[18F]FDG and [18F]FLT PET for the evaluation of response to neo-adjuvant chemotherapy in a model of triple negative breast cancer

Rationale Pathological response to neo-adjuvant chemotherapy (NAC) represents a commonly used predictor of survival in triple negative breast cancer (TNBC) and the need to identify markers that predict response to NAC is constantly increasing. Aim of this study was to evaluate the potential usefulness of PET imaging with [18F]FDG and [18F]FLT for the discrimination of TNBC responders to Paclitaxel (PTX) therapy compared to the response assessed by an adapted Response Evaluation Criteria In Solid Tumors (RECIST) criteria based on tumor volume (Tumor Volume Response). Methods Nu/nu mice bearing TNBC lesions of different size were evaluated with [18F]FDG and [18F]FLT PET before and after PTX treatment. SUVmax, Metabolic Tumor Volume (MTV) and Total Lesion Glycolysis (TLG) and Proliferation (TLP) were assessed using a graph-based random walk algorithm. Results We found that in our TNBC model the variation of [18F]FDG and [18F]FLT SUVmax similarly defined tumor response to therapy and that SUVmax variation represented the most accurate parameter. Response evaluation using Tumor Volume Response (TVR) showed that the effectiveness of NAC with PTX was completely independent from lesions size at baseline. Conclusions Our study provided interesting results in terms of sensitivity and specificity of PET in TNBC, revealing the similar performances of [18F]FDG and [18F]FLT in the identification of responders to Paclitaxel.

Pathological response to neo-adjuvant chemotherapy (NAC) represents a commonly used predictor of survival in triple negative breast cancer (TNBC) and the need to identify markers that predict response to NAC is constantly increasing. Aim of this study was to evaluate the potential usefulness of PET imaging with [ 18 F]FDG and [ 18 F]FLT for the discrimination of TNBC responders to Paclitaxel (PTX) therapy compared to the response assessed by an adapted Response Evaluation Criteria In Solid Tumors (RECIST) criteria based on tumor volume (Tumor Volume Response).

Methods
Nu/nu mice bearing TNBC lesions of different size were evaluated with [ 18 F]FDG and [ 18 F] FLT PET before and after PTX treatment. SUV max , Metabolic Tumor Volume (MTV) and Total Lesion Glycolysis (TLG) and Proliferation (TLP) were assessed using a graph-based random walk algorithm.

Results
We found that in our TNBC model the variation of [ 18 F]FDG and [ 18 F]FLT SUV max similarly defined tumor response to therapy and that SUV max variation represented the most accurate parameter. Response evaluation using Tumor Volume Response (TVR) showed that the effectiveness of NAC with PTX was completely independent from lesions size at baseline. PLOS

