MR susceptibility imaging for detection of tumor-associated macrophages in glioblastoma

Tumor-associated macrophages (TAMs) are a key component of glioblastoma (GBM) microenvironment. Considering the differential role of different TAM phenotypes in iron metabolism with the M1 phenotype storing intracellular iron, and M2 phenotype releasing iron in the tumor microenvironment, we investigated MRI to quantify iron as an imaging biomarker for TAMs in GBM patients. 21 adult patients with GBM underwent a 3D single echo gradient echo MRI sequence and quantitative susceptibility maps were generated. In 3 subjects, ex vivo imaging of surgical specimens was performed on a 9.4 Tesla MRI using 3D multi-echo GRE scans, and R2* (1/T2*) maps were generated. Each specimen was stained with hematoxylin and eosin, as well as CD68, CD86, CD206, and l-Ferritin. Significant positive correlation was observed between mean susceptibility for the tumor enhancing zone and the l-ferritin positivity percent (r = 0.56, p = 0.018) and the combination of tumor’s enhancing zone and necrotic core and the l-Ferritin positivity percent (r = 0.72; p = 0.001). The mean susceptibility significantly correlated with positivity percent for CD68 (ρ = 0.52, p = 0.034) and CD86 (r = 0.7 p = 0.001), but not for CD206 (ρ = 0.09; p = 0.7). There was a positive correlation between mean R2* values and CD68 positive cell counts (r = 0.6, p = 0.016). Similarly, mean R2* values significantly correlated with CD86 (r = 0.54, p = 0.03) but not with CD206 (r = 0.15, p = 0.5). This study demonstrated the potential of MR quantitative susceptibility mapping as a non-invasive method for in vivo TAM quantification and phenotyping. Validation of these findings with large multicenter studies is needed.


Introduction
Glioblastoma (GBM) is the most common malignant primary brain tumor, with a median survival of 15 months, and a 2-year survival rate of 26.5% despite an aggressive treatment regimen including surgical resection and chemoradiotherapy [1]. Several mechanisms have been proposed for treatment failure which include tumor heterogeneity and immunosuppressive tumor microenvironment [2].
Pathologically, GBM is characterized by a heterogeneous microenvironment consisting of not only neoplastic but also non-neoplastic cells including resident and migrated immune cells, vascular cells, as well as other glial cells [3]. Among the non-neoplastic components of GBM microenvironment tumor-associated macrophages (TAMs) are the dominant immune cell population with a key role in tumor growth [4].
TAMs exert a pro-tumor behavior by suppressing the innate immune response, promoting tumor angiogenesis, as well as releasing intracellular iron into the tumor microenvironment [5,6]. These pro-tumor functions have been attributed to the M2 phenotype of TAMs. In contrast, the M1 phenotype of macrophages is known to sequester intracellular iron and induce an inflammatory response against tumor growth [6][7][8].
These opposing roles of macrophage phenotypes in tumor growth and response to treatment [9,10], have been the basis for novel immunotherapeutic approaches aimed at promoting the anti-tumor M1 behavior [11,12]. However, in the absence of non-invasive biomarkers to monitor response to therapy, the assessment of the dominant TAM behavior in the tumor microenvironment requires invasive tissue sampling, which imposes a significant limitation to the advancement and real-time monitoring of novel immunotherapeutic approaches.
Currently, the radiological diagnosis of GBM and assessment of response to therapy are routinely performed through a battery of structural and physiological MRI modalities. Tumor progression is often associated with an increase in the size of contrast-enhanced areas or in the relative cerebral blood volume; however, these features still have significant limitations in the ability to differentiate response to treatment from tumor progression, particularly following administration of immunotherapeutic agents [13]. Given the increasing knowledge about the role of immune microenvironment and the increasing number of immunotherapeutic trials for GBM, there is a great need for biomarkers that can non-invasively monitor the immune profile of the tumor microenvironment [14,15].
MRI techniques that are sensitive to susceptibility such as gradient-echo-based relaxivity (R2*) measurement [16] and quantitative susceptibility mapping (QSM) [17] can provide valuable information on sources of magnetic field inhomogeneity, including iron concentration in the tissue [18][19][20][21]. Considering the differential role of TAMs in iron metabolism [6,7], the goal of this study was to assess the potential of MR susceptibility imaging as a non-invasive biomarker to predict TAM's dominant phenotype in the GBM microenvironment. In particular, we hypothesized that QSM and R2* measurements would correlate with TAM quantity and phenotype in GBM.

