the blood contents, we collected blood from a stab wound on the hepatic surface, and compared its spectral pattern with those obtained from the whole blood. As shown in Fig. 3E, there was a significant reduction in the peaks corresponding to the lipid components of the blood, which implied that biological molecules were decomposing or being lost during the resection of the liver. Further studies are therefore required to develop a deeper understanding of the mechanisms involved in this process, as well as identifying molecules that can be readily degraded during surgical procedures and the metabolic profiles and pathways that occur following the extirpation of specific organs.

Fig. 3. (A) An anesthetized mouse was held in the supine position on the sample stage using a clamp (white arrow), and its liver was exposed (yellow arrow). Representative mass spectra patterns from the living mouse liver (B) and the ex vivo liver following its real-time analysis (C). (D) Upper graph indicates the mean normalized mass spectra from living mouse livers (n=14, blue line) and extirpated livers (n=13, red line). Lower graph indicates the p-values calculated using Welch’s two-sample t-test for each m/z value. The green dotted line denotes the Bonferroni-corrected significance level of 0.05/1190=4.2×10⁻⁵. (E) Mass spectrum of the blood collected from the surface of the liver.

Pharmacokinetics of 5-FdU in living mouse

5-FdU is a fluoropyrimidine nucleoside analog showing anti-oncotic and cytotoxic activity, and this compound is regularly used in cell biology to prevent cell mitosis.²¹⁾ Herein, we tested the efficiency of the non-proximate PESI-MS system for detecting 5-FdU in vivo, because this compound can be readily detected by non-proximate PESI-MS with an m/z peak of 269.1, [M+Na]⁺ (Fig. 4A). Furthermore, the peak for 5-FdU does not overlap with any of the other peaks in the sample derived from biological molecules. Furthermore, the results of a previous study demonstrated that 5-FdU shows very low levels of binding to plasma proteins (0.2–5.9%).²²⁾ Taken together, these properties of 5-FdU make it an ideal candidate to test the validity of our newly developed non-proximate PESI-MS system for pharmacokinetic analysis. Thus, a calibration curve was prepared by measuring 5-FdU dissolved in HBSS at concentrations in the range of 0.1–50 μM using the non-proximate PESI-MS system (Fig. 4B). The ion intensity of 5-FdU was then plotted as a function of its concentration. The completed plot revealed that a high linear relationship (coefficient R²=0.9921) was obtained for 5-FdU concentrations in the range of 0.1–12.5 μM, although the ion intensities became saturated at 5-FdU concentrations over 25 μM. One hundred microlitres of a 100 mM 5-FdU solution was then injected into a mouse via its tail vein (Fig. 4C) and its liver was directly analyzed using the method described in Fig. 3A. The results of this analysis revealed that 5-FdU was delivered to the liver by the systemic circulation system about 1 min after its initial injection (Fig. 4D), and that the ion intensity of 5-FdU gradually decreased to 0.4 μM 35 min after injection (Fig. 4D). This value was therefore in the same range as that found in the human case, where the half-life of 2′-deoxy-5-fluorouridine (an analog of 5-FdU) was 15–22 min.²³⁾ Given that the total blood volume in 8-week-old mice is about 3 mL, as reported previously,²⁴⁾ then the final concentration of 5-FdU detected in the blood would be 3.3 mM. This value therefore represents around one thirtieth of the original concentration of 5-FdU, which was 100 mM. However, the ion intensity measure from the liver in vivo gave rise to a concentration of 1.5 μM, which was about one to two thousandths of the initial concentration. These results therefore demonstrate that there is a discrepancy between the concentration and injection volume for these measures taken in situ at the liver. To determine whether an ion suppression effect was responsible for this discrepancy in the concentration of 5-FdU, we tested various concentrations of blood containing 100 μM 5-FdU and discovered that the addition of blood had a significant ion suppression effect (Fig. 4E). Under these conditions, the suppression effect of blood was calculated to be 100-fold stronger based on the estimated concentration of 5-FdU. These results therefore suggest that other unidentified factors in the blood could be attenuating the detection of 5-FdU.