Introduction
Breast cancer (BC) is a heterogeneous disease composed of several biological subtypes having different clinical course, response to therapy and molecular profile. The lack of expression of Estrogen Receptor (ER), Progesterone Receptor (PR), Epidermal Growth Factor Receptor 2 (HER2) and the absence of HER2 amplification define the TNBC [1]. TNBC represents approximately 15-20% of all invasive breast cancers and is characterized by ductal histology, high mitotic rates and earlier lymph node involvement when compared to other BC subtypes [2]. TNBC is frequently associated to high expression of proliferation markers as Ki67 and cyclins and activation of the beta-catenin pathway [3].
High aggressiveness, as well as non-susceptibility to hormone and targeted therapies, limits the number of therapeutic opportunities and makes the prognosis of TNBC patients poor. NAC with anthracyclines and the mitotic inhibitors taxanes used in sequential or combined treatment, represents the standard pharmaceutical approach for TNBC [4,5,6] and describes therapeutic interventions prior to surgery to reduce size of unresectable tumors and test therapies efficacy. Despite its intrinsic aggressiveness, TNBC is highly responsive to NAC, a phenomenon called "triple negative paradox" [4,6]. Unfortunately, those patients who do not achieve pathological complete response (pCR) present a high rate of relapse. Therefore, much research is focused on the development of biomarkers predictive of clinical response, avoiding the use of ineffective protocols and customizing the optimal strategy. Traditionally, treatment response has been assessed through the application of RECIST, which classifies effectiveness on the basis of tumor shrinkage, using anatomical measurements. However, this parameter represents a later event compared to other changes which may be triggered by treatments [7]. PET allows the non-invasive monitoring of biological aspects related to tumor growth and aggressiveness, like glucose metabolism, cell proliferation and hypoxia [8]. In different types of cancer, the radioligand 2-deoxy-2-[ 18 F]fluoro-D-glucose ([ 18 F]FDG) has been reported as useful tool for early prediction of response or resistance to pharmacological treatment [9]. Considering TNBC, a reduction of [ 18 F]FDG uptake after two cycles of neo-adjuvant chemotherapy has been recently proposed as a powerful marker of patients' outcome [10,11,12], but preclinical as well as clinical studies identified other tracers of potential interest. Among these, the thymidine analogue 3'-[ 18 F]fluoro-3'-deoxythymidine ([ 18 F]FLT) seems to be a potential indicator of tumor response/resistance to therapy [13,14,15]. In fact, the uptake of [ 18 F]FLT reflects the activity of the enzyme thymidine kinase-1 (TK1), well known for its function in the pyrimidine salvage pathway. This enzyme is upregulated during late G1/S phase of the cell cycle, thus representing an indirect marker of cell proliferation.
The high basal [ 18 F]FDG uptake and rate of cell proliferation make TNBC an adequate subtype of BC to investigate response assessment with PET. Many studies have been performed to compare the effect of repeated chemotherapy on [ 18 F]FLT and [ 18 F]FDG uptake [16,17,18,19,20,21,22], but data on TNBC are not conclusive. In this study, we aimed to evaluate and compare the effect of NAC with taxane on [ 18

Animal experiments
All animal experiments were carried out in compliance with the institutional guidelines for the care and use of experimental animals, which have been notified to the Italian Ministry of Health and approved by the ethics committee of the San Raffaele Scientific Institute. Female SCID Hairless Congenic (SHC™) mice (Charles River, Italy) of 6-8 weeks of age were subcutaneously implanted on the back with 1.5 x 10 7 (n = 24) or 2 x 10 7 (n = 14) MDA-MB-468 cells under ketamine/xylazine anaesthesia (i.p., 100 mg/kg / 10 mg/kg). Animals were housed in the animal facility of San Raffaele Scientific Institute and daily monitored for body weight and lesions sprouting; tumor volume was measured with digital calliper twice a week and expressed as (L x l 2 )/2 = (mm 3 ) where L is the long side and l is the short side. Moreover, when tumours reached diameters of more than 15 mm or when mice showed signs of severe illness, they were euthanized by cervical dislocation under isoflurane anaesthesia.

Treatment protocol
PTX was prepared dissolving the drug powder in the vehicle solution: 90% saline, 5% ethanol and 5% Cremophor (Sigma Aldrich S.r.l., Italy). Tumors smaller than 150 mm 3 (small tumors, n = 12) or larger than 150 mm 3 (large tumors, n = 14) were randomized into two groups and treatment started with vehicle (control) or Paclitaxel (treated, 18 mg/kg i.v., two doses per week) for two weeks. Treatment response was evaluated using [ 18 F]FDG and [ 18 F]FLTPET scans, before (baseline) and at the end of treatment. The efficacy was determined according to the RECIST score adapted to the experimental procedure [23]. Indeed, since the standard monitoring of tumor in preclinical setting is usually performed by volume measurement, an adapted RECIST score was used in the study. This index was defined as Tumor Volume Response (TVR) and calculated as the percentage change in median tumor volume measured by calliper at the end of treatment over the median tumor volume before treatment. According to this definition, treatment response was calculated as Partial Response (PR) (TVR, score at least > -30%); Stable Disease (SD), (TVR, score < -30% and < +20%) and Progressive Disease (PD), (TVR score > +20%) [24]. [ 18 F]FDG, prepared for clinical use (European Pharmacopeia VIII Edition), and [ 18 F]FLT [25] were injected with a radiochemical purity > 99%. PET acquisitions were performed as previously described [13]. Identification of hypermetabolic or hyperproliferative lesions was performed using a segmentation method [26], adapted for preclinical use. Briefly, an algorithm based on Random Walks (RW) on graphs has been used to convert DICOM (Digital Imaging and Communications in Medicine) images into a graph where some nodes are known (nodes with target or background label) and others unknown. PET image is then converted in a lattice where voxel SUVs are assigned to corresponding graph nodes and edge weights are computed accordingly. A probability map is then produced, and a threshold p is chosen to discriminate between target and background voxels. Tracers' uptake was expressed as:

