Carbon ion radiotherapy: impact of tumor differentiation on local control in experimental prostate carcinomas

Regarding the effectiveness of carbon ion beams, most systematic experimental investigations have been performed in vitro [1‐3] or in normal tissue complication models in vivo [4, 5] and only very few quantitative data on the response of different tumor types to carbon ion beams are currently available [6‐9]. As in treatment planning for photon radiotherapy, tumors are still considered as biologically homogeneous entities and the relative biological effectiveness (RBE) is computed by biomathematical models [10‐12] to adjust for the dependencies of the RBE on linear energy transfer (LET) and dose. Biological response characteristics of the tumor are described by very few parameters neglecting the impact of additional tumor-associated biological factors on the RBE. However, any parameter that influences the tumor response differently for photons and ion beams is expected to influence the RBE. Therefore, to better understand the tumor response to ion beams, these dependencies have to be identified and assessed quantitatively in systematic preclinical experiments.

In previous studies, we determined the dose response curves for the anaplastic prostate carcinoma R2327-AT1 [13, 14] after 1, 2 and 6 fractions of photons and carbon ions, respectively, for the clinically relevant endpoint “local tumor control at 300 days”. It was found that the dependence on fractionation was much weaker than for normal tissue [4] resulting in smaller RBE-values at low fractional doses as compared to normal tissue. Although two other tumor types showed similar RBE-values at the same LET [7, 8], no systematic investigation of the RBE of tumors varying with respect to growth kinetic and differentiation status have been performed yet.

Within the translationally oriented clinical research group KFO 214 on heavy ion therapy, funded by the German Research Foundation (DFG), the radiation response of a well (-H), moderately (-HI) and poorly (-AT1) differentiated subline of the R2327 prostate carcinoma model was investigated in terms of dose-response curves for the endpoint “local tumor control at 300 days”. As it was found that the radiation responses of the three tumor lines differ much less for carbon ions than for photons, potential influence factors were further investigated by radiological imaging as well as on the histological and molecular level. Furthermore, the relationship between tumor microenvironment and local tumor control and its dependence on radiation quality was evaluated by dedicated irradiation experiments. This contribution gives a brief summary of the overall project and adds to previously published data [6, 15, 16].

Preclinical studies in normal tissues are preferentially performed to evaluate potential side effects of carbon ions and to validate RBE-models. In contrast, tumor experiments aim to decipher biological factors, which influence the tumor response differently for photons and ion beams, and to identify, which tumor entities might benefit most likely from high-LET irradiations. In this context, a systematic study was initiated to quantitatively assess the treatment response of three different tumor lines to photons and ¹²C-ions using a local tumor control assay.

In summary, the following clinically relevant results (Fig. 1) were obtained: (i) For photons, a considerably heterogeneous treatment response was found, documented by a broad range of TCD₅₀-values for the three tumor-sublines. (ii) For carbon ions, the respective dose-response curves were located much closer to each other. (iii) In addition, the slope of the dose-response curve for each tumor subline was comparable or steeper for ¹²C-ions than for photons, and (iv) the resulting RBE increased with tumor grading (i.e. H vs. HI vs. AT1). This increase of RBE predominantly results from a rise of TCD₅₀ with tumor grading in photon treatments while the variation of the treatment response to ¹²C-ions is only small. This supports the conclusion that certain tumor-associated factors might be responsible to render tumors more resistant to photons than to ¹²C-ions. Clearly, these factors are depending on tumor grade. Moreover, also intra-tumoral heterogeneity seems to possess minor impact as documented by the increased slope of the dose-response curve of ¹²C-ions for the very heterogeneous HI-subline as compared to the respective curve for photons. These results allow the conclusion that the response to ¹²C-ions is also less dependent on intra-tumor heterogeneity. Regarding the effectiveness, the highest RBE of ¹²C-ions can be expected for undifferentiated tumors, showing the highest resistance against photon irradiations. A first report on prostate cancer patients in Japan confirmed our results showing very high tumor control rates with reduced toxicity and a comparable 5-year local control rate for carbon ions between low, intermediate and high-risk prostate cancer patients [20].

From a technical point of view, assessment of local control was most difficult in the slow-growing and well differentiated H-tumor because of frequently occurring residual tissue nodules at the end of the follow-up time. This problem was solved by additional histological analysis using lack of proliferative activity within these nodules as secondary endpoint. Interestingly, as the corresponding TCD₅₀-values increased for both, photons and ¹²C-ions, there was only a minor difference in RBE and the above conclusion remains unchanged [6].

