Magnetic Resonance Imaging of Tumor-Associated Macrophages: Clinical Translation

This article is featured in Highlights of This Issue, p. 4057

Tumor-associated macrophages (TAMs) have been associated with tumor progression and poor prognosis in patients with lymphomas (1, 2) and sarcomas (3, 4). New therapeutic drugs that target TAM are currently being developed and are starting to enter the clinic (5–8). It is important to identify tumors that are heavily infiltrated with TAM in order to stratify patients to TAM-modulating therapies and to monitor treatment responses. Various diagnostic tools for TAM detection and quantification have been developed, including gene expression analyses, immunohistochemistry, and fluorescent magnetic nanoparticle labeling (3, 4, 9). However, a major limitation of these methods is their invasive nature and lack of clinical availability. No diagnostic test exists to date that can detect and quantify TAMs noninvasively and repeatedly in patients. This creates a major bottleneck for the development and clinical translation of novel cancer immunotherapies. We addressed this unmet clinical need by developing an immediately clinically applicable molecular imaging test for the selective detection and quantification of TAM in patients. It has been shown in animal models that iron oxide nanoparticles are engulfed by macrophages in tumors and can be detected with MRI by a measurable signal decline on T2-weighted images (10, 11). We therefore hypothesized that clinically applicable ferumoxytol nanoparticles would serve as an MRI biomarker for TAM in patients. The same nanoparticles are not retained in normal tissues outside of the reticuloendothelial system. Ferumoxytol (Feraheme) is currently the only nanoparticle compound that is FDA approved (for treatment of anemia) and readily clinically available as an MRI agent via an “off-label” use. Neither ferumoxytol nor other nanoparticles have been used for TAM imaging in patients thus far. The purpose of our study was to determine if ferumoxytol-enhanced MRI can detect TAM in lymphomas and sarcomas of pediatric patients and young adults. If successful, this imaging test could be immediately utilized as a noninvasive imaging biomarker for TAM, for identifying tumors amenable to TAM-modulating immunotherapies and monitoring these immunotherapies in clinical practice (11).

This prospective, nonrandomized, Health Insurance Portability and Accountability Act (HIPAA)–compliant clinical trial was approved by our Institutional Review Board and was performed under an investigator-initiated IND for “off-label” use of the iron supplement ferumoxytol (Feraheme) as a contrast agent with drug approval from the FDA (IND 111154) and ClinicalTrials.gov Identifier: NCT01336803 and NCT01542879). The primary research objective was to test if ferumoxytol-enhanced MRI can detect TAM in lymphomas and sarcomas of pediatric patients and young adults. Our prespecified hypothesis was that clinically applicable, intravenously injected ferumoxytol nanoparticles can serve as an MRI biomarker for TAM in patients with cancer.

Pediatric patients and young adults with an age of 8 to 35 years, a biopsy-confirmed lymphoma or sarcoma, and willingness to participate in a research imaging study were included in this study. Exclusion criteria were contraindications to MRI; history of allergies against contrast agents or iron compounds; hemosiderosis or hemochromatosis; or pregnancy. The sample size was determined based on preliminary data (12) and power calculations.

In a first step, five patients (two lymphoma and three bone sarcoma) underwent pre- and postcontrast MRI with ferumoxytol nanoparticles to confirm ferumoxytol enhancement of lymphomas and sarcomas. Subsequently, 20 patients (10 with lymphoma and 10 with bone sarcoma) underwent ferumoxytol-enhanced MRI 24 to 48 hours after i.v. injection. T2* relaxation maps were created for each tumor, and T2* relaxation times were measured by operator-defined regions of interest. Lymphomas were diagnosed by a hematopathologist (D. Gratzinger), and sarcomas were diagnosed by a pathologist with specific expertise in the diagnosis of pediatric bone sarcomas (F.K. Hazard) at Stanford University. TAMs were identified using CD68⁺ and CD163⁺ immunohistochemical staining kits.

