[¹¹C]Choline Positron Emission Tomography in Estrogen Receptor–Positive Breast Cancer

Purpose: Novel radiotracers could potentially allow the identification of clinically aggressive tumor phenotypes. As choline metabolism increases during malignant transformation and progression of human mammary epithelial cells, we examined the ability of [¹¹C]choline (CHO) positron emission tomography imaging to detect clinically aggressive phenotype in patients with estrogen receptor (ER)–positive breast cancer in vivo.

Experimental Design: CHO positron emission tomography was done in 32 individuals with primary or metastatic ER–positive breast cancer. Semiquantitative (standardized uptake value) and fully quantitative (net irreversible transfer rate constant of CHO, Ki) estimates of CHO uptake in the tumors were calculated and compared with tumor grade, size, involved nodes, and also ER, progesterone receptor, Ki-67, and human epidermal growth factor receptor-2 scores.

Results: Breast tumors were well visualized in 30 of 32 patients with good tumor background ratios. A wide range of uptake values were observed in primary and metastatic tumors. CHO uptake variables correlated well with tumor grade. For most imaging variables, a poor association was found with tumor size, ER, progesterone receptor, human epidermal growth factor receptor-2, Ki-67, and nodal status.

Conclusions: CHO showed good uptake in most breast cancers and merits further investigation as a breast cancer imaging agent. (Clin Cancer Res 2009;15(17):5503–10)

Breast cancer is the commonest female malignancy with >1.2 million women diagnosed each year worldwide. Over 75% of breast carcinomas express estrogen receptor (ER). Hence, ER-positive breast cancer is usually treated with antiestrogens, such as tamoxifen. Adjuvant hormonal treatment with tamoxifen followed by an aromatase inhibitor for a total of 5 years (1), as well as aromatase inhibitors alone (2), have been shown to improve recurrence-free survival and overall survival. However, patients relapse despite adjuvant hormonal therapy, due to drug resistance. Only approximately half of the patients with recurrence in ER-positive disease respond to endocrine therapy, whereas the remainder show resistance, clinically evidenced by progression (3). The mechanisms underlying this process are complex, but it is generally agreed that, at least in some patients, cross-talk between cell surface receptor kinases and ER may play a role. In this regard, phosphorylation of ERα by the mitogen-activated protein kinase (MAPK; refs. 4–7) and phosphatidylinositol-3-kinase pathways (3) have been implicated. In particular, MAPK-mediated hyperphosphorylation of ERα on S118, S104, and S106 collectively leads to ligand-independent receptor activation, thereby potentially contributing to resistance to antiendocrine agents, such as tamoxifen and aromatase inhibitors (4, 8, 9). A large proportion of breast cancers have elevated MAPK (10), and expression correlates to poor response to endocrine therapies (11). There is an unmet need to develop imaging methods to guide clinicians as to the correct choice of subsequent therapies.

Whereas FDG positron emission tomography (PET) has been extensively used as an imaging biomarker for diagnosis (12) and for monitoring of treatment response in cancer patients (13), newer imaging biomarkers including [¹⁸F]fluorothymidine and [¹¹C]choline (CHO) are being examined for their greater specificity in the assessment of proliferation (14, 15) and choline metabolic activity (16), respectively. The utility of CHO PET in breast cancer has not been studied. Of interest, Aboagye and Bhujwalla (17) previously showed a unique phenotype of increased choline metabolism in the transition of normal human mammary epithelial cells to immortalized, oncogene-transformed, and finally nonmetastatic and metastatic cancer. In their study, they showed a switch from predominantly higher intracellular metabolite levels of glycerophosphophocholine (degradation pathway) to predominantly higher phosphocholine levels (biosynthetic pathway) early in cell transformation (17). Phosphocholine levels then increased in a stepwise manner with progression (17) such that breast cancer cells had the highest phosphocholine levels. Further studies in mammary epithelial cells showed that the aberrant increases in phosphocholine metabolite levels were due to expression of the biosynthetic enzyme, choline kinase-α (18, 19). Coincidentally, choline kinase activity and cellular phosphocholine levels are regulated by the growth factor receptor–MAPK pathway in mammalian cells (17, 20–24), the same pathway that modulates estrogen-independent growth. Beyond breast cancer, the level of choline kinase-α in tumors of patients with non–small cell lung cancer has been shown recently to have a significant effect on survival; in that study, tumor choline kinase-α expression of at least 1.91 times of that expressed in normal tissue fared worse with respect to relapse-free and lung cancer–specific overall survival after definitive lung cancer surgery (25).

