In Vivo Proton Magnetic Resonance Spectroscopy of Pancreatic Neuroendocrine Tumors (PNETs) in a Multiple Endocrine Neoplasia Type 1 (MEN1) Conditional Knockout Mouse Model

Abstract

Purpose

Magnetic resonance spectroscopy (MRS) can improve diagnosis and follow treatment in cancer. However, no study has yet reported application of in vivo ¹H-MRS in malignant pancreatic lesions. This study quantitatively determined whether in vivo ¹H-MRS on multiple endocrine neoplasia type 1 (Men1) conditional knockout (KO) mice and their wild type (WT) littermates could detect differences in total choline (tCho) levels between tumor and control pancreas.

Methods

Relative tCho levels in pancreatic tumors or pancreata from KO and WT mice were determined using in vivo ¹H-MRS at 9.4 T. Levels of choline-containing compounds were also quantified using in vitro ¹H-NMR on extracts of pancreatic tissues from KO and WT mice respectively, and on extracts of pancreatic tissues from patients with pancreatic neuroendocrine tumors (PNETs).

Results

tCho levels measured by in vivo ¹H-MRS were significantly higher in PNETs from KO mice compared to the normal pancreas from WT mice. The elevated choline-containing compounds were also identified in pancreatic tumors from KO mice and tissues from patients with PNETs via in vitro ¹H-NMR.

Conclusion

These results indicate the potential use of tCho levels estimated via in vivo ¹H-MRS in differentiating malignant pancreatic tumors from benign tumors.

Introduction

Pancreatic neuroendocrine tumors (PNETs) are clinically rare, with an incidence of 1 per 100,000 people¹. In the last two decades, however, the reported incidence has increased 2-3 fold, reaching 1000 new cases per year in the United States^2-4. Clinically, PNETs are divided into functional and nonfunctional tumors based on hormone production and associated clinical syndromes⁵. Biochemical assessments are necessary to diagnose a functional PNET⁶, while nonfunctional PNETs are not associated with a characteristic syndrome¹ and they are often diagnosed incidentally during imaging exams for unrelated conditions⁷. However, once a PNET is identified, an approach to distinguish the metabolic features of PNETs may be of value, and such an approach would likely be essential to the longitudinal monitoring of therapy response to ensure successful and appropriate treatment.

In vivo proton magnetic resonance spectroscopy (¹H-MRS) has been explored to characterize breast cancers^8,9, prostate cancers¹⁰ and brain cancers¹¹. Malignant tumors are characterized by alterations in the regulation of a variety of metabolic pathways associated with the proliferation of tumor cells^12,13. Studies have shown that in vivo ¹H-MRS provides information on tumor metabolism, which can be used to improve diagnostic accuracy⁹ and monitor the therapeutic response of tumors^14,15. In vivo ¹H-MRS spectra often exhibit resonance at ~3.2 ppm, which corresponds to cholinergic compounds like free choline (Cho), phosphocholine (PCho), and glycerophosphocholine (GPCho), etc¹⁶. Elevated levels of cholinergic compounds (total choline, tCho) have been associated with malignancy^8,10. However, application of in vivo ¹H-MRS to pancreatic tumors has not advanced to the same extent as it has to other previously mentioned cancers. This may be due to difficulty with application of ¹H-MRS to abdominal cancers arising from breathing or peristaltic motions that can lead to severe imaging artifacts¹⁷. Recently, Su et al¹⁸ demonstrated the feasibility of in vivo ¹H-MRS in the human pancreas, which led to detection of choline. No study has yet reported measurements of ¹H-MRS in malignant pancreatic lesions or evaluated the specificity of tCho measurements from in vivo ¹H-MRS to differentiate pancreatic lesion type.

Recently, we developed a transgenic mouse model based on the homozygous inactivation of the Men1 gene in the pancreas¹⁹. These well-described mice display the characteristics of MEN1 syndrome seen in humans. Specifically, these conditional knockout (KO) mice develop PNETs, recapitulating the phenotype observed in MEN1 patients and providing a powerful tool for the study of tumor progression and treatment. The purpose of this study was to quantitatively determine whether in vivo ¹H-MRS on Men1 KO mice and their wild type (WT) littermates could detect differences in tCho levels between tumor and control pancreas. For comparison we examined tCho levels from in vitro ¹H-NMR on extracts of pancreatic tumor tissues and normal pancreatic tissues from Men1 KO and WT mice respectively. We then conducted in vitro ¹H-NMR measurements on extracts of pancreatic tissues from patients with PNETs. Total Cho levels were compared between tumor tissues and normal adjacent tissues of pancreas from patients with PNETs.

