Tracking Cell Transplants in Femoral Osteonecrosis with Magnetic Resonance Imaging: A Proof-of-Concept Study in Patients

This article is featured in Highlights of This Issue, p. 6105

Osteonecrosis is a debilitating and devastating complication of high-dose corticosteroid therapy: 15%–47% of patients with leukemia and 3%–44% of patients with systemic lupus erythematosus (SLE) develop an osteonecrosis as a result of high-dose corticosteroid therapy (1–4). Seventeen to 22% of these patients progress to hip joint collapse (5, 6), which leads to major long-term morbidities, such as severe joint pain requiring narcotic analgesia, impaired joint function and, severely limited ambulation, ultimately requiring total joint replacement.

The pathogenesis of osteonecrosis is multifactorial and involves corticosteroid-induced ischemia, necrosis of cells of the mesenchymal and hematopoietic lineages, as well as hypertrophy of fat cells that further compress microvessels and thereby perpetuate ischemia (7). Progressive death of cells that provide structural support for the underlying bone ultimately leads to bone collapse. Interventions to save the affected bone are effective only if performed in a preventive manner (8). Various noninvasive and surgical procedures have been used to prevent progression to bone collapse (9). Core decompression has been established in many academic centers and involves drilling a track to an osteonecrosis segment to release the presumed increased pressure in osteonecrosis and facilitate revascularization. Recently, this core decompression procedure has been combined with the delivery of concentrated bone marrow aspirates containing mesenchymal stromal cells (MSC) and enriched osteoprogenitor cells (10–12). These cell transplants are expected to facilitate the regeneration of normal bone marrow in osteonecrosis through direct or indirect mechanisms. However, success or failure of this new treatment can only be diagnosed after several months to years (11, 12). Our group has previously shown in a rat model that intravenously injected iron oxide nanoparticles are taken up by MSCs in the bone marrow and could be tracked with MRI after transplantation into osteochondral defects (13, 14). Others have also noted in vivo labeling capacity of immune cells in the bone marrow (15, 16).

An imaging test, which could directly track transplanted bone marrow cells in vivo, could help us better understand the contribution of these cells to bone repair processes, diagnose complications earlier, and facilitate the development of more successful cell therapies that can prevent bone collapse. The ability to track therapeutic cells noninvasively in vivo could have direct impact on patient management, for example, by stratifying patients with unsuccessful or lost cell transplants to revision surgeries or alternative treatment options. To address this unmet clinical need, we developed an imaging test for the detection of bone marrow cell transplants in osteonecrosis in a “first in-patient” proof-of-concept clinical trial.

This prospective, nonrandomized, HIPAA-compliant proof-of-concept clinical trial was approved by our institutional review board and performed under an investigator-initiated IND (111 154). The study was conducted in accordance with the Belmont Report. We invited pediatric and young adult patients from May 2015 until December 2017 to participate if they met the following inclusion criteria: (i) age 8–40 years, (ii) avascular necrosis of the proximal femur, (iii) planned core decompression with transplantation of autologous bone marrow aspirates, and (iv) willingness to give written informed consent. Patients were excluded if they had: (i) active leukemia, (ii) contraindications to MRI, (iii) hemosiderosis or hemochromatosis, or (iv) if they were pregnant.

We recruited 7 patients (mean age 30 ± 8.4 years; range: 17–38 years) with history of high-dose corticosteroid treatment for leukemia (n = 3), Hodgkin lymphoma (n = 1), asthma (n = 1), SLE (n = 1), or inflammation of unknown origin (n = 1). The patients included 4 female (mean age 30 ± 9.5 years; range: 17–38 years) and 3 male patients (mean age 31 ± 8.7 years; range: 21–36 years). The patients had nine early-stage (ARCO stage II) epiphyseal osteonecrosis in their femoral heads: 3 patients had osteonecrosis of the right femoral head, 2 patients of the left femoral head, and 2 patients had bilateral osteonecrosis.

All patients received core decompression with transplantation of either iron-labeled or unlabeled bone marrow aspirates. To achieve iron labeling of bone marrow cells in vivo, we injected ferumoxytol (Feraheme) intravenously as published previously (13). Four patients with six core decompression procedures received an intravenous injection of ferumoxytol at a dose of 5 mg Fe/kg at 24 to 48 hours before surgery. Three patients with three core decompression procedures did not receive ferumoxytol. Ferumoxytol-labeled or unlabeled bone marrow cells were harvested by an iliac crest aspiration using an autologous cell aspiration and concentration system (Zimmer-Biomet) and mixed with demineralized bone matrix from DePuySynthes. Following core decompression, the graft matrix enriched with labeled or unlabeled bone marrow cells was injected into the access canal by an experienced orthopedic surgeon (S.B. Goodman). Follow-up imaging was performed by MRI.