In a pilot study, we enrolled five patients (n = 2 lymphoma and n = 3 bone sarcoma) who received both pre- and postcontrast MRI scans after ferumoxytol administration. From November 2014 until December 2017, we enrolled 20 additional patients for the main arm of our study, consisting of 10 patients with malignant lymphomas (six males and four females; 20.2 ± 7.2 years old; all classical Hodgkin lymphomas) and 10 patients with bone sarcomas (five males and five females; 17.6 ± 5.4 years old; four Ewing sarcomas and six osteosarcomas). Informed consent was obtained from adult study participants. Parental consent was obtained for children younger than 18 years of age, and pediatric patients gave their assent to participate in the study.

Twenty-four to 48 hours before the planned MRI scan, all patients received a slow i.v. infusion of the FDA-approved iron supplement ferumoxytol (dose of 5 mg Fe/kg, diluted 1:4 in saline, administered over at least 15 minutes). This dose was overall well tolerated by pediatric patients in previous studies (12, 13). For safety evaluations, vital signs including pulse and blood pressure of the patients were measured before, during, and up to 30 minutes after ferumoxytol administration. Twenty-four hours later, patients underwent MRI on a 3T MR scanner (GE Discovery MR750; GE Healthcare), using a short tau inversion recovery sequence (STIR): repetition time (TR) = 4,925 ms, echo time (TE) = 54 ms, inversion time (TI) = 200 ms, echo train length (ETL) = 8, flip angle (FA) = 90°, field of view (FOV) = 360 mm, slice thickness (SL) = 4 mm; a multi-echo, flow-compensated 2D fast spoiled gradient recalled (FSPGR) sequence: TR = 150 ms, TE = 2.2 ms, inter-echo interval 2.2 ms, ETL = 8, FA = 30°, FOV = 460 mm, SL = 5 mm; and a Fast Spin Echo (FSE) sequence: TR = 4,900 ms, TE = 63 ms, ETL = 18, FA = 90°, FOV = 220 mm, SL = 5 mm. We generated T2* maps of all tumors using customized postprocessing software (Cinetool, GE Global Research; ref. 14) and measured T2* relaxation times using operator-defined regions of interest.

Patients underwent tumor biopsy (n = 14) or gross total resection (n = 6) at 9 ± 3 days after the MRI. To correlate MRI findings with the degree of macrophage infiltration in malignant tumors, we performed immunohistochemical staining of macrophages in the tumor tissue on paraffin-embedded tissue specimen. Immunohistochemical analysis was performed on 4-μm-thick sections that were placed on glass slides, baked for 1 hour at 60°C, deparaffinized in xylene, and hydrated in a graded series of alcohol. Endogenous peroxidase was blocked with hydrogen peroxide. Staining was performed on the Benchmark XT automated stainer (Ventana) and with standard heat-induced epitope retrieval except as indicated, and the chromogen used was diaminobenzidine (Enzo). Antibodies used were as follows: CD68 (mouse monoclonal KP1 1:1600, Dako, Leica, pH 6.0 retrieval) and CD163 (mouse monoclonal 10D6 1:50, Novocastra/Leica, Ventana, pH 8.5 retrieval). Hematoxylin (VWR) was used for counterstaining. To detect intracytoplasmatic iron in macrophages, paraffin-embedded tumor tissue sections were stained with Prussian blue (Dako).

Light microscopic images were obtained on an Olympus BX45 upright microscope (Olympus) with UPlanFL 4× /0.13, 20×/0.50, and 100×/1.25 oil immersion lenses, SpotFLEX camera, and acquisition software (Diagnostic Instruments Inc.).

Lymphoma cases were reviewed by an expert hematopathologist (D. Gratzinger) who confirmed the diagnosis of classical Hodgkin lymphoma in each case.

Bone sarcomas were reviewed by an expert in pediatric histopathology (F.K. Hazard). Four patients underwent a biopsy, and six patients underwent a tumor resection. For the whole tumor specimens, the specimens were sectioned longitudinally in the sagittal plane according to routine grossing techniques. An entire cross-section containing the largest tumor volume was oriented according to the MRI scan and mapped out using a grid numbering scheme. Three random tumor areas per patient were selected for histopathologic evaluation, carefully avoiding areas of tumor necrosis.