Subjects
We screened subjects with intra-axial brain mass suggestive of high-grade glioma, who were referred to the radiology department of the Hospital of University of Pennsylvania from 2016 to 2019. Twenty-one adult patients who underwent tumor resection and whose tissue specimens were consistent with the histopathological diagnosis of GBM, based on 2016 WHO classification were enrolled in the study [22]. This study was approved by the Institutional Review Board of the University of Pennsylvania, and all research was performed in accordance with relevant guidelines and regulations. Informed consent has been obtained from patients to include the additional research MRI sequence and histopathological studies as described in the following sections.

Image acquisition
Imaging was performed on a single 3 T scanner (Trio; Siemens) equipped with a 12-channel head coil. Single echo gradient echo MRI was obtained in the axial orientation using the following imaging parameters: in-plane spatial resolution = 0.86 × 0.86 mm 2 , slice thickness = 3 mm, matrix Ex vivo MR imaging was performed to allow comprehensive radiology-pathology correlation. For ex vivo imaging, we selected three subjects whose specimens had the least hemorrhage in order to minimize artifact in susceptibility maps on ultra-high field MRI. The formalin-fixed paraffinembedded tissue samples for these subjects were imaged on a Bruker 9.4 Tesla 8.9 cm vertical bore MR as explained in detail in the supplementary material (see supplementary Fig. 1). T2 TurboRARE, and 3D multi-echo GRE scans were acquired at 60 °C to ensure paraffin melt (See supplementary material and methods).

Histology
A board-certified neuropathologist (M.P.N. with 6 years of experience), who was blinded to MRI susceptibility measurements, assessed the specimens from the initial surgical resections. Nineteen subjects had enough tissue for histological studies. Samples were formalin-fixed and paraffinembedded, and the block that best represented the submitted tissue (often greater than 50% of the entire tissue) was chosen for additional staining.
Each specimen was stained with hematoxylin and eosin, as well as with CD68 (myeloid-specific surface marker), CD86 (M1 TAM marker), CD206 (M2 TAM marker), and ferritin light chain (l-Ferritin). The unstained slides underwent heat-induced epitope retrieval in citrate buffer pH 6.0 (Leica Microsystems) for 20 min. We then performed immunohistochemical staining on the Bond 111 Autostainer with the hematoxylin counterstain and DAB chromogen.
For all subjects, histology slides were digitized at the Children Hospital of Philadelphia pathology core facility at 20 X using an Aperio CS-O slide scanner (Leica Biosystems, Buffalo Grove, IL). These scanned images were preprocessed with QuPath software [23], using an automatic module to adjust the IHC stain vectors and remove unexpected colors due to artifacts. Areas of artifact (e.g. tissue folds, edge artifacts and scanning artifacts) were removed from the analysis through manual tissue segmentation using the QuPath annotations tool [23]. For each case, the positivity percentage of the IHC images (i.e. CD68, CD86, CD206, and L-Ferritin) was then quantified for the whole slide based on a threshold applied to the 3,3' Diaminobenzidine (DAB) signal in the cytoplasm. Refer to the supplementary material table-1 for the parameters used in positive cell detection. To address variation observed in staining intensity, for each IHC image, positive cell quantification was first performed using an initial threshold equal to mean + 2SD of DAB optic density of all cells. To minimize the error rate in positive cell detection, this threshold was then modified by a neuropathologist (M.P.N) accordingly.

