Evaluation of hypoxia in a feline model of head and neck cancer using ⁶⁴Cu-ATSM positron emission tomography/computed tomography

Hypoxia occurs in tumors for a variety of reasons; these include abnormal vessel growth [1, 2], fluid accumulation in the tumor extracellular matrix and rapid proliferation of cancer cells causing high interstitial pressure [2, 3], a breakdown of the diffusion geometry within the tumor, and paraneoplastic or therapy-related anemia leading to decreased oxygen delivery [4]. While tumor hypoxia was initially recognized as a cause for cellular radiation resistance, it is now known to contribute more generally to malignant progression and therapeutic failures [5‐7]. Lack of oxygen within tumors results in relative resistance to ionizing radiation, since the presence of oxygen permits irreversible peroxidation of DNA following ionizing radiation [5]. Furthermore, in acidic, hypoxic conditions, an aggressive cellular phenotype, with increased propensity for angiogenesis, invasion, and metastasis can emerge, an effect mediated by hypoxia-inducible transcription factors [2, 8‐11].

Hypoxia and its contribution to malignant phenotype and treatment failure are well-documented in head and neck squamous cell carcinoma (HNSCC) [6, 9, 11‐17]. Conversely, modulation of hypoxia may provide benefit to patients with HNSCC [18], which underscores the importance of understanding the impact of therapies on tumor hypoxia and developing improved methods to modulate tumor pO₂ and the molecular response to hypoxia. Unfortunately, animal models used to study HNSCC may not completely recapitulate the larger, invasive, and metastatic phenotype observed in human clinical populations. Indeed for many cancers and agents, there is a significant gap between preclinical rodent investigations and the clinical response of patients, suggesting a need to understand the biology of therapeutic interventions in models that more closely mimic human malignancies.

One potential model for HNSCC is head and neck squamous cell carcinoma that occurs spontaneously in pet cats. HNSCC is among the most common cancers affecting cats [19, 20]. Although its causation is not well studied, it is thought that the fastidious grooming behavior exhibited by cats may put the feline oropharynx at risk of exposure to a variety of environmental carcinogens [21‐23]. In addition to sharing histopathologic appearance, feline HNSCC is characterized by invasive, highly recurrent, and sometimes metastatic phenotype that is also observed in people with this cancer [19]. Furthermore, feline and human HNSCC may share their molecular underpinnings including frequent expression of EGFR [24, 25] and Cox-2 [26‐28], as well as mutant p53 [23]. However, to our knowledge, the presence of hypoxia has not been previously studied in feline HNSCC.

A great variety of techniques to detect hypoxia in tumors have been developed. Traditionally, techniques for evaluating tumor hypoxia have comprised tissue probes and immunohistochemical evaluation of tissue [29]. However, these methods have limited clinical application given that they are invasive and provide only focal assessment of oxygenation. To provide a clinically applicable, global assessment of tumor hypoxia, imaging techniques have been applied. In vivo imaging methods include both magnetic resonance (MR) techniques such as dynamic contrast enhanced-MR and nuclear-based imaging modalities, including SPECT (Single Photon Emission Computed Tomography) and PET (Positron Emission Tomography).

PET utilizes the detection of secondary, annihilation photons produced by cyclotron-generated, positron-emitting radionuclides, such as ¹⁸F, ¹³N, ¹⁵O, ¹¹C, ⁶²Cu, and ⁶⁴Cu. Suitable radionuclides are chemically coupled with tracers targeted for detection of particular molecular or physiologic parameters, such as hypoxia. Though activity of the most commonly used PET agent, 2-deoxy-2-(¹⁸F)fluoro-D-glucose (FDG), has been correlated with gene expression induced by hypoxia (HIF-1 α), FDG does not directly detect hypoxia within the tissues [17]. A number of PET tracers specifically designed for the detection of hypoxia have been developed. These include either misonidazole (MISO) or azomycinarabinofuranoside (AZA) coupled to ¹⁸F, or ATSM coupled to a positron-emitting isotope of Cu (⁶²Cu of ⁶⁴Cu) [13‐16, 30, 31]. All such agents rely on the hypoxia-dependent trapping of the tracer in cells that are hypoxic, yet viable. Cu-diacetyl-bis(N₄-methylthiosemicarbazone) (Cu-ATSM) has been demonstrated to exhibit hypoxia associated cellular uptake and is particularly advantageous due to its rapid uptake and strong signal to noise ratio. However, there is also evidence that some tumor subtypes may not demonstrate a direct relationship between Cu-ATSM signal and hypoxia [16, 32].