Immunohistochemical analysis was performed under ×20 magnification. The percent area covered by CD68- or CD163-positive macrophages was quantitated using freely available ImageJ software using color thresholding followed by particle analysis (15). Appropriate color thresholding levels were reached in consensus between two physicians (M. Aghighi and A.J. Theruvath; Supplementary Fig. S1).

A threefold approach was taken to evaluate the ability of MRI to image TAMs: (1) In a pilot study of five patients who underwent both pre- and postcontrast MR imaging, we confirmed tumor ferumoxytol enhancement by comparing pre- and postcontrast tumor T2* data with a paired t test. (2) To determine if ferumoxytol-MRI can differentiate tumors with different TAM content, we compared T2* enhancement data of lymphomas and bone sarcomas with a two-tailed Student t test. (3) To determine if ferumoxytol-tumor T2* enhancement correlates with TAM quantities within a given tumor group, we performed a linear regression analysis. All statistical computations were performed with Microsoft Excel software. An alpha level of 0.05 was chosen to indicate significant differences.

A pilot study in five patients (n = 2 lymphoma and n = 3 bone sarcoma) was conducted to confirm that ferumoxytol nanoparticles cause detectable signal enhancement of pediatric tumors on MR images. T2-weighted MR images before ferumoxytol administration revealed isointense to mildly hypointense signal in all tumors compared with muscle as an internal standard (Fig. 1A). Postcontrast T2-weighted MR images at 24 hours after i.v. ferumoxytol injection demonstrated negative (dark) enhancement of all tumors (Fig. 1B). Corresponding T2* relaxation times of the tumor tissue were significantly shorter on postcontrast scans (8.27 ± 2.12 ms) compared with precontrast scans (13.80 ± 2.80 ms; P = 0.036; Fig. 1C). Skeletal muscle did not show any significant change on postcontrast scans.

Lymphomas showed mildly decreased MRI signal on T2- and T2*-weighted scans with homogenous ferumoxytol enhancement (Fig. 2A and B). Bone sarcomas showed markedly decreased MRI signal on ferumoxytol-enhanced T2- and T2*-weighted MRI scans and markedly inhomogenous enhancement (Fig. 2C and D). The average T2* relaxation time of ten bone sarcomas (7.71 ± 0.60 ms) was significantly lower compared with the average T2* relaxation time of ten lymphomas (14.84 ± 1.03 ms; P = 0.0000006, Fig. 2E).

Histopathologic evaluation of lymphomas revealed the typical mixed background of classical Hodgkin lymphoma with scattered large Hodgkin variants. Immunohistochemistry revealed admixed CD68- and CD163-positive TAMs (Fig. 2F). Histopathologic evaluation of bone sarcomas showed very pleomorphic tumor cells with numerous and atypical mitotic figures, areas of osteoid matrix and/or calcifications, and areas of necrosis. Solid tumor areas contained high densities of CD68- and CD163-positive TAMs (Fig. 2G). The mean density of CD68-positive TAM in lymphomas and bone sarcomas was not significantly different (P = 0.373; Fig. 2E). However, the density of CD163-positive TAM was significantly higher in bone sarcomas compared with lymphomas (P = 0.043; Fig. 2E).

Because the composition of the tumor microenvironment is very different for lymphomas and bone sarcomas, we correlated MRI signal enhancement and TAM density separately for these two tumor types:

Lymphomas demonstrated homogenous ferumoxytol tumor enhancement on MRI (Fig. 3A–D). This corresponded to homogenous TAM distributions in lymphomas on histopathology (Fig. 3E and F). Tumor T2* relaxation times showed a significant, inverse correlation with CD68⁺ stains (r = –0.68; P = 0.031) and CD163⁺ stains (r = –0.76; P = 0.010; Fig. 3G).