To compare the clinical outcome of core decompression with and without labeled cells, we enrolled 5 additional control patients with seven femoral osteonecrosis, which were treated with core decompression and unlabeled bone marrow cell transplants and who did not receive serial MRI. These comprised 1 female and 4 male patients (mean age 30 ± 6.8 years; range: 21–38 years) with history of status post high-dose corticosteroid treatment.

Patients included in our study received a preoperative MRI and a follow-up MR scan at 1 week, 4–7 weeks, and 6 months after bone marrow cell transplantation. MRI has been established as a highly sensitive and specific test for diagnosing early epiphyseal osteonecrosis, before joint damage is apparent on bone scans or radiographs (17). MRI scans were obtained with a 3 Tesla MRI scanner (Discovery 750 MR, GE Healthcare), using a 32-channel torso phased array coil and the following pulse sequences: T1-weighted fast spin echo (FSE) sequence [TR = 600 ms (517–721), TE = 15 ms (5.8–19.7), flip angle (FA) = 90° (90°-160°), slice thickness (SL) = 3 mm (3–4.5)], T2-weighted fat saturated FSE sequence [TR = 4450 ms (2399–4450), TE = 61 ms (57–69), FA = 125°, SL = 4 mm (3–4)], short TI inversion recovery sequences [TR=5200 ms (5131–5282), TE=50 ms (47–54), inversion time=170 ms, FA = 111°, SL = 3 mm], and a flow-compensated 2D fast spoiled gradient recalled (FSPGR) sequence (TR = 21.2 ms, TE = 2.2 ms, inter-echo interval 2.2 ms, FA = 25°, SL = 3 mm).

Images were analyzed using Osirix (Pixmeo SARL). The decompression track was equally divided on coronal T2-weighted MR images in three parts: the proximal, mid, and distal track. Each of these areas was manually outlined by one observer, who measured the signal-to-noise ratio (SNR) as the mean signal intensity of the outlined area, divided by the SD of the background noise, which was measured in phase encoding direction within the field of view and outside of the patient. In addition, the iron signal in the same areas was quantified by measuring T2* relaxation times on corresponding T2* maps, which were generated from FSPGR sequences using the T2 fit map plugin of Osirix. Each of the areas was considered as an independent observation for MRI analyses.

The extent of osseous necrosis on imaging studies was graded according to the Association Research Circulation Osseous (ARCO) classification (18, 19). All patients had an ARCO stage II at baseline. Stable disease at 6 months after surgery was defined as equal or improved ARCO stage. Progressive disease was defined as progression to stage III or IV.

All experiments were analyzed using R version 3.4.4. SNR and T2* relaxation times were pairwise compared between labeled and unlabeled cell transplants, and between decompression track areas with and without visible iron-labeled cells, using a mixed-effects model including a random effect term accounting for correlation among the measures within a same patient. A Fisher exact test was applied for comparison of clinical outcomes of labeled and unlabeled cell transplants. In addition, differences in time to progression of osteonecrosis from surgery between labeled and unlabeled cell transplants were assessed by log-rank tests. Because of the small sample size and exploring purpose of this study, a P < 0.05 without adjustment for multiple comparisons was considered to indicate significant differences between experimental groups.

The overall concept of our study is shown in Fig. 1. Patients with osteonecrosis received an intravenous injection of the iron supplement ferumoxytol prior to a scheduled core decompression to label bone marrow cells with iron, which can be detected by a dark signal on MRI. One to 2 days later, the patients underwent a core decompression, bone marrow aspiration from the iliac crest, and transplantation of concentrated iron–labeled bone marrow cells through the decompression track into the osteonecrosis in the femoral head. MRIs were performed before and within 1 week after the surgery, as well as at 4–7 weeks, and 6 months to track transplanted iron-labeled bone marrow cells in osteonecrosis.