Methods

Animals

All animal experiments were conducted after approval by the Animal Institute of Albert Einstein College of Medicine. Individual animal numbers used in each experiment are identified beneath each experiment description below.

In vivo MRI/MRS

Ten Men1 KO mice and eleven WT littermates with matched age and sex were studied by in vivo MRI/MRS. The KO mice exhibited symptoms of insulinoma progression¹⁹, resulting in significant weight loss compared to their WT littermates (28.7±2.3 g vs. 41.3±3.2 g, p=0.005), as shown in Table 1. MRI was acquired on a 9.4 T Varian Direct Drive animal magnetic resonance imaging and spectroscopic system (Agilent Technologies, Inc., Santa Clara, CA). Mice were fasted for about four hours before MRI/MRS to reduce peristaltic movement. Mice were anaesthetized with isoflurane mixed with room air and placed supinely in an MR compatible holder. Surface body temperature was maintained at 34-35 °C using warm air with feedback from a thermocouple. The abdomen was centered in a 35-mm ID quadrature ¹H volume coil (m2m Imaging Co., Cleveland, OH). After scout images, consecutive coronal slices covering the whole abdomen were taken by a gradient echo imaging sequence gated to the respiratory cycle²⁰ via an air pressure transducer (Small Animal Instruments, Inc., Stony Brook, NY). Water-suppressed ¹H-MRS data were acquired via a LASER sequence²¹ on a cubic voxel placed in the body of the pancreas (controls) or pancreatic tumor in a location defined based on the gradient-echo coronal images. The parameters for ¹H-MRS are as follows: TR=2.2-2.8 s, TE=36 ms, transients=256, complex point=1024, spectral width=4006 Hz. Data acquisition was gated to the respiratory cycle. Shimming routinely resulted in the voxel un-suppressed water signal linewidth (FWHM) of 20-40 Hz or 0.05-0.1 ppm. Water suppression (WS) was achieved using the WET sequence²². Bandwidth of the WS RF pulse was 225 Hz and duration of the WS RF pulse was 3 ms. Non-suppressed water spectra (TR=10.3-12.5 s, TE=36 ms, transients=16) were also acquired from the same metabolic voxel as an internal reference. All spectra were analyzed in the frequency domain using Agilent VnmrJ integration routine with linewidth of 10 Hz FWHM. For relative quantitation of the tCho, the tCho peak at 3.2 ppm was normalized relative to the internal water signal²³ from the same voxel after T1 correction of saturation factors²⁴.

Table 1.

Characteristics of Men1 KO mice and WT littermates.

	Men1 KO (n = 10)	WT (n = 11)	P value
Age (months)	10.7 ± 0.8	13.5 ± 1.5	NS
Gender	6F/4M	5F/6M	NS
Weight (g)	28.7 ± 2.3	41.3 ± 3.2	0.005
tCho peak observed	10/10	8/11
tCho/H2O (× 104)	11.1 ± 2.1	3.3 ± 0.5**	0.003

* Data are expressed as means with standard error.

^**tCho/H₂O ratios are expressed as mean values from eight WT pancreata, in which tCho peak was detected by in vivo ¹H-MRS.

In vitro ¹H-NMR Spectroscopy

Normal pancreata and MEN1 pancreatic neuroendocrine tumors were obtained from patients as part of a tissue procurement protocol approved by the Albert Einstein College of Medicine Institutional Review Board, with all patients signing informed consent. After surgical removal, the samples were flash frozen in liquid nitrogen and stored at −80 °C for 3-12 month before in vitro ¹H-NMR experiments. Pancreatic tissues from a second group of mice (WT: n=5; KO: n=6) and humans (normal: n=8; tumor: n=9) were weighted, homogenized and extracted with 50:50 CH₃CN and water²⁵. Following lyophilization, the pellet was dissolved into 550 μL of D₂O and 5 μL of 10-mM trimethylsilyl propionate (TSP) stock, which was used as an internal standard. ¹H-NMR spectra were acquired with a standard presaturation ¹H sequence on a Bruker DRX 600-MHz spectrometer with a 5-mm TXI probe (Bruker BioSpin Co., Billerica, MA). The raw data were multiplied with 0.3 Hz exponential line broadening before Fourier transformation and processed using the NUTS program (Acorn NMR Inc., Livermore, CA). The chemical shift δ was assigned with reference to the internal standard TSP (δ=0 ppm). PCho, GPCho, choline (Cho), taurine (Tau) and lactate (Lac) metabolite levels were determined and normalized to the internal standard TSP and the tissue weight. Total choline (tCho) levels were calculated as the sum of Cho, PCho, and GPCho levels.

