Deep-tissue localization of magnetic field hyperthermia using pulse sequencing

Abstract Objective Deep-tissue localization of thermal doses is a long-standing challenge in magnetic field hyperthermia (MFH), and remains a limitation of the clinical application of MFH to date. Here, we show that pulse sequencing of MFH leads to a more persistent inhibition of tumor growth and less systemic impact than continuous MFH, even when delivering the same thermal dose. Methods We used an in vivo orthotopic murine model of pancreatic PANC-1 cancer, which was designed with a view to the forthcoming ‘NoCanTher’ clinical study, and featured MFH alongside systemic chemotherapy (SyC: gemcitabine and nab-paclitaxel). In parallel, in silico thermal modelling was implemented. Results Tumor volumes 27 days after the start of MFH/SyC treatment were 53% (of the initial volume) in the pulse MFH group, compared to 136% in the continuous MFH group, and 337% in the non-treated controls. Systemically, pulse MFH led to ca. 50% less core-temperature increase in the mice for a given injected dose of magnetic heating agent, and inflicted lower levels of the stress marker, as seen in the blood-borne neutrophil-to-lymphocyte ratio (1.7, compared to 3.2 for continuous MFH + SyC, and 1.2 for controls). Conclusion Our data provided insights into the influence of pulse sequencing on the observed biological outcomes, and validated the nature of the improved thermal dose localization, alongside significant lowering of the overall energy expenditure entailed in the treatment.


Introduction
Since it was first suggested by Gilchrist and colleagues in 1957 [1], magnetic field hyperthermia (MFH) has been explored as a cancer thermotherapy, using iron oxide magnetic nanoparticles (MNPs) as heating agents to deliver localized thermal energy into tumors (e.g [2,3].). This energy is generated through magnetization reversal processes within the MNPs, driven by the presence of external time-varying magnetic fields (e.g [4,5].).
MFH has been very effective in producing in vivo temperatures higher than 41 C [6]. Such temperatures lead to cell death, by affecting the cellular metabolism in general, and also RNA/DNA synthesis and DNA repair mechanisms, membrane fluidity and stability, as well as the function of cell surface receptors and transmembrane transport proteins [7,8] in particular. The formation of reactive oxygen species during hyperthermia further favors induction of cell death [9,10]. In response to the thermal stress, an altered expression of the proteins Ki67, TOP2A, TPX2, ERK 1/2 and JNK has been observed [11,12]. Interestingly, the cytotoxic effects of hyperthermia can differ depending on the tumor cell phenotype [12].
For clinical purposes, such relatively strong heating effects should be confined only to the tumors. However, to achieve such localization, many different factors need to be considered, including the MNP concentration or 'load' in the tumor tissue, the retention of the MNPs over time, the extent of the heating volume in the tumor, MNP heating parameters, and the amplitude and frequency of the applied magnetic field [4,5]. The topological distribution of the MNPs throughout the tumor is known to be of particular significance, especially when the tumors are near vital organs or major blood vessels, both of which act as thermal sinks, drawing the MFH-induced heat away from the tumor, and reducing its effectiveness. This effect is particularly prominent in a highly perfused organ such as the pancreas, where branches of the truncus coeliacus and other vessels, are present.
In this work, we highlight and explore an aspect of magnetic hyperthermia that is not often considered, but which we find has a distinct effect on localization: viz. the time domain sequencing of the treatment. From experimental studies in tumor-bearing mice, thermal modeling, and reference to published data on thermal dosing, we show that by varying the sequencing, it is possible to distinctly reduce the potentially damaging flow of heat away from the tumor. As part of this, we report on the development of a new orthotopic murine model of pancreatic cancer, which presents a preclinical target that better replicates the target clinical condition than more conventional subcutaneous xenograft models. The work is part of recent efforts to address the challenge of deep-tissue localization by the European Commission supported NoCanTher consortium [13], which focuses on a particularly aggressive cancer, namely locally advanced pancreatic cancer (LAPC)for which new treatment modalities are very much needed. Both the orthotopic model and the pulse sequencing studies described in this paper contributed, in 2020, to a successful application for National Competent Authority permission for a 15-patient clinical feasibility study of pulsed MFH, as an adjunctive treatment alongside standard-of-care chemotherapy, across two sites (Barcelona and Madrid), starting in 2021.