While this report refers to single dose irradiations only, the identical study was conducted for 2 and 6 fractions, already published for the AT1-tumor [14]. Although still under evaluation for the HI- and the H-tumor, there is a clear trend that fractionation increases the TCD₅₀-values in all three tumor cell lines and both irradiation modalities. Again, the shift is larger for photons than for carbon ions, indicating an increasing RBE with decreasing dose per fraction and decreasing differentiation status. The highest RBE for 6 daily fractions (2.67 ± 0.15) was found for the anaplastic AT1-subline [14]. Details on the complete fractionated studies, including the dose dependence of the RBE and the determination of α/β-ratios will be published separately. An interesting side observation of the published study [14] was that in the fast growing AT1-tumor the metastatic rate increased, when the number of fractions raised from 2 to 6. Yet, at least for the given treatment schedules (1, 2 and 6 fractions) the results were not dependent on radiation quality [21].

There is significant evidence in the literature that resistance to photon therapy is associated with both, intrinsic cellular factors conditioned by the evolutionary capacity of cancer phenotypes as well as epigenetic parameters, or the temporal and spatial heterogeneity of the tumor microenvironment caused by structural abnormalities and density of tumor microvessels, dysfunctional blood flow, low pH leading to either chronic or acute hypoxic conditions [22‐25].

For further clarification, a detailed structural and functional characterization of all three tumor lines prior to irradiation was undertaken. As highly aneuploidic subpopulations were present in all three tumor lines, the ploidic status was not considered as a relevant tumor-associated intrinsic factor for the differential radiation response [15]. In contrast, differences were detected with respect to putative cancer stem-like cells characterized as CD24⁺/CD45⁻ cells, which were positively tested for the ability to form new tumors in a functional limiting dilution assays (Glowa et al., unpublished data). The fact that stem-cell properties were detected in H- and HI- but not in AT1-tumors needs further analysis which is presently ongoing.

Dramatic differences were found with respect to the structure and quality of the tumor vascularization and in correlation with the tumor microenvironment, inasmuch as a range of differently oxygenated tumors were detected, with the highest hypoxic fraction in the poorly differentiated AT1-tumors and nearly no detectable hypoxia in the well differentiated H-tumors. These results are in line with a previous report on the same tumor model using TOLD-MRI [10]. In addition, the [¹⁸F]FMISO-TAC-curves in PET were extremely variable between the three tumor sublines indicating also large differences in perfusion [5]. Thus, the investigated tumor-sublines represent a wide range of differently oxygenated tumors allowing dedicated investigation of the role of oxygenation on the radiation response.

To further exploit the role of ¹²C-ions to overcome hypoxia, which is presumably the most important resistance factor in photon therapy, a four-armed dose-response study was performed. Larger moderately differentiated HI-tumors were selected as model tumors because of its proven hypoxia and its extensive heterogeneous treatment response to photons. For larger HI-tumors treated with photons either under ambient or complete hypoxic (clamping) conditions, the detected oxygen enhancement ratio (OER) was clearly below 2, which is in line with previously published in vivo studies [26, 27]. Generally, OERs for single dose irradiations in solid tumors under clamping conditions were found to be lower than in cell culture studies [3, 28], presumably because tumor cells in intact tissues are not only impacted by the intrinsic cellular radioresistance but also by additional factors like cell to cell communication, bystander effects, and the immune response. Moreover, clamping does not only create a transient severe hypoxic state but also reduces the nutrient supply and induces a strong extracellular pressure to the capillaries which might increase secondary tumor cell death and therefore masks the potentially higher OER to some degree. When ¹²C-ions (dose average LET: 65 keV/μm) were applied under identical experimental conditions an up to 15% lower OER was found for larger HI-tumors. The detected decrease of OER for ¹²C-ions is relevant and if confirmed in patients would increase effectiveness dramatically. With this respect, the only available study, which compares the impact of tumor oxygenation for ¹²C-ions and photons in patients, is inconclusive [29].

A detailed comparison with the previous dose-response experiments for the small tumors, however, turned out to be difficult as the TCD₅₀-values after photon and ¹²C-ion irradiations under non-clamping conditions were found to be substantially higher as compared to the previously investigated small tumors. This suggests that larger tumors are not only associated with an increased number of tumor cells but also that volume-dependent alterations of the tumor micromilieu might play a role. To investigate this hypothesis in more detail, the clamping experiments are currently repeated for the small tumors within a new project and a comparison of the response of small and large tumors will be published separately when the results are available.