Bone sarcomas demonstrated a markedly heterogenous ferumoxytol tumor enhancement on MRI (Fig. 1): Some tumor areas showed notably reduced T2* signal, and other areas showed relatively high T2* signal (Fig. 4A). On histopathology, this corresponded to tumor areas that contained clusters of TAM and other tumor areas that did not contain any TAM (Fig. 4B–D). Tumor T2* values correlated significantly with the density of CD68⁺ stains (r = –0.53; P = 0.012) and the density of CD163⁺ stains (r = –0.56; P = 0.007; Fig. 4E).

In addition, we noted that the pattern of MRI enhancement corresponded with TAM and iron nanoparticle distribution on histopathology (Supplementary Fig. S2). A telangiectatic osteosarcoma showed enhancing septa after ferumoxytol injection on MRI, which corresponded to similar septal distribution of CD68⁺ TAMs (Fig. 5).

In summary, tumor ferumoxytol enhancement on MRI correlated significantly with TAM density in lymphomas and bone sarcomas, and showed similar distribution of TAMs on histopathology.

Data showed that ferumoxytol-enhanced MR imaging can detect and quantify TAM in lymphomas and bone sarcomas of pediatric patients and young adults. Ferumoxytol-tumor enhancement correlated with the degree of TAM infiltration in different tumors and different areas of a given tumor. Results integrate the basic science discovery of TAM as markers of tumor progression (3, 16) with the use of novel MRI biomarkers to characterize tumor immune responses (10, 11), and can be immediately applied in a clinical setting.

The inflammatory component of the cancer microenvironment has not yet been a major target for clinical imaging thus far. Macrophage activation has been evaluated in other contexts; for example, inflammatory macrophages in atherosclerotic plaques and arthritic joints have been imaged with nanoparticle-enhanced MRI (17–19).

Macrophages in tumors have been targeted with preclinical imaging probes: Melancon and colleagues imaged TAMs in C6 tumors in rats with the experimental probe PG-Gd-NIR813 and combined fluorescence and MR imaging (20). Pérez-Medina and colleagues (21) imaged TAM in a mouse model of breast cancer with (89)Zr-labeled reconstituted high-density lipoprotein nanoparticles and PET. Locke and colleagues (22) targeted TAM in a mouse model of pulmonary adenocarcinoma with (64)Cu-labeled mannosylated liposomes (MAN-LIPs) and PET. Leimgruber and colleagues (23) targeted TAM in three different cancer models (colon carcinoma, lung adenocarcinoma, and soft-tissue sarcoma) in mice with the experimental probe AMTA680, using fluorescence imaging, MRI, and intravital microscopy. Movahedi and colleagues (24) tracked TAM with ^99mTc-labeled anti-MMR (macrophage mannose receptor) nanobodies and single-photon emission computed tomography/micro-CT, and Jiang and colleagues (25) tracked TAM in hepatomas in mice with Cy7-labeled deoxymannose and near-infrared fluorescence imaging. Although all of these approaches confirm the feasibility to selectively target and image TAM in malignant tumors in mice, none of these previously reported tracers can be applied in patients. Our ferumoxytol-MRI approach has the distinct advantage that it is readily clinically applicable.

Preclinical studies have shown that nanoparticle phagocytosis is a slow process. Intravenously injected nanoparticles are first distributed in the intravascular space, followed by slow extravasation in tumors due to the enhanced permeability and retention effect and subsequent phagocytosis by TAMs (26). Preclinical studies have also shown that ultrasmall iron oxide nanoparticles with a size of less than 50 nm localize to macrophages in tumors at 24 hour after i.v. administration (10, 11). Building on this experience, we also performed our imaging studies at 24 hours after ferumoxytol administration, although this gap between nanoparticle infusion and imaging is not convenient for most patients. Future studies have to show if earlier or later imaging times after nanoparticle infusion could be also used for TAM imaging.