Measurement of Choline Kinase α (ChoKα) mRNA and Protein Expression by Real-time RT PCR as well as Western Blot and Immunofluorescence (IF) assays

Abnormal expression of ChoKα, an enzyme involved in the cholinergic pathway for phosphocholine biosynthesis, was identified in malignant transformation²⁶. In this study, we assessed the mRNA expression levels of ChoKα in pancreatic endocrine tissues from a third group of five Men1 KO and five WT mice by laser capture microdissection (LCM) real-time (RT)-PCR. RT-PCR products were measured as fluorescent signal intensity after standardization with a glyceraldehyde 3-phosphate dehydrogenase (GAPDH) internal control. The following primers for ChoKα were used: Sense primer for ChoKα: 5’-GCCGACTGGAGCAGTTTATC-3’; Antisense primer for ChoKα: 5’-GCTCAAGAGGCAGGTTGTAA-3’.

The protein expression of ChoKα in pancreatic tissues from KO and WT mice was conducted by Western blotting¹⁹. IF staining for ChoKα on sections was also performed from Men1 KO and WT mice using rabbit anti-ChoKα antibody (SC-32907, Santa Cruz Biotechnology, CA) and were incubated at a 1:50 dilution overnight at 4°C. Pig anti-insulin Ab was also used for identification of β cells. Sections were then washed with PBS and were incubated with a 1:200 dilution of anti-rabbit Alexa Fluor 488 and anti-pig Alexa Fluor 647 secondary antibodies for 45 minutes in the dark, respectively. A negative control without primary antibody was also performed.

Statistics

Statistical analysis was performed using Microsoft Excel 2011. Group differences were analyzed using two-tailed Student's t-test with p < 0.05 considered statistically significant.

Results

In vivo MRI and ¹H-MRS

Pancreatic tumors can be identified from coronal images of the abdomen of the Men1 KO mice along with other organs such as liver, gallbladder, kidney, stomach and spleen, as shown in Fig.1A. For WT mice, the pancreas was delineated by first locating the pyloric sphincter of the stomach that connects to the duodenum. Since the pancreas is bound to the duodenum, it is usually found between the duodenum and stomach, as outlined in red in Fig.1B of a coronal image of WT mouse.

Fig.1.

(A) Coronal MRI image of a Men1 KO mouse showing gallbladder (G), kidney (K), liver (L), stomach (ST), spleen (SP) and pancreatic tumors (T). The green box demonstrates the placement of voxel (9.8 mm³) for in vivo ¹H-MRS acquisition. (B) Coronal MRI image of a WT mouse showing gallbladder, kidney, duodenum (D), liver, pancreas (P), stomach and spleen. Pancreas is outlined in red. The green box demonstrates the placement of voxel (8.8 mm³) for in vivo ¹H-MRS acquisition. In vivo ¹H-NMR spectrum of pancreatic tumor of Men1 KO mouse (C) and of pancreas of WT mouse (D) showing total choline (tCho) methyl proton peak at 3.2 ppm and resonances from lipid. The spectra were scaled to the non-suppressed water signal.

In vivo ¹H-MRS data were acquired on pancreata of Men1 KO mice and WT littermates. Fig.1C and Fig.1D display the in vivo ¹H-NMR spectra of pancreatic tumor of Men1 KO mouse and pancreas of WT mouse, respectively. The peak at 3.2 ppm assigned to tCho was detected in the pancreata of all ten Men1 KO mice aged from 7 to 15 months and eight out of eleven WT littermates aged from 7 to 20 months. Besides the tCho peak at 3.2 ppm and residual water peak at 4.7 ppm, lipid peaks were also observed. Voxel size was 11.6±1.2 mm³ in pancreatic tumor of KO mice vs. 8.2±1.1 mm³ in pancreas of WT mice, p=0.05. The signal-to-noise (S/N) ratio of tCho peak from KO mice was higher than that from WT mice with detected tCho peak, e.g., 16.8±2.6 vs. 9.5±1.9, p=0.04. In the ten Men1 KO mice, tCho levels normalized to the un-suppressed water peak from the same voxel, i.e., tCho/H₂O (×10⁴), in pancreata or pancreatic tumors was more than three fold of that in the pancreata of the eight WT littermates, in which we observed a tCho peak, i.e., 11.1±2.1 ×10⁻⁴ vs. 3.3±0.5 ×10⁻⁴, p=0.005 (Table 1). In a separate set of in vitro assessments (unreported) we determined that our detection sensitivity for tCho was approximately 1 mM.