Thermal modeling
Model calculations were based on the Pennes bioheat equation, including both metabolic and perfusion effects [14][15][16]. Tissue-specific parameters [17,18] and temperature dependent changes in the perfusion rate were applied [19]. For the murine orthotopic tumor model it was assumed that the entire pancreas was enveloped by the orthotopic tumor, and that the tissue immediately beyond the edge of the tumor was muscle. The model was implemented using the opensource software package LT Spice, which is well suited to the detailed tracking of temporal effects. Although LT Spice is designed for electrical circuits research, it can be applied to thermal models [20][21][22].

Thermal dosimetry
Thermal doses were quantified using the thermal iso-effect method [23,24], where temperature-time data is converted into an equivalent treatment time at 43 C (CEM43 ¼ 'cumulative equivalent minutes' at 43 C). The conversion factor, based on cell survival data [25], was t R (43-T) where t is the dwell time (in minutes) at temperature T, and R ¼ 0.25 for T < 43 C, or 0.5 for T ! 43 C. For the in silico models an assumed homogeneous and spherical MNP dispersion allowed computation of CEM43Tx metrics, these being the CEM43 doses on the spherical surfaces enclosing the xth volume percentile of the tumor. Particular attention was paid to the CEM43T90 metric [26][27][28] by modeling T(r 90 ,t), where r 90 ¼ ffiffiffiffiffiffiffiffiffi 0:93 p r tumor : For the in vitro experiments it was assumed that heating was homogeneous over the entire cell culture, and the CEM43 conversion was applied directly to the recorded T(t) data.
For the in vivo experiments, where inhomogeneous MNP distributions were anticipated, another approach was taken. T(r probe ,t) data were recorded from a thermal probe pressed into (but not penetrating) the skin of the animal, at a distance of 1.0 ± 0.2 mm (as measured by ultrasound) from the edge of the tumor, and maximal temperatures T max noted for each run. As the MNPs were expected to form a collection of separate deposits in the tissue (injection volumes, see Table 1), it was reasoned that the T max values recorded would be indicative of the thermal doses delivered to the animals.

Magnetic heating apparatus
A solenoidal Preclinical MACH system (Resonant Circuits Limited) operating at 1.048 ± 0.010 MHz was used to deliver a sinusoidally varying magnetic field H 0 sin(xt), where H 0 could be varied from 4.44 to 8.34 kA/m. Thermal data were recorded using fiber optic probe thermometers (TS5 & FOTEMPMK-19, Optocon AG).

Pulse and continuous magnetic hyperthermia treatments
Local ethics committee policies led to a constraint against simple field-on/field-off sequences (each on-off cycle was deemed to be a separate experiment), so high-field/low-field sequences were used instead. For pulse in vitro, H 0 was set to H 0-1 ¼ 5.43 kA/m for 10 min, then raised to In vivo experiments using an orthotopic tumor model Orthotopic pancreatic tumor models of PANC-1 cells were propagated by surgical means. For this, fluorescent PANC-1 (2 Â 10 6 ) cells were suspended in 2% Matrigel TM in PBS and injected into the pancreas. The localization of the orthotopic tumor within the pancreas was substantiated by ultrasound imaging (Vevo700, FUJIFILM Visualsonics Inc.) and by histopathology on excised organs stained with hematoxylin/eosin (Supplementary Figure 1). Tumor volumes were assessed regularly via ultrasound. All animal experimentation was carried out in accordance with the international guidelines on the ethical use of animals, and they were approved by the regional animal care committee (Th€ uringer Landesamt f€ ur Verbraucherschutz, Bad Langensalza, Germany). Eight-and 10-weeks old female nude mice (Rj:Athym-Foxn1nu/nu, Janvier, Germany) were used. Animals were maintained under artificial day-night cycles (14 h/10 h light-dark cycles; 25 C room temperature) and received food and water ad libitum.

