Detection of alterations in membrane metabolism during neoadjuvant chemotherapy in patients with breast cancer using phosphorus magnetic resonance spectroscopy at 7 Tesla

Breast cancer is by far the most prevalent type of cancer in women worldwide (Jemal et al. 2011). Surgical breast cancer treatment is sometimes preceded by a neoadjuvant therapy such as chemotherapy, endocrine therapy, and or targeted therapy. Neoadjuvant therapy is applied in patients with inoperable inflammatory or locally advanced breast cancer and nowadays also in patients with operable breast cancer to reduce tumor volume to enable breast conserving surgery (Liu et al. 2010). Although reduction of the tumor volume is achieved in the majority of patients, a pathologic complete response is only seen in less than 20% of patients according to a recent meta-analyses which included almost 3800 breast cancer patients receiving neoadjuvant therapy (Kong et al. 2011). Because of this low pathologic complete response that is often used to predict survival rate, there is a need for early assessment of neoadjuvant chemotherapy efficacy, which would enable switching to a possibly more successful agent early in therapy. In contrast to imaging, that monitors the tumor’s morphology alterations during neoadjuvant therapy, the monitoring of tumor metabolism may offer a faster and better window to assess therapy efficacy. Phosphorus magnetic resonance spectroscopy (³¹P MRS) is probably the most suitable tool to achieve this, because the metabolites of interest are phosphorus compounds such as the phosphomonoesters (PME) phosphocholine (PC) and phosphoethanolamine (PE) and the phosphodiesters (PDE) glycerophosphocholine (GPC) and glycerophosphoethanolamine (GPE) that are involved in cell membrane anabolism and catabolism respectively (Podo 1999; Glunde et al. 2004). In addition to these metabolites, signals of mobile phospholipids glycerophosphatidylcholine (GPtC) and glycerophosphatidylethanolamine (GPtE), but also metabolites involved in energy metabolism such as inorganic phosphate (Pi) and adenosinetriphosphate (ATP) can be measured with ³¹P MRS.

Although hampered by a relatively low intrinsic sensitivity, ³¹P MRS offers substantial benefits above proton (¹H) MRS in monitoring tumor metabolism. In vivo measurement of these individual ³¹P metabolites is feasible with ³¹P MRS, while in vivo proton MRS can only assess a ‘total-choline’ signal that comprises the sum of monoesters and diesters. In contrast to ¹H MRS, that necessitates water suppression that may bias the total choline signal, phosphorous MRS doesn’t require water suppression. Moreover the chemical shift of Pi, as measured in ³¹P MRS may serve as a proxy for intracellular pH and tumor hypoxia.

The first clinical application of ³¹P MRS in monitoring neoadjuvant chemotherapy in breast cancer patients was reported in 1989 by Ng et al. (1989), after which a number of reports followed (Glaholm et al. 1989; Redmond et al. 1992; Twelves et al. 1994; Leach et al. 1998). All of these studies were performed at low magnetic field, having insufficient spectral resolution to resolve the individual phosphomonoester (PE, PC) signals or individual phosphodiester (GPE, GPC and mobile GPtE and GPtC) signals. Moreover, the sensitivity at these lower field strengths could only detect MR spectra from very large tumors, often involving almost the entire breast. Nonetheless, these previous studies report increased PME signal prior to therapy and decreasing PME during therapy for responders and partial responders. After the publication of Leach et al. (1998), it took 13 years to improve ³¹P MRS in breast cancer to facilitate clinical studies in more relevant tumor sizes. First of all, the field strength was increased from 1.5 T to 7 T, and efficient quadrature RF coils were designed dedicated to the human breast (Klomp et al. 2011). All together this improved sensitivity by an order of magnitude; a factor of 2 from 1.5 T to 3 T and more than a factor of 5 from 3 T to 7 T using the dedicated breast coil (Korteweg et al. 2011). Moreover, improved MRS sequences were developed that include spatial localization techniques and means to maximize sensitivity even further by using adiabatic multi-echo spectroscopic imaging (AMESING) (van der Kemp et al. 2013).

In this work we investigate whether state of the art ³¹P MRSI is sufficiently sensitive to pick up PME alterations in the primary tumor in patients with breast cancer during the course of neoadjuvant chemotherapy.

Within a time frame of 17 months, six patients participated in this study. One patient dropped out after the first scan session due to claustrophobia. Cancer type and receptor status of the breast cancers were assessed from core needle biopsies. Two patients of the group who successfully concluded this study were diagnosed with invasive ductolobular carcinoma, two with invasive ductal carcinoma, and one with medullary carcinoma. Three patients had receptor status ER +, PR +, Her2neu – and two were triple negative. Histopathologic assessment showed that the patient group that concluded the study (n =5) consisted of one full responder and four partial responders. After chemotherapy the full responder still had a fibroadenoma surrounding the tumor location.

The FID spectra during neoadjuvant chemotherapy for the patient group are shown in Figure 2. For comparison, a FID spectrum of a healthy volunteer group consisting of seven healthy volunteers measured four times and with a four times larger nominal voxel size is shown as well (Stehouwer et al. 2014). There are a number of distinct features that distinguish the presented breast cancer spectra from the healthy volunteer group spectrum. These distinct features are most prominent when comparing the patient spectrum acquired prior to chemotherapy with the healthy volunteer group spectrum. Prior to chemotherapy the patient group spectrum shows higher: PE than Pi, PE than GPtE + GPC, and PE than GPtC. The signals of γ-ATP and α-ATP are higher than the PDE signals. In the healthy volunteer group spectrum theses patterns are reversed. During chemotherapy (half way and after chemotherapy) the patient group spectra are showing more resemblance to the healthy volunteer group spectrum.