Malignant tumors can contain different TAM phenotypes: M1 TAM phenotypes counteract tumor progression, whereas M2 TAM promotes tumor growth. Previous experiments showed that both M1- and M2-TAMs take up ferumoxytol (11). However, because the vast majority of TAMs in malignant tumors represent the M2 subtype (27), we did not attempt to discriminate these subpopulations. Other investigators developed mannose-targeted nanoparticles, which provide more specific targeting of M1-TAMs only (24, 25). However, as described above, these probes are not clinically translatable. We found a better correlation between ferumoxytol enhancement and CD163 stains compared with CD68 stains, which is in accordance with previous reports that CD163 is upregulated in M2-TAM (28–30). Recently, a proliferating TAM subpopulation has been identified in high-grade cancers, which was associated with an increased risk of recurrence (31). We would expect these metabolically active, proliferating TAMs to show marked nanoparticle uptake, whereas nonactivated macrophages or monocytes may show less (or no) ferumoxytol uptake. Most of our tumor biopsies/resections were not obtained directly after the MRI scan. Therefore, we could not generate a systematic correlation between T2* signal and iron within specific cell types in the tumor tissue.

Iron oxide nanoparticles might also be useful to study the effect of TAM on tumorigenesis. Shih and colleagues tracked the accumulation of ex vivo iron-labeled macrophages in oncogene LMP1-transformed BALB/c-3T3 tumors in mice (32). The authors found increasing TAM quantities and nanoparticle enhancement with progressive tumor growth. It was also shown that there are focal low signal intensity areas in lymphoma lesions due to iron deposits which are related to a biological inflammatory syndrome (33). Although these patients did not receive iron oxide nanoparticles intravenously, iron deposits were found mostly in patients with aggressive lymphoma, which could be explained by accumulation of iron-laden TAMs. In addition, the authors reported that iron deposits in lymph nodes corresponded to elevated FDG uptake, possibly due to activated macrophages in addition to tumor cells. In lymph nodes, nonneoplastic inflammations could lead to false-positive iron oxide nanoparticle and FDG uptake.

In bone sarcomas, osteoid matrix and calcifications can both appear hypointense on T1- and T2-weighted MR images. This can be differentiated from iron oxide nanoparticle enhancement by comparing pre- and postcontrast MRI scans. However, in a clinical setting, performing two scans on 2 subsequent days is not very practical. We omitted the precontrast scan for some of our patients and focused our analyses on the soft-tissue component of the tumor on 24-hour postcontrast scans, carefully excluding the bone and areas with typical patterns of periosteal new bone formation. However, there is a good chance that some of our evaluated tumor areas contained some calcifications. This is a limitation of our current analyses which might have contributed to the imperfect correlation between tumor T2* signal and TAM density in sarcomas. Our future investigations are dedicated to solving this problem by identifying specific signal characteristics of bone and iron oxides, e.g., on multi-echo gradient sequences, ultrashort TE sequences, and quantitative susceptibility mapping (QSM). QSM has shown excellent results in differentiating calcifications (diamagnetic) from hemorrhages (paramagnetic; ref. 34).

A potential alternative approach is to track macrophages with ¹⁹F MRI. This approach is based on the ability to detect fluorinated compounds with ¹⁹F MRI. Due to the negligible amount of fluorine in the body, ¹⁹F MRI can be used to label and track cells with little background signal (35). Khurana and colleagues described detection of TAMs with ¹⁹F MRI after i.v. injection of a perfluorocarbon nanoemulsion (36), and Shin and colleagues described detection of macrophages in an animal model of colitis after i.v. injection of perfluorocarbon (37). The advantage of this technique is that it is inherently quantitative and allows for precise cell quantification (38, 39). However, although intravenously injectable perfluorocarbons are in principle clinically translatable, they are not immediately available in a clinical setting. In addition, ¹⁹F MRI requires specific surface coils or specific tuning of ¹H surface coils for ¹⁹F detection (40), which would add time and costs to clinical imaging procedures. ¹⁹F MRI does not provide anatomical information but needs to be fused with ¹H MR images, similar to the procedure known from integrated PET/MR scans. Generating ¹H/¹⁹F MRI overlays is a time-consuming process. Finally, there is no current procedural terminology code for ¹⁹F MRI, i.e., these studies cannot be covered by the patient's insurance to date. Our technique has the disadvantage of being less precise with regards to cell quantification, but the advantage to be easily integrated into routine clinical MRI scan procedures.