To investigate this further, in the context of the detection of clinically aggressive breast phenotype, we examined the role of CHO PET as an imaging tool in breast cancer. Given that choline metabolism increases with transformation and progression, we hypothesized that high CHO phosphorylation will be associated with high tumor grade, reflecting a more aggressive phenotype.

Patients with biopsy-proved and ER-positive breast cancer were recruited. We selected this patient population to retain a level of homogeneity with respect to tumor presentation, prognosis, and survival. Recruitment period was for 3.5 y. In this initial study, both primary and heavily pretreated metastatic cases (including lung and omental metastasis and Krukenburg tumors) were included to provide a dynamic range of uptake values. Primary cases were scanned before any treatment. Metastatic cases were those who were progressing in spite of previous therapy in the form of chemotherapy or hormonal therapy and had a chemotherapy-free and radiotherapy-free interval of a minimum of 3 and 4 wk, respectively, before undergoing PET to negate any residual effects of previous therapies. To minimize the effect of partial volume averaging, all individuals scanned required at least one measurable lesion (≥20 mm in the longest dimension), which was confirmed on routine ultrasound (primary cases) or contrast-enhanced spiral computed tomography (metastatic cases). Breast tumors often metastasize to the liver. Tumors in this organ were, however, excluded from analysis because of the high physiologic uptake of CHO (although protocols are under way within our group to overcome this difficulty). Other inclusion criteria were as follows: life expectancy of ≥3 mo, hemoglobin of ≥10 g/dL, granulocyte count of ≥1.5 × 10⁹/L, platelet count of at least 100 × 10⁹/L, adequate creatinine clearance (≥50 mL/min calculated), adequate hepatic function (bilirubin, ≤1.5 upper limit of normal; serum transaminases, ≤2.5 upper limit of normal), and ability to consent and comply fully with the study protocol. The exclusion criteria were pregnant or lactating patients, sexually active patients who were not using adequate contraception (pregnancy tests were done in all women of childbearing potential), patients with uncontrolled medical conditions, patients unable to lie flat, and patients with evidence of cerebral metastases. The study followed the “Declaration of Helsinki's — Ethical Principles for Medical Research involving Human Subjects.” Appropriate ethical permission from the Hospital Research Ethics Committee and a license from the Administration of Radioactive Substances Advisory Committee UK were obtained, and all patients provided written informed consent. A minimum of 10 d between core biopsy and PET scanning was ensured to prevent false-positive uptake of CHO.

CHO was manufactured by standard methods at the local hospital as previously described (26, 27). The mean injected activity of CHO measured 320 MBq (range 188-360 MBq) with mean specific radioactivity of 13,279 MBq/μmol and mean radiochemical purity of >99%. PET scanning was done on an ECAT HR+/962 PET scanner (CTI/Siemens) at Hammersmith Hospital with a 15.7-cm field of view.

An arterial cannula was inserted in the radial artery (size 22 G) to allow for continuous (total 50 mL; 5 mL/min) and discrete (total 100 mL; baseline and 2.5, 3.5, 5, 7.5, 10, 15, 20, 30, 45, and 50 min after injection of CHO) blood sampling throughout the CHO scan. A venous cannula (size 22 G) was positioned in the opposite arm for radiotracer injection. The patient was positioned in the scanner with the region of interest centered in the field of view. Single-bed position dynamic PET scanning was commenced 30 s before i.v. injection of CHO as a single bolus and consisted of 26 frames of varying time intervals (5 × 10, 2 × 20, 2 × 30, 5 × 60, 3 × 120, 2 × 180, 3 × 300, and 3 × 600 s) lasting 65 min. Continuous arterial sampling was commenced at the start of the first frame for 10 min. Radioactivity from both continuous and discrete blood and plasma was assessed by gamma counting. Reverse-phase high-performance liquid chromatography with radiochemical detection was used to analyze the proportion of parent CHO and metabolite, [¹¹C]betaine, in the discrete samples. The “arterial input function” necessary for calculation of the kinetic variables of CHO uptake was derived as the plasma parent fraction.

Correction factors were applied to image data for attenuation, scattered radiation, electronic dead time, and random coincidences. Filtered back projection images were used for analysis. Regions of interest were drawn manually around tumor and normal tissues (normal breast and normal lung) using the Analyze software (Biomedical Imaging Resource, Mayo Foundation). The PET analysis was done in an unblinded manner (by K.C.). Semiquantitative analysis was done, and kinetic parameters were derived using Matlab (version 6.0, The Mathworks, Inc.) and in-house software. Standardized uptake values (SUV; ref. 28) were derived from decay-corrected time activity curves (at midframe time periods for each frame) and normalized for injected activity and body surface area (body surface area = 0.20247 × height (m)^0.725 × weight (kg)^0.425; University of Wisconsin).⁶

Specifically, we derived semiquantitative measures of CHO uptake including the average uptake at 5 to 15 min (SUV_5-15,av), the average uptake at 60 min (SUV_60,av), and the maximal voxel uptake at 60 min (SUV_60,max). The early and late time frames were selected to reflect transport and phosphorylation (with minimal metabolism) on one hand and retention of radioactivity on the other. In other tumor types, CHO is rapidly taken up by tumor within 5 to 10 min of injection and remains in steady-state up to 40 to 60 min after injection (16, 29).