PET evaluation
• standardized uptake value (SUV = [radioactivity in the tumor/injected radioactivity] Ã animal weight); • metabolic tumor volume (MTV = volume (mm 3 of the VOI after segmentation); • total lesion glycolysis (TLG) for [ 18 F]FDG or total lesion proliferation (TLP) for [ 18 Variations in all parameters in sequential scans were normalized to baseline and expressed as percentage of variation (% change) according to the following formula: %change ¼ 100 x ðpost À treatment À pre À treatmentÞ=pre À treatment:

Histological and immunohistochemical analyses
Twelve of the twenty-four female SCID mice implanted with 1.5 x 10 7 MDA-MB-468 cells were treated with PTX (n = 6) or vehicle (n = 6) were sacrificed for histological (H&E) and immunohistochemical (IHC) analyses for Ki67, as already described [27]. Proliferation index (P.I.) was evaluated for each tumor considering the whole number of Ki67 positive nuclei over the whole number of cell nuclei in three randomly selected fields.

Statistical analysis
Data generated were expressed as percentage change between the end and the baseline of treatment, mean value with standard deviation (mean±S.D.). Prism 4 (GraphPad Software Inc., San Diego, CA, USA) was used for the statistical analysis. Parameters of radiotracer uptake were assessed and compared through the Student T-test or the ANOVA test using Bonferroni's multiple comparison; p was considered statistically significant, when < 0.05. The accuracy of PET parameters was evaluated by carrying out the Receiver Operating Characteristic (ROC) analysis in defining the pathological response.

Tumor weight after treatment correlates with Ki67 expression
We firstly evaluated in a separate group of mice bearing MDA-MB-468 cells the effect of Paclitaxel on Ki67 proliferation marker which is used in clinical practice to assess neo-adjuvant chemotherapy [28]. No animal died because of the experimental procedures or showed signs of illness during tumor growth. The results clearly indicate a reduction of Ki67 staining as a consequence of PTX treatment. Moreover, the weight of harvested tumors (mg) significantly correlated with the corresponding Ki67 expression level (Fig 1).

Response of MDA-MB-468 tumors to PTX was independent from the initial size
To better represent tumor variability and to mimic the heterogeneity of the human disease, mice which underwent PET evaluations were inoculated with different concentrations of MDA-MB-468 cells and treatment started when tumors reached a volume smaller than 150 mm 3 (76.7 ± 35.7 small tumors, n = 12), or more than 150 mm 3 (236.8 ± 107.5 large tumors, n = 14). After the whole PTX cycle, treated animals displayed a significant decrease in tumor volume, when compared to animals receiving vehicle (p = 0.018) (Fig 2). In addition, the response to treatment resulted independent from tumor size at the beginning of treatment. Indeed, applying the TVR for the evaluation of response to PTX therapy, a PR was observed in 33% of small tumors and in 29% of mice bearing large tumors. Similarly, 33% and 43% of mice bearing small and large tumors respectively exhibited SD. Finally, a comparable number of mice bearing small (33%) or large tumors (29%) showed an increase in lesions volume being defined as PD (Table 1), indicating that MDA-MB-468 tumors response to PTX is independent from the initial lesion size.