Animal groups
Animals bearing orthotopic tumors were randomized into 7 independent experimental groups. P s and C s : mice treated weekly with systemic chemotherapy (SyC: per kg BW, 50 mg Gemcitabine, i.p., 30 mg nab-Paclitaxel, i.v.) and with MFH (P s : pulsed; C s : continuous) after intratumoral application of MNPs (RCL-01, 0.72 mg Fe per 100 mm 3 tumor volume). 1 S: mice treated with SyC alone. P and C: mice treated with MFH alone (P: pulsed; C: continuous). N: non-treated controls (no MNPs, SyC or MFH). M: mice with intratumoral MNPs alone. Timeline, number of mice per group, and treatment parameters are shown in Figure 1 and

Protein expression analysis of tumor tissue
At day 30, animals were sacrificed and tumor tissues were excised, fixed, and embedded in paraffin. Tissue slides were incubated in antibody solutions (dilutions: 1: 60 to 750 in Dako REAL TM Antibody Diluent) against HSP70 (from mouse, BioGenex), Ki67, CD31 (all from rabbit, Abcam), and pERK (Cell Signaling), and subsequently with a biotin-labeled secondary antibody (goat anti-mouse, 50 min). Finally, treatment with streptavidin-alkaline phosphatase and chromogen (all from DAKO) was done for visualization. Sections were counterstained with hematoxylin. The presence of the respective proteins was estimated semi-quantitatively using a five-category scoring system (0 ¼ less than 5%, 1 ¼ 5 to 25%, 2 ¼ 25 to 50%, 3 ¼ 50 to 75%, 4 ¼ more than 75% of stained cells positive). CD31 as measure of tumor vascularity was estimated through hot spot analysis by Chalkley count [29].
Tumor slices from 2 animals per group were analyzed. Reproducibility was validated by a second independent and blinded examination [14].

Blood analysis (hemograms)
During animal euthanasia the subclavian vein was punctured and blood was collected and immediately measured on an automated hematology analyzer (Sysmex XT-1800i, Japan) according to the user manual. As a measure of therapeutic stress in animals the neutrophil to lymphocyte ratio was calculated [30]. Sample collection was performed by one experienced person in order to minimize experimentation-based animal stress, which could potentially impact data quality.

Circulating tumor cells in the blood
Immediately after euthanasia, blood was collected, transferred into erythrocyte lysis buffer (155 mM NH 4 Cl, 10 mM KHCO 3 , 0.1 mM EDTA, pH 7.3), centrifuged at 250 g (10 min, 4 C), and washed to remove cell debris. The enriched  nucleated cells were then stained with a fluorescence-labeled Alexa-Fluor-488 anti-EpCAM monoclonal antibody (VU1D9, Cell Signaling Technology) and analyzed on a FACSCalibur flow cytometer (Becton Dickinson). To determine the 'exposure' (E) of the body to CTCs as a measure of therapeutically-induced tumor cell spread, total CTCs (in percentage of total blood nucleated cells) per gram body mass was determined. The 'readiness' (R) of tumor cells to be released out of the tumor was determined by the total CTCs normalized to tumor volume (ratio of tumor volume between day 30 and day 0 post MNP application).

In vitro cell culture
The human pancreatic adenocarcinoma cell line (PANC-1, ATCC) was cultured in DMEM medium supplemented with 10% (v/v) fetal bovine serum (Gibco V R ). Cells were kept at 37 C in a humidified atmosphere containing 5% (v/v) CO 2 .

Cell viability assays
Experiments with PANC-1 cells were designed to entail CEM43 heating doses of more than 60 min, which is the anticipated therapeutic dose near intratumoral MNP deposits [31]. For this, PANC-1 cells were supplemented (or not, in the case of controls) with MNPs (FluidMAG-C11-D, diluted to a concentration of 100 mg Fe per mL) and subjected (or not) to MFH. Following the MFH treatments, cell viability was determined by a) the Alamarblue V R assay (24 and 48 h post hyperthermia) [32], b) counting cells via impedance spectroscopy (OMNI Life science); or c) via evaluation of 2 D cell survival at 16 days post hyperthermia [33]. All experiments were carried out at least 3 times (3 biological replicates with quadruple to sextuple samplings for each replicate).