Fig. 4. (A) Single prominent peak of the sodium adduct 5-FdU [M+Na]⁺ was detected with an m/z value of 269.1. (B) Calibration curve for the quantitative analysis of 5-FdU in mouse liver using non-proximate PESI-MS. The ion intensity of 5-FdU has been plotted against the concentration. R² denotes the coefficient of determination. (C) Photograph of the procedure used for injecting 5-FdU into the mouse tail vein. (D) Extracted ion chromatogram of 5-FdU (m/z 269.1). Arrow indicates the time point of the 5-FdU injection. Soon after the injection, 5-FdU was directly detected from the living mouse liver. The two dotted lines indicate the calculated concentration of 5-FdU according to the results in panel B. (E) Graph showing the possible ion suppression effect of blood on the analysis of 5-FdU. Various amounts of blood were added to a 100 μM 5-FdU analyte solution. Data are expressed as relative values to the intensity of a 10 μM 5-FdU standard solution and shown as the mean values±S.D.

CONCLUSION

In this study, we have validated the use of non-proximate PESI-MS for the real-time analyses of living animals in situ. To optimize the PESI-MS system for the real-time analysis of larger samples placed at a reasonable distance from the ion inlet, it was necessary to improve the ion collection efficiency by employing an extended stainless steel sampling tube, which was assisted by an external vacuum pump. The resulting system was subsequently termed non-proximate PESI and could potentially be installed on every mass spectrometer depending on their primary purpose (e.g., quantitative, tandem, or high resolution MS). This modification made it possible for us to directly detect lipid ion species from living animals with high efficiency and reproducibility. We believe that the use of this modification would make a good pilot study in terms of its application to biomedical research. Given that the injected 5-FdU could be directly detected on the serosal surface of the mouse using this technique, it is clear that this non-proximate PESI-MS system has great potential in terms of being used to study the pharmacokinetic profiles of drugs and chemical compounds in vivo. Taken together, the properties and many advantages of non-proximate PESI suggest that this technique has significant potential to be used in various situations.

Acknowledgements

We thank A. Iizuka and Y. Shimura for their assistance with this project. This work was partially supported by SENTAN, JST (to ST), JSPS KAKENHI Grant Number 25713023 (Grant-in-Aid for Young Scientists (A), to K.Y.), and Grant Number 26505003 (Grant-in-Aid for Scientific Research (C), to L.C.C.). This study is performed in collaboration with the SHIMADZU Corporation.