In Figure 3 the PME to PDE signal ratios (PME/PDE, PE/(GPtE + GPC), PC/GPtC) and the PME to Pi metabolic signal ratios, as obtained from spectral fitting the FID group spectra, are shown during the course of chemotherapy, also in comparison to the healthy volunteer data by Stehouwer et al. (2014). Except for PC/GPtC, with the largest error bars, all metabolite signal ratios of the patient group PE/(GPtE + GPC), PME/PDE and PME/Pi show the same decreasing pattern during the course of chemotherapy and all normalize to values observed in healthy volunteers.

The metabolic signal ratios of the averaged group spectra shown in Figure 3 are based on average group signals (no physiological variability) and system noise (i.e. Cramer Rao lower bounds (De Graaf 2007) is represented by the error bars. In Figure 4 the PME over PDE ratios are shown as an average based on the individual patient data –including physiological variation as an error bar– and averaged group data with system noise as an error bar. The largest physiological variation between patients is seen in the PME over PDE ratios prior to chemotherapy as may be expected from this heterogeneous patient group with three ER+/PgR + and two triple negative patients. In preclinical work and animal studies it has been shown that triple negative breast cancers have a low PME to PDE ratio in contrast to ER+/PR + tumors which have a high PME to PDE ratio (Moestue et al. 2010; Esmaeili et al. 2014).

Patient signal ratios of PME over Pi relative to healthy glandular tissue are shown during the course of chemotherapy, in Figure 5. Here, a ratio of 1.0 implies that patients PME over Pi signal ratio equals that of healthy glandular tissue.

Prior to neoadjuvant chemotherapy, four out of five patients show significantly higher PME over Pi signal ratios than encountered in healthy glandular tissue. After chemotherapy, one out of five patients shows a significantly lower PME over Pi signal ratio than encountered in healthy glandular tissue, while the other four patients have PME over Pi signal ratios similar to those in healthy glandular tissue. The PME over Pi ratio is a tissue viability marker that can be used in for instance assessing kidney quality for transplantation (Bretan et al. 1989), where a high ratio designates viable tissue. In radiation therapy a decrease of this ratio is an early marker for tumor cell apoptosis (Sakurai et al. 1998). However, pre-thermoradiotherapy PME to Pi ratios in sarcomas inversely correlate to pathologic complete response (Dewhirst et al. 2005).

Calculated metabolite T₂ values with 95% confidence intervals are shown in Figure 6 during chemotherapy and for healthy volunteers.

Metabolite T₂ values fitted from low-SNR data are prone to be overestimated, as is evident from the asymmetric confidence intervals in Figure 6. Significantly lower T₂ values than in healthy volunteers are seen for PE(HW), PC(P), Pi(P, HW, A) and GPtE + GPC(HW). Overall there is a trend for lower ³¹P metabolite T₂ values in these breast cancer patients, as compared to the metabolite T₂ values in healthy fibroglandular tissue. This matches the lower T₂ values of water in prostate tumors as reported by Roebuck et al. (Roebuck et al. 2009). The significant higher T₂ for the (GPtE + GPC) signal prior to chemotherapy is possibly caused by increased GPC prior to chemotherapy in this patient group, that includes two triple negative breast cancer patient cases. It is likely that GPtE has a T₂-value similar to GPtC, i.e. in the order of 20 ms, while GPC is likely to have a T₂ value that is in the range of the phosphomonoesters or even higher. In calf muscle for instance it exceeds 300 ms at 7 T (Bogner et al. 2009). This implies that on average the phosphodiester GPC signal (and concentration) is higher prior to chemotherapy and decreases during therapy in this patient group, which is contrary to the general, probably too simplistic, paradigm of increasing glycerophosphocholine and -ethanolamine during successful chemotherapy. This is corroborated by a report on the difference in choline metabolic profile of basal like and luminal like breast cancer xenografts by Moestue et al. (2010), where it was found that triple negative/basal like breast carcinomas had high GPC/PC ratios.

For the patient group as a whole the average chemical shift of Pi shows an increase of +0.14 ± 0.05 ppm over the course of the chemotherapy, which corresponds to a shift in pH of 0.19 units, implying a more acidic environment before the start of the chemotherapy, possibly indicative of tumor hypoxia or lactate formation.

In conclusion, using state of the art ³¹P MRSI provides sufficient sensitivity to detect statistically significant phospholipid alterations in primary breast tumors during neoadjuvant chemotherapy, despite the limited and heterogeneous sample size. Moreover, the ³¹P spectrum of the patient group obtained prior to treatment is distinctly different from the ³¹P spectrum of glandular tissue in the healthy volunteer group and transforms its shape during the course of the chemotherapy towards the shape of the spectrum of the healthy volunteer group. In line with our prospective hypothesis, the elevated phosphomonoesters in breast tumors decrease during the course of the chemotherapy. In addition, due to the high spectral resolution at 7 T, a subtle tumor pH increase of 0.19 units is observed during the treatment, possibly indicating tumor hypoxia or lactate formation.