Statistics
Data requiring statistical analyses were evaluated using the SigmaPlot 14.0 program. All data passed the Shapiro Wilk test for normal distribution. Therefore, ex vivo data with one end-point analysis per animal was depicted as mean and standard error of the mean. Since data from in vitro experiments on PANC-1 cells were composed of several biological replicates with multiple data acquisitions each, the mean and standard error of the mean was used. Owing to their normal distribution, the student's t-test was used to compare groups, and differences between groups were considered as statistically significant at a p value equal or less than 0.05. With consideration of the analysis of tumor volumes with low animal numbers per group, data were evaluated using a linear mixed model which includes a quadratic term for time and an interaction term for time and treatment group. Hereto, the effects of therapy (least square means of tumor volumes) at day 28 were determined with a 0.95 level of confidence using the Kenward Rogers degrees of freedom method and Tukey's method for p value adjustments.

Impact of pulse or continuous MFH on orthotopic pancreatic tumors in mice
On treating PANC-1-inoculated orthotopic pancreatic tumors in vivo with MFH combined with systemic chemotherapy (SyC), both pulse heating (P s ) and continuous heating (C s ) were effective in reducing tumor growth (compared to the non-treated group, Figure 2). Thermal probe data showed comparable T max levels in the P s and C s groups (Table 1, Supplementary Figure 3), indicating that comparable thermal doses had been delivered to both groups. The chemotherapy-only, S group in Figure 2 exhibited some tumor inhibition effects, comparable to those of the C S group, but not as marked as for the P s group. Parallel experiments on the hyperthermia-only P and C groups, showed no relevant reduction of tumor growth at all (Supplementary Figure 4), even though the thermal doses in these groups were comparable to those in the P s and C s groups (Table 1).
Although the latter observation may be attributed to an adjunctive effect in the combination of SyC and MFH, the difference between the P s and C s groups evident in Figure  2(B) is not so readily explained. To better understand how different responses might arise despite equal thermal dosing, a series of in silico experiments were performed. Figure 3 shows modeling data for a homogeneous distribution of MNPs in a 125 lL spherical pancreatic tumor. Five treatment scenarios were considered: Aa single 60 min heating cycle; B À 40 min of heating spread over 5 cycles of 8 min, interspersed with 4 min rest (no heating); C À 40 min over 10 cycles of 4 min, with 2 min rest; D À 20 min over 5 cycles of 4 min, with 2 min rest; and E À 10 min over 5 cycles of 2 min, with 1 min rest. T(r,t) curves were calculated for r ¼ 0 to 9.9 mm in 0.1 mm steps. Particular attention was paid to the T(r ¼ 3.0 mm,t) data to determine the CEM43T90 metric, and the input power level was adjusted so that the total CEM43T90 delivered in each of the scenarios was 72 min.
Despite the equal CEM43T90 doses, significant differences were seen in the thermal characteristics of the scenarios. First, and most dramatically, the total energy dissipation levels (input power x power-on time) ranged from 1,065 J (scenario A: continuous) to 267 J (scenario E: pulsed) ( Figure 3). Second, different intra-tumoral temperature distributions were attained (Figure 4(A)), with higher temperatures reached in the pulsed scenarios, indicating a more localized thermal dose. Third, and more subtly, the 'thermal treatment trajectories' (TTTs) were different. This latter point is illustrated in Figure 4(B), where each scenario is represented as a temperature-time histogram, with the amount of time spent in a given 2 C range plotted against the midpoint of that range. (The data in Figure 4(B) are for r ¼ 0 mm, but TTTs can be constructed for any point in the tumor.) An advantage of the TTT visualization is that it allows comparisons with known preclinically-or clinically-validated treatments. For example, the solid line plotted in Figure 4(B) corresponds to the well-known Moritz and Henriques data on thermally-induced epidermal necrosis of porcine and human skin [34]. (Although this data is many years old, it is still critically recognized as valuable [25], and in the absence of specific data on pancreatic tissue damage, it is a useful benchmark of in vivo thermal damage.) The line may be regarded as a 'necrosis threshold', beyond which necrotic damage occurs. As such, it is clear from Figure 4(B) that all five of the scenarios traverse the necrosis threshold, but with quite different TTTs, and different temperature-time end points. As such it is reasonable to expect the biological response of the tissue to be different, depending on the scenario useda result that is consistent with the in vivo P s versus C s results presented above.
To elucidate the impact on cell viability of pulse versus continuous MFH per se under controlled experimental conditions, in vitro experiments with PANC-1 cells were performed. These revealed that both the pulse and continuous modalities were equally cytotoxic in MNP-inoculated PANC-1 cells (at comparable CEM43 thermal doses, see Supplementary Figure 2, and Figure 5). The expression of cellular-stress-associated proteins was also comparable between the modalities (Supplementary Figure 6).
To assess the impact of P s and C s on protein expression in treated orthotopic pancreatic tumors, semi-quantitative analyses of protein expression (4 weeks post-MFH) were performed. A slightly increased cell proliferation (Ki67 expression) was observed, and a differential impact of P s and C s on the HSP70 and pERK expression indicators of cell stress (P s increased HSP70 and pERK expression; C s decreased HSP70, but increased pERK expression in comparison to non-treated control tumors, Table 3). Interestingly, P s rather increased whereas C s decreased tumor cell proliferation (Ki67 expression, Table 3). Moreover, tumor vascularity (CD31) remained unaltered in all treatment groups (Table 3). In contrast, pulse or continuous MFH as monotherapy either decreased protein expression in tumors (HSP70 and Ki67) or it maintained it unaltered (pERK, and CD31, Table 3).

