AcidoCEST MRI Evaluates the Bone Microenvironment in Multiple Myeloma

Abstract

Purpose:

Multiple myeloma (MM) is an incurable disease of malignant plasma cells in the bone marrow (BM). Adaptive responses to hypoxia may be an essential element in MM progression and drug resistance. This metabolic adaptation involves a decrease in extracellular pH (pHe), and it depends on the upregulation of glucose transporters (GLUTs) that is common in hypoxia and in cancer cells. CEST MRI is an imaging technique that assesses pHe indirectly by the exchange rate of magnetic saturation transfer between labile protons on a solute and water. Thus, this study aimed to determine the feasibility of acidoCEST MRI for pHe measurement using an orthotopic mouse model of MM compared with GLUT1 immunofluorescence staining as a reference.

Procedures:

Orthotopic BM engrafted MM xenografts were established in NSG/NOD mice using the human RPMI8226 myeloma cell line. AcidoCEST MRI was performed approximately six weeks after intravenous challenge, before and after intravenous administration of iopamidol. BM pHe values were generated via fitting the CEST spectrum with the Bloch-McConnell equations. Samples were decalcified, sectioned, and immunostained for GLUT1 expression. Pearson’s correlation was used to assess the relationship between pHe and [H₃O⁺] versus GLUT1 expression.

Results:

Ten mice underwent acidoCEST MRI followed by immunofluorescent histologic analysis. A strong negative correlation was seen between pHe versus GLUT1 expression (r = − 0.75, p < 0.001). After transformation of pH to [H₃O⁺], a strong positive correlation between [H₃O⁺] and GLUT1 expression was observed (r = 0.8, p < 0.001).

Conclusions:

AcidoCEST MRI can measure the extracellular pH of bone marrow affected by multiple myeloma. In this MM orthotopic mouse model, pHe measured by acidoCEST MRI showed strong correlations with the metabolic phenotype of BM tumor assessed by immunofluorescent histological assessment of GLUT1 overexpression.

Introduction

Multiple myeloma (MM), the most frequent cancer to involve the skeleton, is an incurable disease of malignant plasma cells accumulated in the bone marrow (BM). MM incidence is increasing, accompanying the aging population [1]. Despite newer treatments aimed at improving survival rates, patients with refractory and relapsing disease still have a poor outcome [2]. Previous studies point to adaptive responses to hypoxia as an essential element in MM progression and drug resistance, with newer drugs designed to target these mechanisms [2].

Cancers, including MM, tend to favor glycolysis even in aerobic conditions (known as the Warburg effect), due in part to a loss of hypoxia-dependence for the expression of normally hypoxia-responsive genes such as glycolysis enzymes, glucose transporters, and carbonic anhydrases [3–4]. The resulting lactic acidosis and increased conversion of CO₂ to carbonic acid allows both hypoxic response and the Warburg effect to decrease the extracellular pH (pHe) [5]. Lower pHe has been implicated in tumor aggression and metastasis [6]. Specifically, in MM, hypoxia-responsive gene activity and reduced pHe allow tumor cells to evade the immune response [7], increase endothelial cell survival, and reduce apoptosis [8]. As cancer progresses, cells can outgrow the ability of the vascularity to maintain normoxic conditions [4], and hypoxia-responsive genes can become further upregulated even in Warburg-phenotype cancers [9] such as MM [10], thus amplifying these acidification-associated changes.

The degree of pHe reduction due to these hypoxia-responsive metabolic changes depends in part upon the upregulation of glucose transporters (GLUTs) to provide increased glucose uptake. Although most tissues express the glucose transporter GLUT1 at only low levels in adults [11], Warburg-phenotype cancers, including MM [3, 8] express high levels of this hypoxia-responsive gene, making it an attractive marker of Warburg effect changes, hypoxia, and tumor cells in general [12]. Variability in GLUT1 expression levels exists in individuals affected with high-GLUT1 cancers, such as MM, possibly due to regional variability in tumor microenvironment conditions such as oxygen tension [9, 13].