Our primary goal was to determine whether feline head and neck tumors, particularly feline HNSCC, exhibit biologically relevant hypoxia. For our purposes we considered levels of hypoxia sufficient to confer cellular radioresistance or to induce of HIF1α signaling to be biologically relevant. Such consequences occur below 1% O₂ (7.5 mmHg). In addition, we planned to evaluate the utility of ⁶⁴Cu-ATSM PET to detect hypoxic tumors in cats. To accomplish these aims, all cats were imaged with ⁶⁴Cu-ATSM PET/CT and were also evaluated using at least one other technique to measure intratumoral hypoxia including a fluorescent probe and/or immunohistochemical detection of pimonidazole. Herein, we demonstrate that most feline head and neck tumors concentrate ⁶⁴Cu-ATSM and that this signal is concomitant with low intratumoral oxygen levels and pimonidazole uptake. Feline HNSCC provides an opportunity to explore the modulation of tumor oxygen and vascular physiology in a clinically relevant system.

The twelve cats included in this study ranged in age from 7–16 years (mean = 12 ± 2.8 years), comprised 8 females and 4 males, and were all of common domestic (rather than purebred) origin. Of the twelve primary masses examined in the cats, six were squamous cell carcinomas (HNSCC), three were sarcomas, and three were benign lesions, (Table 1). Size of the masses ranged from 1.4 cm (benign) to 8.7 cm (malignant) maximum diameter with a mean of 4.0 ± 2.0 cm.

With the exception of the bone cyst, all lesions demonstrated at least regional 64-Cu uptake (Table 1). Tmax/M ratios were significantly higher (P < 0.005) than Tav/M ratios, reflecting heterogeneity of uptake in tumors, which in three tumors (both osteosarcomas and one HNSCC) included signal voids. For the tumors exhibiting signal voids, pre and post contrast CT images were compared. Based on Hounsfield Unit (HU) analysis, the tumoral regions exhibiting no 64-Cu uptake were also devoid of CT contrast enhancement, which demonstrates lack of perfusion and, likely, necrosis. Pre and post contrast measurements in the HNSCC were 41 and 40 HU respectively, while in the osteosarcoma, in an area without mineral attenuation, values were 40 and 42 HU pre and post contrast; this compares to an area with contrast enhancement and 64-Cu uptake in the same tumor, of 37 and 122 HU pre and post contrast. In the second osteosarcoma, histopathology of the entire tumor was performed (Figure 1) and this demonstrated that the signal void occurred within a necrotic cavity communicating with a cutaneous ulcer.

⁶⁴Cu - ATSM uptake was highest for HNSCC (Median T_max/M = 11; Median T_av/M = 3.8) than for sarcomas (Median T_max/M = 7.3; Median T_av/M = 2.2) and the benign masses (Median T_max/M = 6.0; Median T_av/M = 1.9). However, given the small numbers and variability in the data, there were no statistically significant differences in comparing uptake parameters between HNSCC (P = 0.24 for T_max/M; P = 0.09 for T_av/M) and other tumor types or between malignant and benign tumors (P = 0.15 for T_max/M; P = 0.21 to T_av/M).

Quantitative detection of tumor O₂ using the intratumoral fluorescent probe confirmed, using a different technique, that tumors with ⁶⁴Cu-ATSM uptake also exhibit regions of very low oxygenation, ranging from 0.6 to 2.6 mmHg, which would be expected to have biologic consequences including radioresistance and HIF1α induction (Table 1). Conversely, the tissues in the region of the bone cyst that did not take up ⁶⁴Cu-ATSM, appeared to be normoxic (Table 1).