In vitro ¹H-NMR

In order to understand what compounds contribute to the elevated tCho levels observed in pancreata or pancreatic tumors in Men1 KO mice compared to WT littermates, tissues from the pancreata or pancreatic tumors were extracted and studied with high-resolution ¹H-NMR. In ¹H-NMR spectra of extracts from pancreatic tissue, Cho, PCho, and GPCho resonate at 3.207 ppm, 3.226 ppm and 3.234 ppm, respectively (Fig.2). These choline-containing peaks can be differentiated in ¹H-NMR spectra of extracts, but are not discernible within in vivo ¹H-MRS spectra. In vitro data, illustrated in Fig.3A revealed a three-fold increase in PCho, GPCho and tCho in pancreatic tissue from Men1 KO mice in comparison with pancreatic tissue from WT littermates. In vitro ¹H-NMR data on human pancreas tissues also showed significantly increased levels of cholinergic compounds in tumor tissues compared to normal tissues (Fig.3B). PCho levels were significantly higher than GPCho level in human tumor tissues (p=0.002), and demonstrated a trend towards higher levels in mouse KO pancreatic tissues (p=0.166). Taurine levels were higher in human tumor tissue than in human normal tissue (p=0.001). The lactate levels were not significantly different between WT vs. Men1 KO mouse pancreatic tissues or between human normal vs. tumor pancreatic tissues.

Fig.2.

In vitro ¹H-NMR spectrum of pancreas extract of Men1 KO mouse pancreatic tumor (A) showing Lac doublet and resonates from choline-containing compounds. Expanded region of ¹H-NMR spectra from KO mouse pancreatic tumor (B) and WT mouse pancreas (C) showing Cho, PChO and GPChO peaks at 3.207 ppm, 3.226 ppm, and 3.234 ppm respectively. Also shown are Tau (N)CH₂ and (S)CH₂ peaks at 3.263 ppm and 3.431 ppm.

Fig.3.

Integrals of metabolites (Cho, PCho, GPCho, Tau, Lac and tCho) relative to TSP standard per mg of tissue (I_M/I_TSP/mg tissue) in (A) pancreatic tissue extracts of WT mice (n = 5, solid bars) and Men1 KO mice (n = 6, striped bars) and (B) tumor tissues (n = 9, striped bars) and normal adjacent tissues (n = 8, solid bars) of pancreata from the patients with PNETs were estimated by in vitro ¹H-NMR spectra. Values represent mean ± standard error, *p < 0.05, ^#p < 0.01 for group comparison.

Increased mRNA and Protein Expression of ChoKα in PNETs from Men1 KO Mice

mRNA and protein expression of ChoKα showed a significant increase in the endocrine tumor tissues of the pancreas from Men1 KO mice compared with normal endocrine tissues of the pancreas in WT control mice (Fig.4A). Furthermore, increased ChoKα protein levels were confirmed in pancreatic endocrine tumors from Men1 KO mice compared to normal pancreatic tissues from WT mice by Western blot (Fig.4B) and IF assays (Fig.4C). Taken together, these findings indicated that ChoKα is significantly upregulated upon in the development of pancreatic endocrine tumor.

Fig.4.

Increased mRNA and protein expression of Choline Kinase-α (ChoKα) in PNETs from Men1 KO mice. ChoKα mRNA was significantly increased in PNETs from Men1 KO mice (n = 5) compared with normal pancreatic endocrine tissues from WT mice (n = 5) by LCM real-time RT PCR assay (A). ChoKα expression was significantly increased in the PNETs from Men1 KO mice compared with normal pancreatic endocrine tissues from WT mice by Western blot (B) and IF staining (C).

Discussion

In vivo ¹H-MRS is evolving as a sensitive tool for the evaluation of benign versus malignant lesions and for monitoring therapy response, in which tumors are characterized by their choline resonance at ~3.2 ppm, which is known to be associated with malignancy. Total choline levels estimated via in vivo ¹H-MRS may serve as a better predictor than the tumor volume in predicting tumor progression²⁷ and in response to therapy²⁸ since metabolites, such as total choline, reflect changes in tissue metabolism and tumor progression. However, the use of in vivo ¹H-MRS in pancreatic cancer research is challenging, especially in mouse models due to the abdominal location of the pancreas and associated motion artifacts from breathing and bowel movements, which result in partial volume, localization and acquisition errors in the MRS measurement process. Nonetheless, the application of motion-sensitivity reducing techniques in the MR measurement has been demonstrated effective at reducing associated errors¹⁸. In addition to respiratory gating, we found that fasting mice for about four hours prior to imaging substantially improved imaging quality by reducing peristaltic motions. We also optimized the imaging protocol, by comparing T1- vs. T2-weighted approaches, matrix size (in-plane resolution), slice thickness, view direction (axial vs. coronal), echo time (TE) and determined that delineation of the pancreata or pancreatic tumors was most reliable when a T1-weighted protocol with in-plane resolution of 0.16×0.16 mm², slice thickness of 0.2 mm and coronal acquisition were used. Using these combined approaches, the pancreata and pancreatic tumors can be reliably imaged in mice (Fig.1A&B).