In MM xenografts using high-GLUT1 cell lines (such as RPMI8226 or MM1.S), this diversity in GLUT1 expression has been linked to the ability of MM to survive treatment, through the survival of isolated minimal residual disease cell populations expressing the highest GLUT1 levels [10]. Considering also that the degree of GLUT1 expression in other cancers has been noted as an indicator of high disease severity, poor prognosis, and resistance to treatment [14], and that GLUT1 has value as a therapeutic target [8, 15], GLUT1 expression analysis is of high clinical relevance in characterizing MM xenograft models. In fact, considering that standard diagnostic tools like ¹⁸F-FDG uptake inconsistently correlate with GLUT1 expression [16] or only moderately correlate with pHe [17], the development of our understanding of variability in GLUT1 expression and pHe may have even higher undiscovered clinical importance. Together, GLUT1 expression and pHe may provide mutually corroborating information about the aggressiveness of individual MM tumor sites, due to their metabolic links: GLUT expression levels are correlated with the production of lactate [18] and CO₂ [15], while the acidification ability of cancer cells is similarly associated with the production of lactate and CO₂ [5]. Thus, pHe measurement may provide a readout of metabolic status and aggressiveness that parallels GLUT1 expression analysis.

Several imaging modalities have been employed to quantify pHe, including optical-based, positron emission tomography, magnetic resonance spectroscopy, and magnetic resonance imaging (MRI) methods [19]. Of the available techniques sensitive to pHe, MRI using the chemical exchange saturation transfer (CEST) technique has gained particular attention due to its high sensitivity and spatial resolution [20]. CEST MRI assesses pHe by the exchange rate of magnetic saturation transfer between labile protons on a solute and water. This process occurs in a pH-dependent way and is then translated as an altered MRI signal [20].

The CEST effect can be induced using endogenous or exogenous solutes, but the latter has the benefit of being able to specifically probe pHe, as compared to a combination of intra-and extracellular pH, and provide calibrated, high precision values [19, 21]. Iodinated contrast agents used for clinical computed tomography have an excellent safety profile. They have been repurposed for CEST MRI since the chemical structure contains exchangeable amide protons with resonances sufficiently far from water, allowing for selective irradiation at clinical field strengths. Various analytical approaches allow for pHe measurements that are concentration-independent [22].

AcidoCEST is one particular CEST MRI technique that uses iopamidol and has been studied before to measure in vivo pHe in various settings [22–24]. To date, acidoCEST has not been used to study intra-osseous tumors or MM. This study aimed to determine the feasibility of acidoCEST MRI for in vivo pHe measurement using an orthotopic mouse model of MM compared with GLUT1 immunofluorescence staining as a reference standard.

Methods

Mouse Models

All animal procedures were conducted with the approval of the Institutional Animal Care and Use Committee (IACUC) of the Greater Los Angeles VA Healthcare System in accordance with applicable guidelines and regulations. Orthotopic bone marrow (BM) engrafted MM xenografts were established as previously described [25] in fourteen 6–8-week-old male NSG/NOD scid gamma mice (Stock Number 005557, The Jackson Laboratory, Bar Harbor, ME), with mean initial weight of 27 ± 2 grams. In brief, human RPMI8226 myeloma cells were transfected with a luciferase reporter vector (pGL4.50, Promega, Madison, WI) and injected via tail veins (10–15 × 10⁶ cells/mouse). Engraftment of the cells into the mice skeletons was visualized at day 15 and prior to the MRI, with an intraperitoneal injection of 200 μL of D-luciferin using In Vivo Glo substrate (30 mg/mL diluted in sterile saline) (Promega) followed by optical imaging/X-ray analysis using the IVIS Lumina XRMS platform (Perkin Elmer, Boston, MA) [25]

MRI Acquisition

MRI was performed approximately six weeks after intravenous challenge on a 7T system with a 20 cm bore and 72 mm transceiver volume coil (Bruker Biospin, Inc., Billerica, MA). The mice were anesthetized using 2.0% isofluorane and maintained at a respiration rate of 35–40 bpm with body temperature of 37°C. Imaging was performed with the mice in the supine position to minimize artifacts, and respiratory gating was not used. Initially, a sagittal rapid acquisition with relaxation enhancement (RARE) sequence was performed to visualize the anatomy and tumor affected regions. The image parameters were TR=4000 ms, TE=45 ms, matrix size=256×256, field of view (FOV)=4 cm, in-plane resolution=156×156 μm, slice thickness=2 mm, number of averages=1, and number of slices=6. A fellowship-trained musculoskeletal radiologist (E.Y.C., 7 years of experience) selected a single image from the RARE series, which included an adequately visualized tumor affected region characterized by high signal intensity in bone, to be focused for subsequent imaging (Figure 1). AcidoCEST MRI was then performed using the CEST fast imaging with steady-state free precession (CEST-FISP) [26] sequence before and after intravenous administration of iopamidol 370 mg I/mL (Isovue, Bracco Imaging S.p.A., Milan, Italy). The CEST-FISP sequence parameters were: CEST saturation time=6 seconds with 10 square-shaped pulses for 600 ms each, saturation power=3.5 μT, saturation frequencies=40 (− 3300 to −900 Hz in 600 Hz increments, −750 to 750 Hz in 150 Hz increments, 810 to 2700 Hz in 90 Hz increments, 3000 Hz, and 3300 Hz), TR=3.6 ms, TE=1.8 ms, flip angle=15°, matrix size=128×160, FOV=4 cm, in-plane resolution=313×250 μm, slice thickness=2 mm, and number of averages=1. The entire CEST-FISP sequence was performed twice pre-contrast. Then iopamidol was administered as a 200 μL intravenous bolus within 60 seconds with subsequent infusion at 400 μL/hour for the next 30–40 minutes. Six post-contrast sets of the CEST-FISP sequence were then acquired.