MR images before ferumoxytol administration showed a focal osteonecrosis lesion in the proximal femoral epiphysis with a typical serpiginous border on T1- and T2-weighted MR images (Fig. 2A). All osteonecrosis were consistent with stage II lesions according to the ARCO classification: the joint surfaces were intact and there was no evidence for subchondral fractures. This is important, because only joints without signs of bone collapse can be rescued by a core decompression. Next, patients received an intravenous injection of ferumoxytol. Postcontrast MR images showed a significant hypointense (dark) enhancement of the normal bone marrow on T2-weighted MR images (Fig. 2B).

The patients underwent core decompression, harvest, and concentration of iron-labeled bone marrow cells from the iliac crest and transplantation of iron-labeled bone marrow cells into the decompression track. T2-weighted MR images after injection of iron-labeled marrow cells demonstrated hypointense (dark) signal in the decompression track and osteonecrosis (Fig. 2C, 3A and B). In comparison, control patients who had received unlabeled bone marrow cell transplants did not show hypointense signal changes in the access canal (Fig. 3C and D). SNR for ferumoxytol-labeled cell transplants were significantly lower compared with unlabeled cell transplants (33.82 ± 12.43 vs. 129.56 ± 10.93; P = 0.002; Fig. 3E). Likewise, T2* relaxation times, which represent more robust measures of tissue iron concentrations, were significantly lower for ferumoxytol-labeled cell transplants than for unlabeled cell transplants (9.04 ± 0.7 vs. 13.7 ± 2.50; P = 0.02; Fig. 3F)

Within the decompression track of patients who had received iron-labeled cell transplants, we noted areas that showed strong iron signal, presumably representing areas where iron-labeled cells were delivered and areas that showed no iron signal, presumably representing areas where iron-labeled cells were not delivered. We divided each decompression canal in three areas (proximal, mid, and distal decompression canal) and compared SNR and T2* relaxation times of areas where cell transplants could be visually detected or not detected. SNR and T2* relaxation times were significantly lower for areas where cell transplants could be visually detected or not detected (68.45 ± 32.41, P = 0.002; 14.2 ± 2.18, P = 0.007; respectively; Fig. 3E and F).

To evaluate the long-term implications of ferumoxytol administrations on a decompression surgery, we investigated the MRI signal of transplanted cells over time and found a slow decline of the iron signal (Fig. 2D and E): compared with unlabeled controls, SNR and T2* relaxation times of labeled cell transplants were not significantly different at 4–7 weeks (P > 0.05; Fig. 3G and H). This implies either metabolization of the iron label or disappearance of the cell transplant or a combination of both.

To evaluate whether the ferumoxytol administration prior to the core decompression had any effect on bone repair, we compared clinical outcomes of patients who did or did not receive ferumoxytol. Of six femoral heads treated with labeled cell transplants, one (17%) progressed to collapse. Of ten femoral heads treated with unlabeled cells, three (30%) progressed to collapse (Table 1). This difference was not significant (P = 1, Fisher exact test), suggesting that ferumoxytol administration before a core decompression did not adversely affect clinical outcomes. In addition, time to progression of osteonecrosis from surgery between labeled and unlabeled cell transplants were also not significantly different (Fig. 3I; P = 0.8)

We noticed that one joint that progressed to collapse after administration of iron-labeled cells had apparently received less cells compared with all other joints that did not collapse, as indicated by less iron signal in the treated decompression track (Supplementary Fig. S1). We quantified the hypointense (dark) area in the access canal through operator-defined regions of interests. In the femur that progressed to collapse, 3.1% of the area of the decompression track contained labeled cells. In the femur that did not progress to collapse, 16.6% ± 3.5% of the decompression track were covered by cells.

Overall, the collective successful outcome of osteonecrosis treated with core decompression plus cell transplants was better than previously reported for core decompression alone: in our study, 12 of 16 femurs (75%) showed no collapse within 1 year or more after the intervention, compared with success rates of 53%–71% for core decompression only, reported previously (20).

Our data showed that a simple intravenous injection of the iron supplement ferumoxytol before a scheduled core decompression led to iron labeling of the bone marrow in patients. After harvest from bone marrow and transplantation into osteonecrosis lesions, the iron-labeled bone marrow cells could then be tracked with MRI in the early postoperative period. We previously proved in animal models that intravenously injected ferumoxytol nanoparticles are phagocytosed by bone marrow cells (13, 14) and are slowly metabolized over time (21). We found that ferumoxytol nanoparticles accumulate in different cell populations in the bone marrow, which are capable of phagocytosis, including MSCs, macrophages, dendritic cells, and osteoprogenitor cells. In our previous work, we could show that ferumoxytol was taken up by MSCs, stored in lysosomes, and had no effect on the viability and differentiation potential of the iron-containing cells. Furthermore, MSCs could be tracked in rodents with MRI (13).