REFERENCES

• 1) M. Wilm, M. Mann. Analytical properties of the nanoelectrospray ion source. Anal. Chem. 68: 1–8, 1996.
• 2) D. I. Carroll, I. Dzidic, R. N. Stillwell, M. G. Horning, E. C. Horning. Subpicogram detection system for gas phase analysis based upon atmospheric pressure ionization (API) mass spectrometry. Anal. Chem. 46: 706–710, 1974.
• 3) R. B. Cody, J. A. Laramée, H. D. Durst. Versatile new ion source for the analysis of materials in open air under ambient conditions. Anal. Chem. 77: 2297–2302, 2005.
• 4) J. Balog, T. Szaniszlo, K. C. Schaefer, J. Denes, A. Lopata, L. Godorhazy, D. Szalay, L. Balogh, L. Sasi-Szabo, M. Toth, Z. Takáts. Identification of biological tissues by rapid evaporative ionization mass spectrometry. Anal. Chem. 82: 7343–7350, 2010.
• 5) J. Balog, L. Sasi-Szabó, J. Kinross, M. R. Lewis, L. J. Muirhead, K. Veselkov, R. Mirnezami, B. Dezső, L. Damjanovich, A. Darzi, J. K. Nicholson, Z. Takáts. Intraoperative tissue identification using rapid evaporative ionization mass spectrometry. Sci. Transl. Med. 5: 194ra93, 2013.
• 6) Z. Takáts, J. M. Wiseman, B. Gologan, R. G. Cooks. Mass spectrometry sampling under ambient conditions with desorption electrospray ionization. Science 306: 471–473, 2004.
• 7) K. Hiraoka, K. Nishidate, K. Mori, D. Asakawa, S. Suzuki. Development of probe electrospray using a solid needle. Rapid Commun. Mass Spectrom. 21: 3139–3144, 2007.
• 8) L. C. Chen, K. Yoshimura, Z. Yu, R. Iwata, H. Ito, H. Suzuki, K. Mori, O. Ariyada, S. Takeda, T. Kubota, K. Hiraoka. Ambient imaging mass spectrometry by electrospray ionization using solid needle as sampling probe. J. Mass Spectrom. 44: 1469–1477, 2009.
• 9) K. Yoshimura, L. C. Chen, D. Asakawa, K. Hiraoka, S. Takeda. Physical properties of the probe electrospray ionization (PESI) needle applied to the biological samples. J. Mass Spectrom. 44: 978–985, 2009.
• 10) K. Yoshimura, L. C. Chen, Z. Yu, K. Hiraoka, S. Takeda. Real-time analysis of living animals by electrospray ionization mass spectrometry. Anal. Biochem. 417: 195–201, 2011.
• 11) K. Yoshimura, L. C. Chen, M. K. Mandal, T. Nakazawa, Z. Yu, T. Uchiyama, H. Hori, K. Tanabe, T. Kubota, H. Fujii, R. Katoh, K. Hiraoka, S. Takeda. Analysis of renal cell carcinoma as a first step for developing mass spectrometry-based diagnostics. J. Am. Soc. Mass Spectrom. 23: 1741–1749, 2012.
• 12) K. Yoshimura, M. K. Mandal, M. Hara, H. Fujii, L. C. Chen, K. Tanabe, K. Hiraoka, S. Takeda. Real time diagnosis of chemically-induced hepatocellular carcinoma using a novel mass spectrometry-based technique. Anal. Biochem. 441: 32–37, 2013.
• 13) M. K. Mandal, L. C. Chen, Y. Hashimoto, Z. Yu, K. Hiraoka. Detection of biomolecules from solutions with high concentration of salts using probe electrospray and nano-electrospray ionization mass spectrometry. Anal. Methods 2: 1905–1912, 2010.
• 14) M. K. Mandal, L. C. Chen, Z. Yu, H. Nonami, R. Erra-Balsells, K. Hiraoka. Detection of protein from detergent solutions by probe electrospray ionization mass spectrometry (PESI-MS). J. Mass Spectrom. 46: 967–975, 2011.
• 15) S. Takeda, K. Yoshimura, K. Hiraoka. Innovations in analytical oncology—Status quo of mass spectrometry-based diagnostics for malignant tumor. J. Anal. Oncol. 1: 74–80, 2012.
• 16) S. Garimella, W. Xu, G. Huang, J. D. Harper, R. G. Cooks, Z. Ouyang. Gas-flow assisted ion transfer for mass spectrometry. J. Mass Spectrom. 47: 201–207, 2012.
• 17) Z. Luo, J. He, Y. Chen, J. He, T. Gong, F. Tang, X. Wang, R. Zhang, L. Huang, L. Zhang, H. Lv, S. Ma, Z. Fu, X. Chen, S. Yu, Z. Abliz. Air flow-assisted ionization imaging mass spectrometry method for easy whole-body molecular imaging under ambient conditions. Anal. Chem. 85: 2977–2982, 2013.
• 18) J. A. Wegelin. A survey of Partial Least Squares (PLS) methods, with emphasis on the two-block case. Technical Report 371, Department of Statistics, University of Washington, Seattle, 2000.
• 19) R. E. Fan, K. W. Chang, C. J. Hsieh, X. R. Wang, C. J. Lin. LIBLINEAR: A library for large linear classification. J. Mach. Learn. Res. 9: 1871–1874, 2008.
• 20) F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, E. Duchesnay. Scikit-learn: Machine learning in python. J. Mach. Learn. Res. 12: 2825–2830, 2011.
• 21) S. Hatse, E. De Clercq, J. Balzarini. Role of antimetabolites of purine and pyrimidine nucleotide metabolism in tumor cell differentiation. Biochem. Pharmacol. 58: 539–555, 1999.
• 22) K. G. Csaky, F. P. LaCreta, B. S. Warren, W. M. Williams. 5-Fluoro-2′-deoxyuridine elimination by the isolated perfused rat liver. Cancer Res. 48: 3561–3565, 1988.
• 23) S. A. Van Der Heyden, M. S. Highley, E. A. De Bruijn, U. R. Tjaden, H. J. Reeuwijk, H. Van Slooten, A. T. Van Oosterom, R. A. Maes. Pharmacokinetics and bioavailability of oral 5′-deoxy-5-fluorouridine in cancer patients. Br. J. Clin. Pharmacol. 47: 351–356, 1999.
• 24) A. C. Riches, J. G. Sharp, D. Brynmor Thomas, S. Vaughan Smith. Blood volume determination in the mouse. J. Physiol. 228: 279–284, 1973.