Figure 1.

Spine MRI of a representative mouse. (A-D) Four consecutive sagittal RARE MR images demonstrate the spine on slices 3 (B) and 4 (C), but optimal visualization is clearly present on slice 4. Note the high signal intensity in the L3, L5, and L6 vertebrae representing tumor infiltration (arrows). (E) Sagittal CEST-FISP MR image at the same location as RARE slice 4 (C), also demonstrates high signal intensity tumor affected regions (arrows).

MRI Analysis

AcidoCEST MR images were processed using MATLAB (R2018b, Mathworks, Inc., Natick, MA). The two pre-contrast image sets were averaged at each saturation frequency and a Gaussian spatial smoothing algorithm was applied to improve signal-to-noise ratio [27]. The same algorithm was applied to the six post-contrast image sets. The averaged pre-contrast image was subtracted from the post-contrast image at each saturation frequency to eliminate CEST signals from static endogenous sources. Pixels with contrast below 2√2 were considered below the noise floor and discarded from the analysis [28]. A fellowship-trained musculoskeletal radiologist (A.F.L, 7 years of experience) drew regions of interest (ROIs) outlining the entire tumor affected bones, selected based on high signal on RARE images with minimization of volume averaging and phase artifacts. Within the ROIs, pixel-wise parametric maps of BM pHe values were generated via fitting the CEST spectrum with the Bloch-McConnell equations [24]. Seven parameters were fit as previously described [24], including pH, agent concentration, T1 and T2 relaxation time constants of water, B₀ value, and two scale factors to account for baseline changes in the CEST spectrum. Only pixels with pHe measurements between the range of 6.3–7.4 were included in the mapping since previous phantom data has reported this range to be the most reliable [29]. Average BM pHe values for each ROI were generated via fitting the CEST spectrum with Bloch-McConnell equations [24].

Micro-Computed Tomography

The mice were sacrificed by CO₂ intoxication followed by cervical dislocation upon completion of the MRI and the imaged osseous structures were harvested en bloc and fixed in 10% buffered formalin for 24 hours. Samples were imaged using a μCT scanner (Skyscan 1076, Bruker, Kontich, Belgium) using the following parameters: 1-mm aluminum filter, 90 kV, 112 μA, rotation step=1°, voxel size=35 μm isotropic, and exposure time=90 ms/rotation. Tomographic images were reconstructed using the Feldkamp conebeam backprojection algorithm with noise and ring-artifact reduction.

Immunohistochemistry and Histologic Analysis

After μCT scanning, trimmed carcasses were decalcified in repeated changes of 10% disodium ethylenediaminetetraacetic acid (EDTA, pH 7.0 at room temperature) until complete demineralization as assessed by X-ray imaging. Blocks of tissue containing the CEST-imaged regions were dissected out manually and embedded in paraffin after dehydration through graded alcohols and Pro-Par solvent (Anatech, Battle Creek, MI). Seven-micron sections were collected for immunofluorescent and hematoxylin-eosin (H&E) staining. For immunostaining, antigen retrieval was performed in 10 mM Tris, 1 mM EDTA, 0.05% Tween 20, pH 8.0 for 30 minutes at 80°C, and sections were stained with 1:800 rabbit monoclonal anti-GLUT1 antibody EPR3915 (Abcam, Cambridge, MA) overnight at room temperature, and Dylight-649-conjugated 1:200 anti-rabbit secondary (Vector Labs, Burlingame, CA), in PBS containing 1% BSA and 0.05% Triton X-100. Propidium iodide was included in the secondary as a nuclear stain. Slides were coverslipped with Everbrite hardset aqueous mounting medium (Biotium, Fremont, CA).