Δ[ 18 F]FDG and Δ[ 18 F]FLT SUV max are similarly influenced by PTX treatment
PTX treatment determined a reduction of both [ 18 F]FDG and [ 18 F]FLT uptake, which were found to be only slight for SD and more marked for PR, as shown in PET images (Fig 3). F]FDG, with a significant reduction in PR (-62.56% ± 45.1%, p = 0.039), a stable trend in SD (7.74% ± 39.7%) and variable but not significant changes in PD (+7.91% ± 37.4%). FDG SUV max variations appeared significantly different between partial responders and nonresponders, that included both stable and progressive disease with statistical significance (p = 0.003). In detail, [ 18 F]FDG SUV max decrease in PR was significantly higher than that of vehicle (p = 0.0001, Fig 4) and that of PD and SD considered alone (p = 0.024 and p = 0.030 respectively, Fig 4) while [ 18 F]FLT SUV max decrease in PR was significantly higher only than that of vehicle and SD (p = 0.026 and p = 0.049 respectively, Fig 4). PTX treatment induced also a comparable reduction, although not significant, of both MTV and TLG or TLP, indicating that [ 18

SUV max variations represent a better parameter to evaluate response to therapy
Our data indicated that variations of [ 18 F]FDG SUV max offered a better accuracy in defining response to NAC with PTX and in differentiating pathological partial responders from nonresponders. The area under the curve (AUC) of ROC curves for [ 18 (Fig 5). According to ROC analysis, a cut-off value of -80.4% offered for [ 18 F]FDG ΔSUV max the best accuracy in predicting non-responder lesions, with a sensitivity and specificity of 89% and 75%, respectively. ΔSUV max for [ 18 F]FLT was also an accurate prognostic factor leading to an optimal cut-off value of -70.7%, (100% sensitivity and 50% specificity), but resulted inferior to [ 18 F]FDG.