Systemic effects after treatment of orthotopic pancreatic tumors in mice with pulse or continuous MFH
When looking at the systemic effects of MFH, pulse sequencing resulted in less peripheral heating of tissue than continuous hyperthermia. This is evident in the maximum rectal temperatures recorded during MFH, plotted as a function of the injected volume V i of magnetic heating agent ( Figure  6(A)). Although there is scatter in the data points, it appears that the body temperatures of the continuous (C and C s ) group animals increased faster as a function of increasing V i than was the case for the pulse (P and P s ) group animals.  Table 1 for day 0 tumor volumes and Table 2 for group notations.
Hemogram data showed different impacts of P s and C s on certain blood components, notably the white cell count, which is unchanged in P s , but significantly dampened in C s (Figure 6(B), Supplementary Figure 7). Interestingly, magnetic field hyperthermia as a monotherapy seemed to rather increase the white blood cell count and decrease some red blood cell parameters ( Figure 6(B), Supplementary Figure 8), whereas the MNPs alone had no distinct effect at all (Supplementary Figure 9).     Table 2. Bars depict mean and standard error of the mean with n ¼ 3. Ã p .05; ÃÃ p .01; ÃÃÃ p .001.
The measure of therapeutic stress in animals by the neutrophil to lymphocyte ratio [30] revealed that P s was much less stressful than C s ( Table 4). As a monotherapy, both P and C were apparently stressful for animals, albeit to a lesser extent than C s (Table 4).
Regarding the release of PANC-1 circulating tumor cells (CTCs) after the combined magnetic hyperthermia therapy, no effect on the body exposure (E) levels was seen in P s , while it was reduced in C s , as compared to the non-treated group (Figure 7(A)). Moreover, some differences between the modalities were observed in the tumor-volume-adjusted readiness (R) levels, with the P s higher, but the C s levels lower than that of the non-treated group (Figure 7(A)), albeit none were statistically significant. Interestingly, when applied as monotherapy, both P and C treatments significantly reduced E, and also reduced R, by tendency, (Figure 7(B)).