Fluorescent images for analysis were acquired on a slide scanner (AxioScan Z1, ZEISS, Thornwood, NY) with a 20× objective, 630 nm LED illumination for the Dylight 649 fluorophore, and a 4-band dichroic and emission filter set. Exposure time was held constant at a level that prevented data clipping due to overexposure, except in the case of bright legbones, in which case the exposure was exactly halved, and the data scaled up by a factor of two. H&E sections were imaged using the 10× objective and transmitted light.

Using Zen Lite software (ZEISS, Thornwood, NY), ROIs on immunostained slides were manually defined by tracing the marrow space of individual bones, and the average pixel brightness values in the Dylight 649 channel were recorded. Hemorrhagic regions and section wrinkles were excluded from analysis. Raw fluorescence levels were normalized by the level observed in unaffected marrow spaces in the same slide. Specifically, low but detectable cellular signal is expected in unaffected marrow sites, including megakaryocytes, as predicted by mRNA studies [30].

Statistics Analysis

All statistical analyses were performed using R (version 4.0.2, R Development Core Team, 2020, Vienna, Austria) and RStudio (version 1.3.959). Descriptive statistics were performed. pH was transformed to the acidic proton concentration, ([H₃O⁺], measured in moles per liter) using the formula pH = −log₁₀[H₃O⁺] since glucose usage follows a linear correlation with [H₃O⁺] production, either as lactate [31] or CO₂. Pearson’s correlation was used to assess the relationship between MRI pHe and [H₃O⁺] versus GLUT1 expression for the whole bone ROIs. p < 0.05 was considered to represent significant findings.

Results

Fourteen mice received tumor cell injections, resulting in successful engrafted tumors in the bone marrow of all mice as observed with bioluminescence imaging on day 15 after challenge, which increased at six weeks. Two mice expired during the experiment and motion artifacts during MRI acquisition compromised images in two other mice. In total 10 mice underwent the complete acidoCEST MRI protocol followed by histologic analysis. Based on tumor burden and quality of visualization of bony boundaries on the anatomic RARE sequence, the spine was selected for CEST-FISP imaging in 8 cases (Figure 1) and the lower extremity in 2 cases.

Extensive involvement of the axial and peripheral skeleton was present, with bone tumor presenting as intra-medullary high signal intensity compared with normal-appearing BM, sometimes associated with endosteal scalloping, cortical erosion, or extra-osseous extension. A broad range of pH values were seen for intraosseous regions, which were subsequently analyzed histologically (Figures 2 and 3).

Figure 2.

Bioluminescence, MR, micro-CT, and histologic correlation on a representative mouse. (A) Bioluminescence image shows skeletal tumor engraftment, greatest in the sacral region. (B) Sagittal RARE MR image depicts bone marrow signal abnormalities throughout the lumbosacral spine corresponding to tumor infiltration. (C) Sagittal CEST-FISP MR image demonstrates the ROIs used to calculate the average pHe (dashed yellow regions). pHe measured 7.1 ± 0.3 and 6.8 ± 0.27 within the S1 and S4 vertebral bodies, respectively. (D) Sagittal micro-CT image shows diffuse rarefaction of trabeculae, most pronounced in the sacral vertebral bodies, including S1 and S4. (E, F) Magnified fluorescent images stained with GLUT1 antibody (red channel) and propidium iodide (blue channel) from S1 (E) and S4 (F) show normalized GLUT1 expression levels of 0.99× in S1 and 1.5× in S4. (G, H) Corresponding H&E-stained images from S1 (G) and S4 (H) confirm diffuse bone marrow infiltration by myeloma tumor cells, with characteristic “clockface nuclei” (black arrows), perinuclear hofs (white arrows), and common mitotic figures (red arrowheads). Bars=50 μm.

Figure 3.

Bioluminescence, MR, micro-CT, and histologic correlation on a different representative mouse. (A) Bioluminescence image shows skeletal tumor engraftment, greatest in the skull and sacral regions. (B) Sagittal RARE MR image depicts diffuse bone marrow signal abnormalities throughout the spine corresponding to tumor infiltration. (C) Sagittal CEST-FISP MR image demonstrates the ROIs used to calculate the average pHe (dashed yellow regions). pHe measured 7.0 ± 0.27 and 6.7 ± 0.38 within the S1 and S2 vertebral bodies, respectively. (D) Sagittal micro-CT image shows diffuse rarefaction of trabeculae, most pronounced in L6 (arrow) but also within the S1 and S2 ROIs (dashed yellow regions). (E) Low magnification immunofluorescent image stained with GLUT1 antibody (red channel) shows heterogenous, but diffusely increased signal. Dashed yellow regions represent the vertebral body contours. (F, G) Magnified fluorescent images stained with GLUT1 antibody (red channel) and propidium iodide (blue channel) from boxed regions in (E) show normalized GLUT1 expression levels of 1.25× in S1 and 1.99× in S2. (H, I) Corresponding H&E-stained images from boxed regions in (E) confirm diffuse bone marrow infiltration by myeloma tumor cells. Bars=50 μm.