Radiologic image analysis
All sequences were registered to post-contrast T1, using the FSL MRI toolkit. Regions of abnormal contrast enhancement and necrosis were segmented using a semi-automated segmentation tool (ITK-SNAP) [24] followed by manual editing by two board-certified neuroradiologists (S.A.N. with 6 years of experience and J.B.W. with 4 years of experience). Areas of susceptibility artifact on source DSC (dynamic susceptibility contrast) and areas with high signal intensity on pre-contrast T1 source were excluded from these segmentations as areas of hemorrhage.
QSM images were reconstructed from single echo gradient-echo MRI magnitude and phase data using the morphology-enabled dipole inversion (MEDI) algorithm [25], implemented in MATLAB (MathWorks, Natick, Massachusetts). A brain mask was obtained from the magnitude image using the FSL brain extraction tool [26]. Subsequently, phase unwrapping was performed using Laplacian unwrapping [27]; transmit phase was removed by fitting and subtracting a fourth-order 3D polynomial [28]; and background field removal was performed using the regularization-enabled sophisticated harmonic artifact reduction for phase data (RESHARP) algorithm [29], prior to susceptibility inversion with MEDI using the default regularization parameters.
A threshold of 0.055 was applied to QSM values to exclude areas of artifact and venous structures. The thresholded QSM maps were reviewed and confirmed by a boardcertified neuroradiologist (S.A.N.), who was blinded to the immunohistochemistry results. The above-mentioned segmentations were applied to the QSM images and the voxel count and the average susceptibility of the segmented regions were recorded for each subject.
The T2* maps for the two ex-vivo samples were generated by a pixel-wise monoexponential fitting of the multiecho GRE images using MATLAB [30].
The Cerebral Blood Volume (CBV) maps were manually registered to the T1 post contrast scan using the ITK-SNAP manual registration tools. The T1 post contrast tumor enhancing region segmentations were then applied to the CBV maps. For each subject, the mean CBV values in these regions were then normalized to a standard region (approximately 500 mm 3 ) in the contralateral frontal cortex white matter to calculate relative CBV (rCBV) values.

Pathology-radiology region of interest (ROI) analysis
For the cases with ex vivo MRI, we undertook an ROI approach for direct correlation analysis between the corresponding ROIs on IHC and MRI images. First, areas of highly dense tumor were determined by a tissue classification model (1 μm/pixel; random trees; QuPath 2.0.1) [23] that was trained by a neuropathologist (M.P.N) on H&E images (Fig. 2). Subsequently, using QuPath annotations tools*, 16 ROIs (0.25 mm 2 each) were randomly placed in areas of tumor, identified on H&E images as explained above. The number of ROIs per slide was proportional to the total area of tumor dense regions on each slide. For each ROI on the corresponding IHC images, the total number of positive cells was counted as described previously (see Histology section and supplementary Table 1). To replicate ROIs on the corresponding coordinates on the T2* maps, IHC images were co-registered with MRI images using HistoloZee software (http:// picsl. upenn. edu/ softw are/ histo lozee/; see Fig. 2) [45]. In summary, the corresponding cut-plane in the MRI was first identified using manual affine MRI transformation to find the best contour match between the MRI image and each histology image (45). Co-registered histology and MRI images were then overlayed in ITK-SNAP software to replicate corresponding ROIs on MRI images [24]. For each ROI, the mean T2* values [24] were then used to calculate the effective transverse relaxation rate (R2*= 1/T2*).

Statistical analysis
We performed Shapiro-Wilk test to assess normality of data. For the correlation analyses between MRI metrics and histology, we used Pearson's R for normally distributed data and Spearman's Rho for non-normally distributed data. Multiple regression analyses were used to assess the potential contribution of increased rCBV in the enhancing regions to the prediction of TAM distribution. Mean and standard deviation was reported for descriptive statistics. P values below 0.05 were considered statistically significant. Statistical analysis was conducted in the R statistical environment (version 3.6.1; http:// www.r-proje ct. org/).