Radioresistance of tumors due to hypoxia is clinically of highest relevance as oxic tumors have a much higher disease-free survival than hypoxic tumors, as has been shown in head and neck cancer patients [30]. Therefore, the observed reduction of the OER for ¹²C-ions is a very important finding for the treatment of hypoxic tumors. In a first patient cohort treated with carbon ions, Japanese colleagues verified a smaller OER of ¹²C-ions in uterine cancer [29] and our findings in the experimental prostate carcinomas confirm this, however, further analyses are necessary. Although the obtained promising OERs for ¹²C-ions might in principal be used to overcome radioresistance evoked by severe hypoxia, the dependence of OER on LET is still an open question. In vitro the OER for high LET irradiations decreases with increasing LET and is expected to be negligible at LETs higher than 200 keV/μm [31]. Our preliminary results based on dose-response studies with oxygen ions (dose average LET: 101 keV/μm) using the same tumor model also indicates a small OER near to 1.

Finally, structural and functional assessments were performed in HI-tumors to decipher some of the biological mechanisms, responsible for the differential effects of ¹²C-ions and photons. In-house synthetized [¹⁸F]FAZA in combination with static PET measurements in a dedicated small animal scanner (Inveon Micro-PET/SPECT/CT (Siemens Medical Solutions, Knoxville, USA)) has been established in larger HI-tumors. A significant tracer-uptake prior to treatment followed by a clear reduction 1 week after irradiation was found for photons as well as for ¹²C-ions in a first pilot study. The hypoxia imaging tracer [¹⁸F]FAZA is a well-established alternative to the first generation tracer [¹⁸F]FMISO and is known to correlate well with both, autoradiography and the hypoxia marker pimonidazole [32]. Yet, results so far are reported to be ambivalent. No significant general reoxygenation was seen for SiHa cervix tumors in mice after fractionated irradiations with 10 or 25 Gy photons [32], while reoxygenation has been shown after 2 weeks of fractionated radiotherapy for patients with head and neck cancer in some but not in all cases [33]. Radiation-induced reoxygenation seems to depend on tumor type and on the intrinsic characteristics of the individual tumor. Tumor cell inactivation, tissue shrinkage, vessel damage and altered perfusion are biological factors associated with oxygenation changes after irradiation. A very striking observation was the extremely fast emerging of vascular disruptions already 18 h after single doses of carbon ions, while similar changes were delayed after photon irradiations. Nevertheless, a clear correlation between vessel integrity and hypoxia or reoxygenation is currently missing.

In spite of existing vascular structures, oxygen delivery to neoplastic and stromal cells is frequently reduced or even abolished by increased vessel distances, severe structural abnormalities of tumor microvessels, disturbed microcirculation and increased interstitial pressure. This can lead to false negative results in PET measurements as the tracer may not reach the hypoxic regions. To independently assess functional microenvironmental disturbations, measurements with doppler ultrasound as well as DCE-MRI were performed. Ultrasound measurements revealed a high blood flow followed by a rapid and dramatic decrease after carbon ions and later a steady state in the first 4 weeks after irradiation. Besides, this initial effect no other significant variation was detected between the two radiation modalities as well as for different dose levels. A further refinement of these results is expected from photoacoustic imaging [34], which allows assessing the oxygen saturation of tumor vessels based on optical detection of oxy- and deoxyhemoglobin. Similar to the ultrasound measurements, preliminary results of DCE-MRI revealed a faster and higher contrast enhancement after single doses of ¹²C-ions as compared to photons, which is in line with the more prominent vessel damage observed on the histological level. This first assessment is based on a qualitative rather than quantitative data analysis and a more detailed investigation using pharmacokinetic modeling to extract perfusion-related tissue parameters is ongoing.

Although the picture of the differential response to photon and ¹²C-ion irradiation is not yet fully completed, the described radiobiological research program has systematically documented the higher efficacy of ¹²C-ion therapy in a syngeneic experimental tumor model. It has been shown that tumor-associated resistance factors can be partly overridden by high-LET ion beam therapy. Moreover, the results provide evidence that the increased effectiveness of high-LET radiation is related to tumor-associated factors, preferentially those which are responsible to render tumors more resistant to conventional photon treatments. Dose-response studies revealed that hypoxia is one of the dominant radioresistance factors and that its impact can be reduced by high-LET ion beams on a clinically relevant level. This could potentially improve the probability of local tumor control relative to conventional photon treatments. Pronounced vascular disruptions associated with enhanced cell inactivation together with a rapid disturbance of the tumor microenvironment were identified as the most striking mechanistic effects of carbon ions and these results served as starting point for various ongoing research projects.