GLUT1 signal showed normal staining patterns [11], in general with negligible background signal and bright signal in structures like perineurium and small blood vessels in muscle. In the hematopoietic marrow of the axial skeleton and long bones of the leg, unaffected marrow showed low but detectable cellular signal, which was used as within-slide regions for normalization of expression levels. In affected marrow, tumor cells with features characteristic of multiple myeloma were found densely packing the marrow space, as shown on H&E slides (Figure 2). These cells tended to strongly express GLUT1, with varying levels of signal intensity (Figures 2 and 3).

In general, the high variability of signal levels, combined with inter-modality differences in spatial resolution (i.e., 2 mm MRI versus 7 μm histology slice thicknesses) caused difficulty between direct comparisons on an MRI voxel and histologic pixel basis. However intra-tumoral heterogeneity could be well visualized on both the pH maps and GLUT1 images (Figure 4).

Figure 4.

AcidoCEST pH pixel map, GLUT1 immunofluorescence, micro-CT, and H&E correlation in an exemplary example (same mouse as Figure 2). (A and B) Sagittal pH pixel maps at lower (A) and higher (B) magnification show intra-tumoral heterogeneity within S4, with lower pH at the cranial and caudal portions. (C) Corresponding GLUT1 immunofluorescence image shows differential staining, with higher GLUT1 expression in the cranial and caudal portions (asterisks), and a central region of lower expression. Dashed yellow lines represent the vertebral body contours. (D) Magnified micro-CT image of S4 shows regions of cortical bone erosion (white arrows) and edge of trabecular bone destruction (open arrow). (E) Corresponding H&E-stained image shows diffuse infiltration by tumor cells, virtually replacing all bone marrow. Regions of cortical bone erosion (white arrows) and trabecular destruction front (open arrow), as previously observed on the micro-CT image, are confirmed. Bar=500 μm.

Forty-seven tumor affected bones were identified on the MR images (between 3–7 per mouse), and when analyzed using a whole bone ROI approach, a strong negative correlation was seen between MRI pHe versus GLUT1 expression (r = −0.75, p < 0.001). After transformation of pH to [H₃O⁺], a strong positive correlation between [H₃O⁺] and GLUT1 expression was seen (r = 0.8, p < 0.001) (Figure 5).

Figure 5.

Scatterplots of immunohistochemical versus acidoCEST results. (A) Scatterplot of GLUT1 expression versus pH measurements showing a strong negative Pearson’s coefficient of correlation of −0.75 (p < 0.001). (B) Scatterplot of GLUT1 expression versus H₃O⁺ concentration showing a strong positive correlation of 0.8 (p < 0.001). Open circles highlight ROIs from Figure 1, and open triangles show ROIs from Figure 2.

Discussion

In this study we showed, for the first time, that the pHe of BM affected by MM could be measured using acidoCEST MRI. Whereas assessment using current standard MRI sequences, including fast spin echo-type sequences like RARE, can identify myeloma tumor-affected regions by relying on larger-scale organizational differences between water and fat within BM [32], CEST MRI can account for differences in tumor cellular biology and the microenvironment, which are relevant to aggressiveness and prognosis [7–8]. Indeed, our data demonstrates an inverse correlation between mean pHe measurements (positive linear correlation with H₃O⁺concentration) and BM tumor GLUT1 expression, a marker of aggressive metabolic phenotype.