Subjects
We studied 21 (mean age: 64.1 +/− 9.86 years; 13 M; 8 F) patients with treatment naïve glioblastoma who underwent tumor resection. Two patients were IDH-mutant and the remaining patients were IDH wild-type. Among the 19 subjects with in vivo MRI, one subject was excluded because of distortion of MRI data due to significant motion artifact. Also, among the three subjects that were selected for exvivo imaging, one was removed due to tissue distortion during MRI imaging. One subject underwent both in-vivo and ex vivo MRI imaging of the tumor specimen.
Multiple regression model using both mean susceptibility and rCBV as predictors demonstrated significant effect of susceptibility, and insignificant effect of rCBV in prediction of CD68 (p = 0.039 and 0.18, respectively) and CD86 (p = 0.01 and p = 0.5, respectively).
In addition, mean susceptibility from the combination of tumor's enhancing zone and necrotic core correlated with the corresponding l-Ferritin (r = 0.72; p = 0.001) and CD86 (r = 0.63; p = 0.005) positivity percent for the entire section area, and demonstrated a trend toward significant correlation with CD68 (r = 0.46; p = 0.06).

Ex vivo MRI
A total of 16 ROIs were randomly placed on the tumor dense areas of IHC slides from the two blocks that had ex vivo MRI. We observed a significant positive correlation between mean R2* values (68.1 ± 37.1) and CD68 positive cell counts for the corresponding ROIs (r = 0.6, p = 0.016; Fig. 2). Similarly, mean R2* values significantly correlated with CD86 (r = 0.54, p = 0.03) but not with CD206 positive cell counts (r = 0.15, p = 0.5; Fig. 2).Of note, we did not find a significant correlation between R2* measurements and l-ferritin positive cells (r = 0.09; p = 0.7).