In addition to the fluorescent O₂ detection probe, pimonidazole immunohistochemistry was also used to investigate tumor hypoxia. When pimonidazole was administered 24 hours prior to biopsy, there was minimal detectable immunostaining in samples, regardless of dose. Whereas in three tumors, in which pimonidazole was administered within an hour of biopsy, there was intense immunohistochemical staining. The discrepancy in staining between samples collected 24 hours or 1 hour before biopsy suggests that pimonidazole tissue adducts are relatively short-lived in cats [33]. The patient with osteosarcoma was severely compromised by the primary tumor and systemic metastasis and died following imaging. Thus the entire tumor was available for examination and spatial comparison of pimonidazole and ⁶⁴Cu-ATSM uptake (Figure 1). This comparison suggests a similar distribution of pimonidazole and ⁶⁴Cu-ATSM in this tumor.

Several additional tissues, distinct from the primary tumor, demonstrated ⁶⁴Cu-ATSM uptake, including lymph nodes (medial and lateral retropharyngeal lymph nodes, mandibular lymph nodes, and superficial cervical lymph nodes) draining the primary tumor in six of the cats with malignancies. In one of these six cats, there was additional assessment of a mandibular lymph node evaluated by fine needle aspiration cytology, which demonstrated reactive change rather than metastatic neoplasia.

Two of the cats had fluid within the tympanic bulla that demonstrated ⁶⁴Cu-ATSM uptake. One cat demonstrated signal associated with a necrotic maxillary molar. Three of the cats had ⁶⁴Cu-ATSM uptake within the thyroid glands. In one cat with bilateral thyroid uptake, clinical hyperthyroidism was confirmed by serum thyroid panel. In another case, a large thyroid gland with increased ⁶⁴Cu-ATSM uptake on PET/CT was confirmed as a thyroid adenoma at necropsy. In the third cat, there was PET signal in an enlarged thyroid gland, but disease was not confirmed with serum panel or histopathology. The cat with osteosarcoma that died immediately following PET/CT had a diffuse increase in pulmonary signal and at necropsy there were multiple 2–4 mm metastatic nodules in its lungs.

In two cats, intratumoral hypoxia was evaluated before and after treatment with an antiangiogenic agent, either a galectin-1 targeted peptide (Anginex) or a multiple tyrosine kinase inhibitor that targets VEFGR2 (toceranib, Palladia™, Pfizer Animal Health, Kalamazoo, MI). PET/CT and intratumoral oxygen probe measurements were performed one week apart with treatment administered in the intervening interval. Similar location of the probe was attempted as outlined in the materials and methods. After one week, there was minimal change in tumor size as measured by CT, with both tumors classifiable as “stable” when applying the RECIST (Response Evaluation Criteria in Solid Tumors) system used for human tumors [38]. Nor was there appreciable change in CT appearance. However, both tumors exhibited a slight increase in T_max/M. While T_av/M increased slightly in the Anginex-treated cat, there was a slight decrease in T_av/M in the toceranib-treated cat, with select regions of this second tumor exhibiting less radiopharmaceutical uptake (see Figure 2; Table 2). Intratumoral probe measurements demonstrated variability in certain regions of both tumors (Table 2). In the toceranib-treated tumor, pO₂ values were consistently increased at each location. In the Anginex- treated tumor the three regional measurements demonstrated decreased, increased, and stable pO₂ levels, respectively.