Discussion
When treating orthotopic pancreatic tumors in mice in combination with a systemic chemotherapy protocol including gemcitabine and paclitaxel (SyC) [35], both the pulse (P s ) and continuous (C s ) magnetic hyperthermia modalities were found to be more effective in reducing tumor volumes than SyC alone on day 27 of the study. To the best of our knowledge, this is the first study reporting that MFH in Table 3. Impact of pulse and continuous MFH in combination with SyC (top) or alone (bottom) on protein expression in orthotopic PANC-1 tumors excised from mice on day 30 as indicated in Figure 1.   Table 2), given as relative cell numbers normalized to the untreated control group. WBC: white blood cells, RBC: red blood cells, HGB: hemoglobin, HCT: hematocrit MCV: red blood cell mean volume, MCH: mean red blood cell hemoglobin, MCHC: mean cell hemoglobin concentration. Bars depict mean and standard deviation of the mean. Ã p .05; ÃÃ p .01; ÃÃÃ p .001. Table 4. Neutrophil to lymphocyte ratio of animals as estimated at day 30. Group notation as in Table 2. combination with systemic chemotherapy is effective in inactivating pancreatic tumors. Until now such promising effects have only been reported for hyperthermic intraperitoneal gemcitabine chemotherapy [36] or for modulated electrohyperthermia [37]. In this work, our goal was to achieve deep-tissue localization, and to circumvent nonspecific heating of tumor-adjacent tissues and organs. Two factors influenced our approach. One was that we anticipated an inhomogeneous distribution of the MNPs in the tumors [24,38], and localized 'hot-spots' [39] of 43-45 C or even higher temperatures formed in the very immediate surroundings of the MNP deposits, as a consequence. From this perspective, normalization was applied to the administered MNP dose per unit volume of tumor, in anticipation of the initial tumor volume per se being a non-determining factor in the biological outcomes. A second factor was an anticipated strong thermoregulation effect from the surrounding internal organs [40], which is much stronger in orthotopic than in more conventional subcutaneous models [27], and which led us to explore the pulse MFH modality both in silico and in vivo.
Pulse MFH has occasionally been reported in the literature, notably by Ivkov et al. [41], who used it to reduce morbidity due to peripheral eddy current heating in animals. All animal tissues are electrically conductive due to their high water content. As such, the time-varying magnetic fields of MFH induce circulating currents, as per Faraday's law. This effect was pronounced in Ivkov's studies, where large-amplitude fields (32,76, and 104 kA/m) were used [41]. Most MFH laboratories use much lower amplitudes, of order 10-15 kA/ m [42], for which the eddy currents, especially in small animal models, are negligible. This may explain why pulse MFH has been seldom used to date, with the exception of a few reports on its use, to obtain high thermal doses and cell death in vitro [43]; to maintain constant temperatures in a murine treatment model [44]; and as part of a pulsatile drug release strategy [45,46]. No previous studies have focused on dose-localization and biological outcomes per se, as in this work.
From the evidence of our in vivo experiments on orthotopic tumors, it appears that the presence of SyC was favorable for tumor regression, particularly in tumor areas between the MNP deposits. Since SyC was administered weekly during the whole experimentation time, it was expected that there would be a continuous stress impulse on DNA integrity of dividing (tumor) cells [47]. Moreover, MFH treatments interfere with the cellular repair mechanisms [48]. This increases the uptake of gemcitabine and nab-paclitaxel by changing the cell membrane permeability and therefore diffusion into the target cells (e.g [49].). As such, the observed reduction of tumor volumes is consistent with the MFH acting in a synergistic manner to SyC [31,50].
At a cellular level, our in vitro results show that the pulse and continuous MFH modalities have comparable effects in inactivating pancreatic PANC-1 tumor cells, at least under the conditions tested, viz. at thermal doses slightly above the toxicity threshold (CEM43 ¼ 60-90 min), and when the maximum temperatures during pulsed MFH were not higher than during continuous MFH (46 C). In contrast, the temperature pulsations in the orthotopic tumors in vivo might well have led to maximum temperatures that were higher than the continuous MFH group, in particular in the vicinity of MNP deposits, leading to a higher tumor reducing impact.
Moreover, P s induced cellular stress in orthotopic pancreatic tumors (e.g., increased HSP70 and p-ERK expression compared to non-treated controls, 30 days post MNP application), but also some propensity toward cell proliferation. Since pulse or continuous MFH as monotherapy attenuated the expression of stress and proliferation markers (i.e., Ki67 and HSP70), we suggest that SyC is the main driver of this effect. Therefore, it may be preferable to use more than two hyperthermia sessions in future, in order to strengthen their impact, at least when using low overall tumor temperatures. It is also noteworthy that there is uncertainty in all hyperthermia applications, due to the lack of real-time 3 D temperature monitoring and accurate image resolution for 3 D MNP distribution. Both may have affected our observation on therapeutic efficacy to some extent.