A modified version of Patlak's (Graphical) analysis described previously for [¹⁸F]fluorothymidine (14) and thymidine imaging (30) was used to derive the rate constant for the net irreversible uptake of CHO from blood to form phosphocholine in tissue (Ki). The original Patlak technique (31) allows estimation of blood-to-tissue transfer constants from multiple time uptake data. This method is valid if there are no metabolites. If the parent tracer is significantly metabolized, as in the case of CHO to [¹¹C]betaine, then a correction factor is needed to exclude the contribution of metabolites in the exchangeable tissue space. This modification was used to derive Ki for CHO in tumor tissue as follows:

wherein V is

A is total tissue radioactivity (kBq/mL), K_x is Ki for CHO (mL plasma/s/mL tissue), C_Tot is total blood radioactivity (kBq/mL), C_x is radioactivity of parent compound determined by radio-high-performance liquid chromatography (kBq/mL), C_m is the radioactivity of metabolite (kBq/mL), τ is the elapsed time from time t = 0 (s), V_ox is steady-state space of exchangeable region occupied by parent, V_om is steady-state space of exchangeable region occupied by the metabolite, and V_b is blood volume. We compared the kinetic and semiquantitative variables of CHO in breast tumors. Tumor PET uptake variables were also compared with that in normal tissues.

We compared each tumor uptake variable to histologic grade as a measure of tumor aggressiveness, ER, progesterone receptor (PR), human epidermal growth factor receptor-2 (HER-2), and the number of involved axillary nodes (primary patients only). Tumor tissue was fixed in formalin and embedded in paraffin wax. Then 4-μm thick sections were cut with a microtome and stained with H&E. Grading of tumor was done using Elston and Ellis's grading system (32). ER and PR scores were graded using the McCarty's H-scoring system (33): 1+, weakly positive; 2+, moderately positive; 3+, strongly positive. The 4B5 (Rabbit Monoclonal Primary Antibody, Ventana Medical Systems, Benchmark XT System) antibody was used for HER-2 immunostaining, and scores were 0 (no staining), 1+ (faint membrane staining in >10% cells), 2+ (weak circumferential staining in >10% cells), and 3+ (strong and thick membrane staining). A positive HER-2 result was a score of either 2+ confirmed positive on fluorescence in situ hybridization or 3+. Anti–Ki-67 antibody, NCL-Ki-67-MM1 (Novocastra Laboratories), was used to stain primary tumor biopsies. The numbers of total and Ki-67–positive cells were then manually counted in eight randomly selected fields of view using a BX51 Olympus microscope (Olympus Optical) at ×400 magnification. In patients with metastatic disease, the original grade from their primary tumors was used for comparison with CHO uptake variables. It is well recognized that although expression of markers such as ER, PR, HER-2, and epidermal growth factor receptor may alter, the grade of breast tumors does not change in the transition from primary to metastasis (34). The comparisons of PET variables with histology were done on a lesion-by-lesion basis. Pathology reports were reviewed by a single, independent pathologist (S.S.).

The association between the PET variables and tumor grade, size, ER, PR, HER-2, Ki-67, and number of involved axillary nodes of primary cases was determined by Spearman's correlation test using GraphPad Prism (version 4.00 for Windows, GraphPad Software). The significance level was set at 5%, two-sided tests. P values of ≤0.05 were reported as significant.

A total of 32 patients were recruited (12 primary untreated, 20 metastatic) as described in Table 1. Their mean age was 52 years (range, 33-72) and the median tumor size imaged was 34 mm (range, 20-72). Tumor was not visualized in one lobular primary (patient 6) and one case with lung metastasis (patient 22). All other primary and metastatic tumors were visualized with good tumor-background ratios (Fig. 1A-D). Five primary patients (Table 1A) had pathologically involved nodes of which CHO PET detected three lesions correctly (patients 8, 11, and 12). Of the remaining two patients in whom PET failed to detect involved nodes, one (patient 1) had a single 10-mm involved node and the other (patient 5) had two nodes (both <8 mm in diameter). In patients with nodal disease, regions of interest were drawn around visible nodes in the axilla. In these patients, CHO uptake values were in general higher in the primary tumor than in involved axillary nodes, except in patient 12, in whom the metastatic axillary node had a marginally higher uptake (Table 1A); this was not a function of tumor size.