The biologic effects and clinical consequences of intratumoral hypoxia have been the focus of decades of research. It is well-established that hypoxic cells in vitro and in animals are relatively radiation resistant [39]. Furthermore, it has been demonstrated that patients with hypoxic tumors, including HNSCC, are more likely to experience treatment failures both locally and systemically [12, 18, 39]. Therefore, a variety of methods to increase tumor oxygenation or to target hypoxic cells within tumors have been investigated. Traditionally, these efforts have included measures such as hyperbaric oxygen administration, inhalation of carbogen gas, and the use of nitroimidazoles as hypoxic cell radiation sensitizers [18]. More recently, agents that specifically target hypoxic cell populations have been developed [40]. Finally, it has also been observed that anti-angiogenic and anti-vascular therapies may also modulate tumor oxygenation [1, 41]. However, despite these various efforts, clinical gains have been modest. While a multitude of factors may contribute to the gap between experimental and clinical results, two issues are particularly problematic. First, of particular importance in the targeting of tumor hypoxia, the assessment of relevant molecular and biologic surrogate endpoints is challenging in humans [42]. Second, rodent models for human cancer have significant limitations that do not always permit direct clinical translation [43]. In this study, we demonstrate the application of developing technology to assess tumor oxygenation in a clinically relevant model, spontaneous feline HNSCC.

There are a variety of methods for evaluating tumor oxygenation and these have been thoroughly reviewed elsewhere [29, 42]. All of these techniques have strengths and limitations, with no single technique offering complete characterization of this dynamic, complex phenomenon [42]. Imaging technology, by providing a noninvasive, three-dimensional, real-time assessment of hypoxia, is particularly promising as a clinical tool. In this study, we investigated hypoxia using ⁶⁴Cu-ATSM. Cu(II)-conjugated ATSM enters cells by either passive diffusion or endocytosis where is reduced and trapped, likely with the dissociation of reduced Cu(I) from ATSM, within hypoxic, yet viable cells [44, 45]. Normoxic cells are able to oxidize the reduced copper, which then is transported out of cells, either passively or, more likely, using a variety of chaperones or transporters [42, 45]. In preclinical studies, data demonstrate that tumor cells vary in their uptake of Cu-ATSM even at constant pO₂, implicating factors such as variable transporter expression, microenvironmental pH, cellular metabolism or the existence of alternative retention mechanisms [32, 45]. Advantages of Cu-ATSM include, rapid uptake, strong signal to noise ratio, the availability of a variety of Cu isotopes with variable half-lives and emission spectra, and some potential for therapeutic as well as diagnostic utility [46‐48]. Cu-ATSM agents have subsequently been used to image multiple tumors [16, 32, 44, 46, 49‐53] and hypoxic tissues [54, 55].

In this study, we demonstrate that most (11 of 12) feline head and neck tumors take up ⁶⁴Cu-ATSM with Tav/M and Tmax/M greater than 1.5 and 4.3, respectively. In studies that have investigated Cu-ATSM in human cancer patients, Tav/M ratios ranging from 2.6 – 3.5 have been used as arbitrary cutoff points for defining hypoxic and normoxic tumors [56]. Indeed these levels of radionuclide uptake have been associated with clinically relevant endpoints such as response to treatment and survival. However, these studies have not documented intratumoral hypoxia using independent methods making it difficult to determine whether these T/M ratios are best for determining actual hypoxic state. Furthermore, tumors with significant radiopharmaceutical uptake also demonstrate regions with quantitatively low pO₂ (less than 7.5 mmHg) or an affinity for pimonidazole, a hypoxia specific marker that forms adducts when the pO₂ is less than 10 mmHg. Conversely, the bone cyst that failed to take up ⁶⁴Cu-ATSM, with T/M ratios was normoxic based on peritumoral pO₂ measurements. These results support the hypothesis that ⁶⁴Cu-ATSM uptake occurs in hypoxic rather than normoxic feline tumors. However, complete spatial correlation between distribution of ⁶⁴Cu-ATSM was only possible in one case in which the animal died following imaging and the entire tumor, an osteosarcoma, was available for sectioning and evaluation. Additionally, no proof of ⁶⁴Cu-ATSM uptake or lack thereof in these tumors’ normoxic cells was available. Subjectively, there appeared to be concordance between pimonidazole and ⁶⁴Cu-ATSM uptake. Interestingly, in a xenograft study, ⁶⁴Cu-ATSM uptake failed to correlate with nitroimidazole staining in a sarcoma, while demonstrating a strong correlation in both a carcinoma and a glioma [32].