Regarding biological outcomes, the in vivo data show that neither P s nor C s affected the risk of tumor cell spread into the blood. Specifically, neither P s nor C s significantly increased the body's exposure (E) to CTCs compared to nontreated control animals, and there was no significant difference in E between P s and SyC treatments. The readiness (R) metric, which depicts the 'theoretical readiness' of tumor cell spread, was found to be higher after P s treatment than in the C s , SyC and N groups. However, this was not statistically significant, as the reduced tumor size at day 27 in the P s group led to large measurement uncertainties. Nevertheless, we can speculate that if R were elevated, it might be related to changes of the cell membrane fluidity due to magnetic hyperthermia and SyC [8], which might subsequently alter the cell migration behavior.
It is also notable that the effects encountered in relation to CTCs were not associated with tumor vascularity nor reactive changes in the blood flow, since the level of CD31 as marker for vascularity [51,52] was unaltered after treatments. In contrast to our findings, whole body hyperthermia (41.8 C) [53] and intraperitoneal hyperthermia [54] have been reported to increase the number of CTCs in the blood as a consequence of increased intratumoral blood flow. This is rather unlikely to occur in the orthotopic pancreatic tumors studied here, since they are poorly vascularized.
The hemograms of treated mice indicated that P s had almost no impact on the blood composition, whereas C s did. Both the P and C monotherapies affected cells of the blood compartment. Such effects may be associated with high demand, consumption, and new production of certain blood cell types as a consequence of immunological responses that occurred as a result of focal tumor cell destruction (apoptosis, necrosis) and bleeding. Further investigations are needed to clarify the corresponding reasons for such differences in more detail.
Systemic stress as measured by the ratio between neutrophils and lymphocytes [30] was not significantly increased by P s treatment, whereas it was by C s treatment, as compared to non-treated controls. Interestingly, when applied as monotherapy, both P and C treatments seem to induce higher systemic stress. In general, the ratio of neutrophils to lymphocytes closely correlates to the magnitude of stressor, such as the level of circulating glucocorticoid in animals [30]. Stress hormones are also produced by immunologic reactions. Cortisol is transported in the blood by the corticosteroid-binding globulin, and the body temperature seems to play a role in modulating the hormone release to the tissues [55]. Experimental stress was negligible, since blood samples were normalized to controls and animal experimentation was performed by only one experienced person.
Lastly, we note that the MFH conditions employed for the in vivo experiments here should not, and do not, translate directly across to the conditions that will be employed in the forthcoming Nocanther clinical study [13]. Other considerations, including eddy current heating, and the acceptable volumes of injected magnetic heating agents, are very different between animals and humans, and must be taken into account [56]. Thus, the magnetic field amplitudes and frequencies to be employed in the NoCanTher study are to be 4-5 kA/m and 300 kHz, rather than the 5.4 and 8.3 kA/m and 1.05 MHz used here. This reduction in H o and f will be compensated for by larger injection volumes, which in the clinic will contain 35 mg of Fe, compared to the ca. 1.0-2.5 mg of Fe used here.

Conclusions
Using orthotopic pancreatic tumors in mice, we found that both pulse and continuous magnetic field hyperthermia (MFH) treatments impact tumor growth in combination with systemic chemotherapy (SyC; gemcitabine and nab-paclitaxel). Pulse heating plus SyC gave the most sustained results and was the only one for which tumor volume reduction was still present at day 27 of the study. In silico experiments showed that different pulse scenarios could deliver the same thermal dose (e.g., as measured by the CEM43T90 metric) as a given continuous treatment, but with better localization and less energy expenditure. In vitro, both modalities were effective in inactivating isolated pancreatic cancer cells. Systemically, pulse MFH induced less peripheral response than continuous MFH, as evident in lower core temperature rises in the animals, and lower levels of blood-borne stress markers (viz. the neutrophil to lymphocyte ratio). Given its anti-tumor effectiveness and its high tolerance in animals, we conclude that pulse MFH should be considered favorably as a thermal treatment modality for pancreatic cancers in combination with systemic chemotherapy.