Two patients scanned (Table 1B) presented with de novo metastasis (patient 24 who had liver metastasis and patient 25 who had bone metastasis); the metastases were not assessed by PET scanning. Patient 23 (Fig. 1B) had a large inflammatory carcinoma involving the entire breast with metastasis in the axilla and bone. Other patients with metastases had been heavily pretreated with chemotherapy, hormonal theraphy, and/or radiotherapy but were still progressing. All metastatic lesions found on computed tomography were identified on PET except patient 22 whose basal lung metastasis was not visible on CHO PET. In total, 31 lesions (Table 1) were present in 20 metastatic patients, and thus, all analyses are on a per tumor basis. In three metastatic patients (patients 13, 21, and 23), relatively large differences in CHO PET uptake variables were seen between different tumors in the same patient. Two metastatic patients had biopsies of the metastasis taken just before PET (patients 13 and 15) and the grade matched that of the corresponding primary tumor.

CHO pharmacokinetics in breast tumors was characterized by a very rapid delivery such that steady-state was reached for all tumors within 5 to 10 minutes (within 1 minute for low uptake tumors). Within the bounds of experimental error, including patient movement, radioactivity levels then remained relatively constant for the duration of the scan. Typical examples of time-versus-radioactivity curves are shown in Fig. 2. Semiquantitative and kinetic variables were derived for tumor regions of interest (Fig. 3A and B). The median SUV_60,av was 9.1 × 10⁻⁵ m²/mL (range, 2.1-22.6); median SUV_5-15 was 9.3 × 10⁻⁵ m²/mL (range, 2.7-21.6); median SUV_60,max 16.8 × 10⁻⁵ m²/mL (range, 4.8-31.7); and median Ki was 12.5 × 10⁻⁴ mL plasma/mL tissue/s (range, 2.1-29.9). Notably, the median tumor CHO uptake values in stage IV metastatic disease were higher than that of stage II (Fig. 3; Table 1).

Imaging variables derived from normal breast tissue (contralateral normal breast in the case of primary breast cancers) or normal lung were found to be low compared with tumor. Compared with tumor, median Ki and SUV_60,av were ∼10-fold lower in normal breast (median Ki = 1.4 × 10⁻⁴, 95% confidence interval = 6.9 × 10⁻⁵ to 2.2 × 10⁻⁵ mL plasma/mL tissue/s; median SUV_60,av = 1.2 × 10^–5, 95% confidence interval 0.6 × 10⁻⁵ to 1.7 × 10⁻⁵ m²/mL) and normal lung (Ki = 3.1 × 10⁻⁴, 95% confidence interval = 1.8 × 10⁻⁴ to 4.4 × 10⁻⁴ mL plasma/mL tissue/s; SUV_60,av = 1.8 × 10⁻⁵, 95% confidence interval = 1.3 × 10⁻⁵ to 2.5 × 10⁻⁵ m²/mL).

Tumor grade and size are well-known prognostic indicators (35, 36). There was a poor association between PET variables and tumor size (Table 2). This may be due to the fact that tumor size per se is not a measure of malignant transformation and progression. Analysis of maximal voxel activity rather than average tumor activity seemed to improve the association of PET variables with tumor size, but the correlation coefficient was still below 0.5. Tumor grade was available for all primary cases and two metastatic cases. SUV_60,max correlated well with histologic grade of primary tumors (Table 2; P = 0.005, r = 0.70, Spearman's test), and grade also correlated well with imaging variables when primary and metastatic tumors were combined (the grade of the metastatic lesion was by definition the same as the original primary tumor); Ki had the best comparative correlation (P = 0.0002, r = 0.53, Spearman's test; Table 2). Overall, the association of imaging variables with grade increased in the order SUV_5-15 < SUV_60,av < SUV_60,max < Ki. We could not find any association of CHO PET variables and ER, PR, HER-2, or pathologically involved nodes (Table 2), a fact that is largely due to the patient inclusion criteria for this study. Ki-67 did not correlate with CHO uptake variables.