While we were able to measure hypoxia using at least one other technique in 11 of the 12 tumors, technical problems precluded the use of all three techniques in every case. The intratumoral probe was not operational at the time of evaluation of the first three cats. We also limited our quantitation of tumor pO₂ to a small number of regions within the tumor. Studies of human tumors suggest that dozens of measurements may be needed to fully map tumor oxygenation. However, our goal was simply to verify the presence or absence of hypoxia in a few intratumoral or peritumoral areas rather than to provide a complete mapping of each tumor.

While the use of pimonidazole has been studied in the dog [33, 34], we were unable to find reports of the use of this marker in cats. Therefore, doses were selected based on those reported in dogs. Many drugs, including the nitroimidazole, metronidazole, have similar or identical doses in both cat and dog. We were unable to perform additional procedures such as biopsy in the imaging facility, which necessitated a separate anesthetic episode. Our initial plan had been to administer the pimonidazole concomitant with the ⁶⁴Cu-ATSM to permit side-by-side comparison between the two. However, at the doses used, we were not able to detect pimonidazole in cat biopsy samples collected 24 hours after administration. In contrast, pimonidazole staining was strong and easily visualized when pimonidazole was administered shortly before biopsy. These data suggest that the pimonidazole adducts may turn over more quickly in feline tumors than in dogs [33]. Factors that may have influenced pimonidazole staining intensity in the cat include species specific pharmacokinetic variables such as serum half life, which in humans is about 5 hours and only 15 minutes in the mouse. Therefore recommended doses are several times higher in the mouse than in humans. Unfortunately, these data are not available for the cat. It is possible that with far larger doses of pimonidazole we would have been able to visualize adducts in our biopsy specimens obtained 24 hours after administration. Other factors that could have contributed to poor retention of pimonidazole in tissues include rapidly changing tissue perfusion or rapid turnover of cells in the tumor. HNSCC in cats is considered a rapidly progressive malignancy therefore it is possible that tumor growth kinetics may have also contributed. Pimonidazole dose optimization should be performed in feline tumors to better utilize this technique.

It is not surprising to see heterogeneous distribution of hypoxia within tumors, therefore significant differences between the T_max/M and T_av/M in these PET studies is expected. However, signal voids were also observed in areas with poor perfusion (based on CT contrast studies), which would presumably be hypoxic. In one cat with osteosarcoma, the signal void corresponded to a necrotic cavity identified at necropsy. It is possible that poorly perfused regions contain necrotic rather than viable cells. Since uptake and retention of Cu-ATSM requires intact cell and likely lysosomal membranes, it is unlikely that Cu-ATSM would accumulate in these necrotic regions [45]. A compounding factor in the specific case of the osteosarcoma may be the high interstitial pressures in bony areas of osteosarcomas leading to decreased perfusion [57‐59].

In this study, while strongest ⁶⁴Cu-ATSM uptake was observed in HNSCC, sarcomas and benign tumors also exhibited uptake and significant hypoxia. Thus, hypoxia is not a characteristic of tumor type or malignancy. The increased uptake among feline HNSCC coupled with intratumoral probe and pimonidazole data support that these tumors are significantly hypoxic like their human counterpart. However, we cannot rule out that some other characteristic of HNSCC, in addition to hypoxia, has influenced Cu-ATSM uptake and retention such as the expression of specific transporters or metabolism. It has been hypothesized that altered redox state associated with glycolytic metabolism in some tumors might also promote reduction and trapping of Cu-ATSM. It is likely that the use of multiple methods to investigate tumor hypoxia may yield the most accurate assessment.

Regardless of whether ⁶⁴Cu-ATSM uptake is a direct reflection of tumor hypoxia, studies of human HNSCC indicate the clinical significance of this tracer. SUV_max[53] and T_av/M ratio [56, 60] cutoffs have been successfully used to predict recurrence after radiation and prognosis, respectively, in human cancer patients. It was not our objective to correlate these data with prognosis in cats nor was it feasible given inconsistent follow-up therapy in these cases. However, in using spontaneous HNSCC to investigate the biologic impact of therapeutic intervention, these data may guide selection of appropriate thresholds.