Published studies have investigated the relationships of both pHe and GLUT1 expression with other clinical measures, and tend to observe only moderate correlations. For instance, ¹⁸F-FDG PET is an established component of the work-up of patients with MM [17], but FDG uptake and pHe in cancers have been shown to correlate only moderately (r = −0.53 to −0.59) [33]. Similarly, although GLUT1 expression is known to have prognostic value [9], and is known to be spatially modulated in patterns similar to those of low-pHe invasive regions [6], its correlation with FDG uptake has been reported to be only moderate (r = 0.25 to 0.66) [16, 35]. Although varying experimental contexts and methods preclude fully conclusive comparisons of these results with our data, acidoCEST and GLUT1 expression do share certain characteristics that might explain a relatively higher mutual correlation than with measures like FDG uptake. First, a constant perfusion is used in the acidoCEST protocol rather than a single bolus injection, which provides stable signals for acidoCEST compared ¹⁸F-FDG PET, and variations in vascularity and perfusion would have a diminished effect on which cells can be probed. Conversely, ¹⁸F-FDG PET signal depends on multiple factors [36], including vascular perfusion, delivery to the cell surface, metabolic trapping, and even tumor size [37]. Second, although ¹⁸F-FDG can ideally reveal uptake capacity, the metabolic fate of glucose can vary with cell phenotype, allowing a cell’s acidification ability to vary for a given amount of glucose absorbed. Thus, variability of acidification capacity can only partially be explained by variability of glucose uptake, in the absence of additional information regarding cellular metabolic phenotype. We propose below that GLUT1 expression provides combined information about both glucose uptake capacity as well as cellular metabolic phenotype.

Glucose metabolism and extracellular acidification are cellular, rather than tumor-level, characteristics that can display diversity across a cell population, such as within a single MM lesion [38], and certainly across multiple sites [13]. Because diversity of clonal subtypes [38] within a MM patient’s lesions may be related to treatment outcome, possibly due to selection of minimal residual disease cells during therapy [10], methods that specifically evaluate cellular phenotype, rather than tumor architecture, may be of critical value. Cellular metabolism includes aerobic as well as anaerobic pathways, even in Warburg-phenotype cancers like MM [5, 13], and acidification due to the reaction of aerobically-produced CO₂ with carbonic anhydrases may be a major component of acidification in cancer [34]. Both of these metabolic acidification pathways are interrogated by the measures we used, because acidoCEST is agnostic to the source of extracellular acid, and because of specific features of GLUT1 expression: GLUT1 upregulation can both cause and be caused by acidosis-related changes in cancer cells. It directly increases glucose uptake capacity, allowing increased glycolysis-based acidosis, and it also is tied to the expression level of carbonic anhydrases through their shared mechanism of hypoxia-inducible factor mediated transcription in both the Warburg effect and in hypoxia [12]. Thus, GLUT1 has some ability to function as a dual reporter of cancer cells’ capacity for both aerobic and anaerobic metabolic acidosis. This ability may explain the previously published diagnostic value of GLUT1 and highlights the value of our observations of pHe variation in the MM xenograft model.

As acidoCEST MRI and ¹⁸F-FDG PET have different mechanisms, ideally both molecular imaging modalities can be used in a complementary fashion to access tumor metabolism. This complementary role could be important in future research, especially in assessing early disease, prognosis, and evaluating tumor metabolic changes in response to treatment [39]. In a simultaneous ¹⁸F-FDG PET/acidoCEST MRI study using a pancreatic cancer mouse model, Goldenberg and colleagues demonstrated the synergistic combination of assessing glucose uptake and tumor acidosis for improving differentiation of a metformin-treated group from a control group during treatment [23].

Our study had limitations. First, we had a small sample size. Second, respiratory gating could have improved motion artifacts and signal-to-noise ratio. Third, we used a single slice acquisition protocol, resulting in limited coverage. Future studies could employ multislice [39] or 3D [40] sequences. Fourth, imaging of bone can be challenging due to its relatively lower T2 values, and in the future CEST with ultrashort echo times may yield higher signal-to-noise and higher quality images. Fifth, total acquisition time was relatively long, which increases the potential for motion artifacts and misregistration. Although the overall imaging time is largely due to the CEST saturation time, some improvements can be made, such as utilization of rapid k-space acquisition techniques such as compressed sensing [41]. Sixth, tissue pH measurements with acidoCEST are influenced by T1 relaxation time. The CEST effect of iopamidol is greater in tissues with longer T1 relaxation times [21], but the effect on pH quantification requires additional study. Finally, we employed a whole bone ROI approach for correlation analysis due to differences in spatial resolution. Future studies could utilize a direct MRI voxel to histologic pixel correlation approach by employing a combination of much thinner MRI slices with use of serial histologic sectioning.

Conclusions

AcidoCEST MRI can measure the extracellular pH of bone marrow affected by multiple myeloma. In this MM orthotopic mouse model, pHe measured by acidoCEST MRI showed an inverse correlation with BM tumor metabolic phenotype assessed by immunofluorescent histological assessment of GLUT1 overexpression.