Discussion
In this study, we examined the role of MRI susceptibility biomarkers, QSM and R2*, to determine the distribution of TAMs in GBM microenvironment based on their differential role in iron metabolism using both in-vivo data from clinical MRI and the proof of concept ex-vivo MR experiment. We observed a significant positive correlation between in vivo QSM-based susceptibility and ferritin light chain. Previous studies have shown that the distribution of ferritin and iron closely overlap in the brain [16]. While ferritin heavy chain predominates in neurons [16,31], in line with their high iron uptake and peroxidase activity, l-ferritin is predominant in macrophages and microglia favoring their role in long-term iron storage [16,32]. Hence, our findings indicated that QSM measurements could predict iron storage in tumor-associated macrophages. We also observed that macrophage counts (as identified by general CD68 marker) significantly correlated with in vivo QSM, with higher susceptibility associated with higher CD68 cell counts. Of note, these results were replicated with CD86 positive cells, representing an M1 phenotype; but not with CD206 positive cells, representing M2 phenotype. This finding is in line with the role of M1 phenotype in storing intracellular iron, and M2 phenotype in releasing iron in the tumor microenvironment [7]. The M1 phenotype, is known to exert anti-tumor functions not only by triggering the inflammatory response but also by sequestering iron, on which tumor cells depend for their growth and proliferation [6,8]. In contrast, M2 phenotype promotes tumor growth and proliferation by releasing iron into the tumor microenvironment, as well as through anti-inflammatory functions [6,8,12]. Our results also demonstrated that rCBV did not have a significant relationship with TAM markers.
In this study, we also performed an ex vivo MR imaging on paraffin-embedded blocks from three subjects with minimal hemorrhage to allow comprehensive radiologypathology correlation [33]. Using a similar T2* relaxometry technique, several studies in preclinical tumor models demonstrated successful mapping of TAMs in treatment naïve tumors and in the context of immunotherapy [19,20,33,34]. We also observed that R2* values correlated with both CD68 and CD86 positive cell counts in multiple ROIs randomly placed on areas of highly dense tumor and coregistered with MRI. This direct histology and MRI correlation provided a proof of concept for the application of MR susceptibility imaging in predicting the distribution of TAMs in human GBM microenvironment.
Given their role in tumor growth and response to treatment, TAMs have been the focus of recent research for the development of imaging biomarkers to assess prognosis as well as response to treatment in GBM [35,36]. Multiple studies involved macrophage modulation in tandem with systematic treatment for GBM; for example, in vivo models have shown that anti-CD47 treatment has the ability to differentiate macrophages phenotypes into the M1 phenotype [37], thereby inducing anti-tumor effects [38] and inhibiting tumor growth. Furthermore, processes such as CD40 activation [39], Toll-like receptors (TLR) stimulation [40,41], and TAM depletion [42,43], have all been investigated in targeting the recruitment of macrophages and therefore modulating tumor progression. On the other hand, the association between TAMs immune function and iron metabolism in tumor microenvironment has been investigated in various disease models. For instance, Shenoy et al., found that iron-loaded macrophages resisted conversion from pro-inflammatory to immunosuppressive polarity, thus indicating the integral role of iron quantification in monitoring GBM response to therapy and predicting its prognosis [44].
A variety of MRI and PET imaging techniques have been used to quantify TAMs in gliomas [45]; for example, in a pilot study of 10 adult subjects with GBM who underwent ferumoxytol-enhanced MRI, measurements of susceptibility obtained after ferumoxytol administration correlated with iron-containing macrophage concentration [35]; however, this approach requires slow intravenous administration of ferumoxytol followed by next day delayed imaging which may limit its widespread use. Translocator protein (TSPO) specific PET radiotracers have also been used to monitors TAMs; however, the two main limitations are lack of phenotype (M1 vs. M2) specificity [45], and over-expression in tumor cells [46]. To the best of our knowledge, our study is the first to provide both in vivo and ex vivo evidence for the Fig. 2 Ex vivo MRI and immunohistochemistry correlations. There is a statistically significant positive correlation between mean R2* values and CD68 as well as CD86 positive cell counts on corresponding ROIs (n = 16) but not with CD206 positive cell counts application of non-invasive MRI imaging for TAMs quantification and spatial mapping in human GBM.
Our study had several limitations. Most importantly, we had a small sample size for in vivo and especially the ex vivo experiments given that most of the subjects were alive at the time of this study and we limited the ex-vivo experiments to subjects with a large amount of tumor tissue to avoid exhaustion of the specimen. Furthermore, one of our three ex-vivo samples was excluded from the study due to tissue distortion during the MRI procedure. Hence, to perform a correlation analysis between MRI susceptibility measures and IHC positive cell counts on the remaining two samples, we used a random ROI placement approach. While the small sample size was a limiting factor, the direct spatial co-registration of the two modalities [47] provided proof of concept for the application of MRI susceptibility imaging in the prediction of TAMs' distribution. In addition, the heterogeneous nature of glioblastoma microenvironment [48] imposes an inherent challenge to investigate inter-modality correlations between iron metabolism and TAMs' distribution. In our in vivo experiments, we tried to mitigate this limitation through computationally intensive whole-slide analyses that provides a comprehensive measurement of entire tissue sections and thus average out confounders due to TME heterogeneity. Finally, the presence of intratumoral hemorrhage may affect quantitative susceptibility mapping and T2* relaxometry; however, we did use source images of DSC perfusion sequence and other sequences to exclude areas of intratumoral susceptibility that were felt to reflect hemorrhage and the same methodology can be used to identify and exclude areas of hemorrhage from QSM maps in the clinical setting. In addition, areas of hemorrhage were excluded from tumor segmentation in ex vivo experiments.
In conclusion, we demonstrated encouraging evidence that MR susceptibility imaging can quantify the iron content of GBM and provide a non-invasive method to assess TAMs' phenotypic distribution, though further studies with larger samples size are needed to corroborate these results. We envision future applications of this research will include monitoring changes in iron content and macrophage composition during conventional GBM treatment regimens, as well as TAM modulating clinical trials, such as those described previously, to monitor treatment success and predict outcomes.

Supplementary Information
The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s11060-022-03947-3. Data availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Conflict of interest
The authors declare no potential conflicts of interest.
Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments.
Informed consent Informed consent was obtained from all individual participants included in the study.