Unexpectedly, certain other tissues in these cats exhibited ⁶⁴Cu-ATSM uptake. Uptake in lymph nodes draining the primary tumor was seen in 8/12 cats. These lymph nodes exhibited normal contrast enhancement on CT and only mild to moderate enlargement. In one case, the lymph nodes exhibited reactivity rather than metastasis. While hypoxia is recognized in metastatic or primary tumors occurring in lymph nodes, its presence in reactive lymph nodes has not been previously documented, to the authors’ knowledge [61, 62]. It is interesting to consider how hypoxia in draining lymph nodes might influence the development of regional metastasis. Two cats also had ⁶⁴Cu-ATSM uptake in association with presumptive otitis media. Hypoxia has been demonstrated in rats with otitis media [63].

Hyperthyroidism is common in elderly felines and occurs secondary to adenomatous hyperplasia, thyroid adenomas, or least commonly functional thyroid carcinomas [64]. Two of the three patients with ⁶⁴Cu-ATSM uptake in the thyroid had clinically proven functional hyperthyroidism prior to the scan. There are limited data concerning hypoxia in non-malignant disorders of the thyroid, though low level vascular endothelial growth factor (VEGF) expression, which is hypoxia inducible, has been observed in follicular adenomas and adenomatous goiter of the thyroid in humans [65]. This may be caused by the hypermetabolic state and increased oxygen consumption [66] of the thyroid cells in these conditions. Human thyroid carcinoma metastases, though not present in these patients, were also demonstrated hypoxic when imaged with ^99mTc-HL91, a nitroimidazole, and SPECT [67]. Confirmation of hypoxia in other tissues using another technique could not be easily performed in these cases due to inaccessibility of lesions and invasive nature of the other techniques used.

Two cats were evaluated before and after different antivascular therapies. It has been proposed that modulation of tumor vasculature may affect intratumoral hypoxia and preclinical studies have supported this notion [1, 68]. In this study, treatment was accompanied by only slight changes in Cu-ATSM uptake. Since we do not have data from untreated cats to demonstrate pattern on Cu-ATSM uptake over time, it is not possible to determine whether the changes observed were drug specific. However, in both cats, there was a slight increase in T_max/M possibly indicating regional vascular compromise. However, these changes may be within range of error, as the inverse quartic relationship between partial pressure of oxygen and Cu-ATSM uptake results in steep slope within the initial decline of pO₂, while at low pO₂, slight changes may be insufficient to alter uptake of ⁶⁴Cu-ATSM [13]. At the same time, in the cat treated with a tyrosine kinase inhibitor targeting VEGFR2, a slight decrease in T_av/M occurred concomitantly with increased quantitative pO₂ as measured with the intratumoral probe. Furthermore, focal areas in the periphery of the tumors had decreased signal, suggesting that further investigation into dose and time frame of anti-angiogenic therapy administration as a hypoxia modulator might be useful.

Despite their experimental utility, rodent models fail to completely recapitulate human cancer and to provide the degree of heterogeneity that is characteristic of human clinical populations. The gap between xenograft and genetically-engineered mouse models and human clinical studies are well recognized. Furthermore, as function of animal size, the tumors seen are considerably smaller from that expected in a human clinical population. Feline HNSCC may provide a relevant alternative to rodent models for this disease.

All of the feline HNSCC studied exhibited regional evidence of biologically relevant hypoxia, regardless of measurement technique. Therefore, in addition to morphologic, clinical and molecular similarities, feline and human HNSCC also share physiologic characteristics, further demonstrating how closely the disease in cats mimics its human counterpart. We also preliminarily illustrate, using anti-vascular agents, that feline tumors can be used to study the biologic consequences of interventions and to develop and apply surrogate endpoints. It is reasonable to assume that such studies could be used to address specific issues of clinical translation and inform the development of more effective human trials.