Our finding of higher TAM quantity in sarcoma over lymphoma is novel and has not been described in the scientific literature. However, it can be linked to different outcomes of lymphomas and bone sarcomas. Our patients with lymphoma all had classical Hodgkin lymphomas, which have an excellent prognosis in children and young adults (41). Conversely, the majority of bone sarcomas investigated in our study were osteosarcomas, which have a worse prognosis, especially when metastasized (42). In principle, our finding of different TAM quantities confirms the known link between the quantity of TAM and aggressiveness of malignant tumors.

Our results have immediate clinical implications. Because the quantity of TAM in malignant tumors has been associated with tumor progression and metastases in breast cancer (43), Hodgkin lymphoma (2), and osteosarcoma (16), our ferumoxytol-enhanced MR imaging test might be suitable as a new prognostic marker for aggressive cancers and thus unfavorable outcomes. Patients with TAM-rich tumors, as demonstrated by ferumoxytol-enhanced MRI, could be directed to novel TAM-modulating immunotherapies, whereas patients with minor or absent TAM responses could be directed to alternative treatment options (44). Our TAM imaging test could be utilized to monitor tumor response to standard chemotherapy and new immunotherapies, determine patient-specific doses of TAM-modulating therapies, and evaluate potential interactions with other drugs or radiotherapy. Because clinical trials of new therapeutic drugs and combination therapies are expensive and take years to complete, the immediate value and health care impact of our new imaging test could be immense.

In conclusion, our data show that ferumoxytol-enhanced MRI can be used as a surrogate biomarker for TAM in pediatric lymphomas and bone sarcomas. This noninvasive imaging test may help to study the effect of immune responses on tumor progression and outcomes in pediatric patients and serve as a new biomarker for monitoring of new TAM-targeted therapies.

No potential conflicts of interest were disclosed.

Conception and design: A.J. Theruvath, L.L. Pisani, S.J. Holdsworth, H.E. Daldrup-Link

Development of methodology: M. Aghighi, A.J. Theruvath, L.L. Pisani, R. Alford, A.M. Muehe, S.J. Holdsworth, F.K. Hazard, H.E. Daldrup-Link

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Aghighi, A.J. Theruvath, A. Pareek, L.L. Pisani, R. Alford, A.M. Muehe, T.K. Sethi, S.J. Holdsworth, F.K. Hazard, D. Gratzinger, S. Luna-Fineman, R. Advani, S.L. Spunt, H.E. Daldrup-Link

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Aghighi, A.J. Theruvath, A. Pareek, L.L. Pisani, R. Alford, T.K. Sethi, D. Gratzinger, S. Luna-Fineman, H.E. Daldrup-Link

Writing, review, and/or revision of the manuscript: M. Aghighi, A.J. Theruvath, A. Pareek, A.M. Muehe, T.K. Sethi, F.K. Hazard, D. Gratzinger, S. Luna-Fineman, R. Advani, S.L. Spunt, H.E. Daldrup-Link

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Aghighi, A.J. Theruvath, L.L. Pisani, A.M. Muehe

Study supervision: A.J. Theruvath, A.M. Muehe, R. Advani, H.E. Daldrup-Link

(Other) I confirmed the sarcoma diagnoses, stained slides with immunohistochemistry, interpreted staining and took pictures for figures: F.K. Hazard

We thank members of the Daldrup-Link lab for their helpful reviews and discussions of our study design and research results. We thank Eileen Misquez for her excellent administrative assistance. We thank Vandana Sundaram from the Quantitative Sciences Unit at Stanford University for her excellent statistical consulting.

This work was supported in part by a grant from the NCI, grant number R21CA190196 (H.E. Daldrup-Link), by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, grant number R01 HD081123-01A1 (M. Aghighi, A.J. Theruvath, A.M. Muehe, S. Luna-Fineman, and H.E. Daldrup-Link), and by a Post Doctoral Fellowship Award in Translational Medicine by the Pharmaceutical Research and Manufacturers of America (PhRMA) Foundation (T.K. Sethi).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.