The purpose of this study was to assess if tCho peaks at 3.2 ppm can be detected in pancreatic tumors and normal pancreata in mice and if tCho levels measured using in vivo ¹H-MRS may be reflective of tumor presence and thus diagnosis of PNETs in our Men1 conditional KO mouse model. To the best of our knowledge, this is the first study wherein the tCho peak was detected in mouse pancreata using an in vivo ¹H-MRS approach. Our data showed that the tCho signal was detected in 10/10 pancreata or pancreatic tumors of Men1 KO mice and 8/11 pancreata of WT littermates. We also found, for the first time, that tCho levels, determined using tissue water as an internal reference²³, in pancreatic tumors or pancreata of Men1 KO mice, are significantly higher than that in WT littermates. tCho levels in pancreata from Men1 KO mice are about four fold that seen in WT littermates. The elevated tCho levels in pancreatic tumors measured using in vivo ¹H-MRS are likely due to elevated PCho and GPCho levels, which were estimated using in vitro ¹H-NMR on pancreatic tissue extracts (Fig.3A). Human pancreatic tumor tissues also demonstrated significantly high levels of tCho (Fig.3B). Finally, evidence supporting an abnormal choline phospholipid metabolism was suggested by a more than five-fold increase of choline kinase α mRNA expression in Men1 KO mouse pancreatic tissue compared to WT mouse pancreatic tissue.

It is known that higher tCho, especially higher PCho, is associated with increased cell replication, membrane synthesis, and phospholipid components in malignant tumors^29,30. However, this is the first study to report the elevated tCho levels in a Men1 KO mouse model using in vivo ¹H-MRS technique, as well as higher PCho, GPCho and tCho levels in both mouse and human pancreatic tumor tissues, via an in vitro ¹H-NMR method. These results suggest that the identification of the level of choline and cholinergic compounds in pancreata using noninvasive MRS can potentially play a major role in early detection and diagnosis of PNETs and in predicting therapeutic benefit in PNETs. Further studies examining Cho-levels associated with tumor progression, at different tumor stages and correlated with treatment outcomes for PNETs in preclinical and clinical settings are warranted. However, it is not known if in vivo ¹H-MRS can differentiate malignant nonfunctional PNETs from benign nonfunctional PNETs, as there is no currently available mouse model with nonfunctional PNETs. Notwithstanding, we are in the process of applying this method with human patients including patients with nonfunctional PNETs.

Conclusions

Our results demonstrate the feasibility of imaging pancreata in live mouse using conventional imaging methods coupled with fasting and respiratory gating, and of using tCho levels measured via in vivo ¹H-MRS in conjunction with MRI in characterizing malignant and normal pancreatic tissues. tCho levels were found to be significantly higher in pancreata from Men1 KO mice compared with WT littermates. Our in vivo findings were replicated with in vitro tissue extracts. This method allows for a noninvasive, presurgical assessment of the proliferative activity of pancreatic tumors. In vivo ¹H-MRS may be useful in differentiating invasive cancer from benign lesions by the higher total choline levels as observed in our Men1 KO mouse model. The higher choline compounds as well as taurine, observed in human pancreatic tumor tissues imply that this in vivo ¹H-MRS method can potentially differentiate malignant pancreatic tumors from benign tumors and maybe useful for monitoring of treatment response in PNET patients.

Abbreviations

¹H-NMR

proton nuclear magnetic resonance

¹H-MRS

proton magnetic resonance spectroscopy

ChoKα

Choline Kinase α

Cho

choline

GAPDH

glyceraldehyde 3-phosphate dehydrogenase

GPCho

glycerophosphocholine

Lac

lactate

Men1

multiple endocrine neoplasia type 1

PCho

phosphocholine

PNET

pancreatic neuroendocrine tumor

Tau

taurine

tCho

total choline

TSP

trimethylsilyl propionate