To our knowledge, this is the first report describing the use of CHO PET in breast cancer patients. Choline metabolism is elevated in the majority of tumors and accounts for increased phospholipid synthesis required for cell membrane turnover (37). In mammary epithelial cells, the increased intracellular phosphocholine levels is associated with malignancy, increasing with transformation and progression (17). The elevated phosphocholine levels in breast cancer cells may be due to elevated choline kinase-α expression (18, 38). Choline kinase-α has been identified as an oncogenic protein and may act in the G₁-S transition to promote cell division. The choline kinase-α–regulated genes that affect G₁-S transition have been reported recently (39). Increased choline metabolism has also been shown to be regulated by the growth factor receptor–MAPK pathway (17, 21, 22, 40), a pathway that is thought to play a central role in resistance to endocrine therapy (4, 6, 11). From these in vitro studies and by analogy to studies of choline kinase-α in biopsies from lung cancer patients (25), phosphorylation of choline to phosphocholine may have prognostic value in ER-positive breast cancer patients. To date, CHO PET has been explored largely in prostate cancer imaging (16, 41–43). This study investigates CHO PET imaging in breast cancer.

Median CHO uptake was ∼10-fold lower in normal breast and lung tissues compared with tumors and was higher in the more aggressive stage IV tumors compared with stage II tumors; we did not have enough patients in the other categories to make a comparison for all patients. This finding is in keeping with the in vitro findings of Aboagye and Bhujwalla, which showed elevated phosphocholine levels with transformation and progression (17). The low background and high signal-to-noise contrast suggests that CHO may be a useful imaging probe for breast cancer. At this stage, we are unable to report the relevance of CHO uptake to overall survival. Of note, patients with stage IV breast cancer showed a large range of uptake values. Although we were careful in selecting patients, this finding may be due in part to the confounding effect of previous treatment (as in metastatic cases) and may not completely reflect CHO uptake in de novo metastases. Regarding its sensitivity for routine diagnosis of cancer, CHO-PET did not detect two target lesions, one primary untreated lobular cancer (3 cm) and one (2 cm) metastatic lung lesion. These two lesions were seen clearly on magnetic resonance imaging and computed tomography, respectively. The lung lesion was located at the base of the lung and therefore subject to continuous respiratory movements, which may be the reason that it was not visualized. Other nontarget lesions (<2 cm) were not detected by CHO PET presumably because of their small size. This will suggest that the technique may be best used in conjunction with other imaging technologies, such as computed tomography.

We assessed a number of semiquantitative and fully quantitative CHO PET variables in this setting. SUV did not change significantly between the early imaging time point (5-15 minutes) and late imaging time point (60 minutes) despite the increased proportion of plasma [¹¹C]betaine levels over time (data not shown). We used grade as the gold standard for comparison of imaging variables. Of note, the CHO uptake was largely similar in different tumors within the same patient (by definition, same grade); although in few patients differences were seen between the primary tumor and involved axillary lymph node(s) consistent with clonal evolution. With respect to grade, SUV_60,max was the best imaging variable in primary tumors. For all tumors combined, the association of imaging variables with grade increased in the order SUV_5-15 < SUV_60,av < SUV_60,max < Ki. The fully quantitative estimate, Ki, was expected to be robust because it describes better the conversion of CHO to the intracellular metabolite [¹¹C]phosphocholine in tumor without contribution by the metabolite [¹¹C]betaine. Interestingly, although detailed kinetic modeling may theoretically better reflect tumor properties, the differences in correlation values with grade were not markedly better compared with SUV measures. No correlation with ER, PR, HER-2, or pathologically involved nodes was found; a fact that may be largely due to the selection of ER-positive cases. The absence of a correlation with Ki-67 (primary cases) is interesting and suggests that CHO uptake does not reflect proliferation.

Further to its prognostic value, CHO PET may also be useful as a pharmacodynamic end point for clinical evaluation of choline kinase-α inhibitors, which are currently in preclinical development (44–46), and potentially drugs that act on the MAPK pathway (21). In this context, a correlation between expression of choline kinase in the breast cancer specimens and the PET outcome would greatly improve the understanding of the findings and provide further support to the potential use as a pharmacodynamic noninvasive tool. Furthermore, it will be interesting to assess whether semiquantitative variables are still useful for monitoring drug response because the therapeutics may themselves alter the systemic metabolism of CHO. Two areas that need exploring further are application of this technology to tumors smaller than 2 cm in size to gain an improved understanding of sensitivity and specificity of these radiotracers and application in bone and liver metastases. These data have broad prognostic implications for use of PET scanning in breast cancer to visualize tumor biology in vivo. In conclusion, CHO showed good uptake in most breast cancers and merits further investigation as a breast cancer imaging agent.

No potential conflicts of interest were disclosed.

We thank Sian Thomas and Helena Marconell for help in recruiting patients for this study and all the chemists and radiographers from GE-IMANET for their help in radiotracer production and PET scanning.