Discussion
The objective of this study was to evaluate PET as an accurate tool to discriminate TNBC treatments' responders. With this purpose, we used SCID mice bearing human MDA-MB-468 lesions of different size and, after classification of responders using an adapted RECIST criteria based on volumetric measurement of tumors, we evaluated response to PTX treatment comparing in the same set of mice [ 18 F]FDG and [ 18 F]FLT PET. Several breast cancer cell lines are currently used to study triple negative tumours; we took advantage of MDA-MB-468 which has been identified as ER-, PR-and HER-basal breast cancer cell, as approximately the 80% of TNBC [29]. Moreover, MDA-MB-468 cells display high Ki67 and EGFR expression and form cohesive grape-like or stellate structures consistent with the more invasive phenotype characterizing the TNBC human situation [30,31].   Tumor response to PTX treatment appeared variable in our study, revealing a high heterogeneity of volume variations, which was independent from the initial lesion size. The histopathological characteristic of MDA-MB-468 tumor and its typical microenvironment might act on PTX distribution and efficacy. Indeed, the presence of poor vascularized sub regions within the tumor although mimicking the clinical situation might influence PTX response [32].
Although TNBC represents an invasive and highly aggressive subtype of BC, using pCR as a surrogate endpoint, there are evidences that TNBC is a chemo-responsive disease [4]. However, while patients with TNBC responding to NAC have a relatively good prognosis, those without response display an extremely poor outcome, with a higher risk of relapse [4]. Hence, the possibility to evaluate the early efficacy of NAC is of fundamental importance for the clinical management of patients, tailoring the best treatment option on the basis of the initial response. NAC for TNBC, which is usually performed with a combination of taxanes and anthracyclines, has been performed with taxanes alone to focus the study in understanding changes in glucose metabolism and proliferation as potential markers of TNBC responsiveness. Indeed, [ 18 F]FDG and [ 18 F]FLT PET have been used to evaluate changes in glucose metabolism and proliferation triggered by treatment in our model of TNBC, which has been known to not show an inflammatory phenotype that could produce a bias in the interpretation of the results obtained using [ 18 [30,31], the effect of RGD-PTX seemed to be not significantly related neither to [ 18 F]FDG nor to [ 18 F]FLT [21]. In our TNBC model PTX treatment clearly demonstrated an effect on proliferation as depicted by the significant correlation between tumor reduction and Ki67 reduction. Nevertheless, MDA-MB-468 bearing mice performing PET imaging displayed that [ 18 F]FLT variations were not more indicative than [ 18 F]FDG SUV max variations in defining response to therapy.
Only few data are available to support the use of [ 18 F]FLT as a marker of TNBC response to NAC, although it can provide higher specificity, since its accumulation in inflammatory areas is less significant than [ 18 F]FDG [33]. It has been demonstrated that [ 18 F]FLT PET is able to detect therapy-induced proliferation changes as early as 1 week after FEC (5-fluorouracil, epirubicin, cyclophosphamide) chemotherapy, discriminating between responders and SD patients with stage I-IV breast cancer [34]. In another study, the predictive value of changes in [ 18 F]FLT SUV after the first cycle of chemotherapy was demonstrated in patients with metastatic breast cancer [35]. Monitoring response to NAC therapy is of great importance since it allows the early switch for alternative treatment. Moreover, in a small population of locally advanced BC patients, Crippa et al. demonstrated the good sensitivity, specificity and AUC of [ 18 F]FLT PET for early monitoring of response after a single cycle of NAC [36]. On the other hand, in a heterogeneous population of primary BC patients, [ 18 F]FLT revealed only a marginal predictive value of therapeutic response after one cycle of NAC, displaying a good AUC despite highly variable chemotherapy regimens [37]. In a similar way, Woolf reported a reduction of [ 18 F]FLT SUV max after a single cycle of NAC, but they demonstrated that neither the baseline value, nor the variation of SUV max after therapy was able to predict treatment response in an heterogeneous population of primary BC [38]. Recent studies report that TNBC tumors usually have higher metabolic activity than those of other phenotypes, being aggressive and [ 18 F]FDG avid [39]. Indeed, [ 18 F]FDG uptake has been largely used to efficiently distinguish TNBC patients responsive to different NAC therapies, compared to other subtypes such as HER-2 positive tumors [40,41,42]. The nature of PET images (i.e., low spatial resolution, high statistical uncertainty noise, and ill-defined margins) renders the MTV and TLG/TLP quantification particularly hard. Moreover, the inaccuracy of visual delineation of tumor is subjected to both intra and inter-operator variability. In order to avoid these limitations, we took advantage from a graph-based algorithm for MTV delineation [26] which differs from conventional approaches since it is more accurate in noisy and low contrast images. Evaluation of tumour response to PTX has been made according to the RECIST criteria adapted to the use of volumetric measurements and it has been defined as tumor volume response (TVR), considering the reduction of tumour volume instead of that of longest diameter [23,43]. Relying on the categorization of responders using TVR evaluation, we observed a general and similar decrease of both [ 18  Many studies apply ROC curves to define an optimal threshold value of radiotracers uptake, able to discriminate responders [14,36,41,42]. The differences in the published threshold value are caused by several factors, which have to be taken into account, such as the definition of good pCR, the time of PET, or the chemotherapy regimen. It has been noted that the use of different therapeutic agents may affect [ 18 F]FLT uptake regardless the direct effect on proliferation, depending on their influence on nucleoside metabolism and on cell cycle [22].
The use of Paclitaxel on a chemotherapeutic regimen has been shown to have minimal effect on [ 18 F]FLT uptake, since it induces cell cycle arrest in an advanced point which doesn't affect TK1 expression or change cell proliferation, even though it reduces tumor growth [22]. Moreover, it has to be considered that other mechanisms, including the use of salvage or de novo pathway for DNA synthesis, could influence [ 18 F]FLT uptake [14].

Conclusions
In conclusion, many works have been performed investigating the role of [ 18