Other investigators reported the ability to track iron oxide nanoparticle–labeled neural stem cells (22, 23), autologous mesenchymal stem cells (24), and dendritic cells (25) with MRI in patients. In these previous studies, autologous cells were first harvested, then iron-labeled in cell cultures, and then transplanted. Our approach is different in that we labeled bone marrow cells in vivo by intravenous injection of an FDA-approved iron supplement. The iron-labeled cells could then be detected with MRI. This in vivo labeling approach is more practical in a clinical setting because it does not require any manipulation of the therapeutic cells and thereby, enables bone marrow cell harvest and transplantation in one surgery.

Corticosteroid-induced osteonecrosis leads to bone collapse in about 26% of patients (6). Core decompression can prevent or delay bone collapse in at least 50% of these patients (20). The addition of bone marrow cell transplants to classical decompression surgeries has improved outcomes compared with core decompression alone (11, 26, 27). It is discussed controversially, whether bone marrow cell transplants support osteonecrosis repair directly or indirectly: Wang and colleagues and Lee and colleagues suggest that mesenchymal stromal cells and osteoprogenitor cells can directly regenerate bone injuries (28, 29), Murphy and colleagues and Linero and colleagues suggest that MSCs support tissue repair through indirect paracrine mechanisms (30, 31), and Lim and colleagues and Pepke and colleagues question any therapeutic effect of bone marrow cells (32, 33). However, Lim and colleagues and Pepke and colleagues also mention that numerous factors such as number of cells, area of transplantation, disease stage, and follow-up period play an important role in clinical outcome. Our imaging test could be used to correlate the number and distribution of transplanted bone marrow cell transplants with outcomes. This information could be used to understand and optimize the effect of bone marrow cell transplants on bone repair outcomes. In addition, our imaging test could discriminate between successfully engrafting or lost cell transplants in the early postoperative period. This information is important for the treating physician, who could stratify patients with failed cell transplants to alternative operative and nonoperative treatment options.

Our data showed that iron labeled and unlabeled cell transplants showed similar success rates in preventing bone collapse, confirming that our in vivo labeling approach did not affect the efficacy of the therapeutic cells. Previous studies showed that high intracellular iron concentrations and iron overload can impair chondrogenic differentiation (34, 35) as well as osteogenic differentiation of MSCs and inhibit osteoblast activity, while facilitating osteoclast function (36). Our group has shown previously that this effect is dose-dependent and that the proliferation and function of MSCs is not impaired when the cells are loaded with less than 10 pg Fe per cell (37). Our investigations in preclinical models showed that intravenous injection of ferumoxytol (28 mg Fe/kg) leads to uptake of approximately 4.3 pg Fe per cell, which is below this threshold (13).

It has been reported previously, that MSCs and osteoprogenitor cells bear the potential to migrate to the site of osteonecrotic bone lesions (38, 39). In principle, our imaging test should be also suitable to investigate this process. Because of the avascular center, a newly developed osteonecrotic lesion takes up less ferumoxytol than surrounding healthy bone marrow. In case of an intrinsic migration of ferumoxytol-labeled bone marrow cells from normal marrow to the site of injury, osteonecrotic lesions could become hypointense (dark) over time in MRI. However, tracking intrinsic in vivo migrations of ferumoxytol labeled cells to osteonecrosis might involve fewer cells and therefore, require more sensitive pulse sequences.

We found that the quantity of bone marrow cells transplanted into osteonecrosis correlates with clinical outcomes, which is in accordance with previously published findings (10, 12, 40, 41). Hernigou and colleagues analyzed the number of transplanted progenitor cells in patients with osteonecrosis and found that femoral heads that received a low number of transplanted cells had higher risks of bone collapse than femoral heads that received a high number of cells (10). In another study, Hernigou and colleagues evaluated the amount of cells needed for treatment of nonunion of the tibial shaft in 60 patients and found a positive correlation between the volume of mineralized callus and the number of cells and concentration (40).

Bone marrow cells from patients with osteonecrosis may have less bone regeneration potential compared with healthy patients (12). Future studies will compare the in vitro bone-forming capacity of MSCs and osteoprogenitor cells in patients with osteonecrosis and healthy controls. In patients who underwent bone marrow transplantation, MSC from the donor might be more effective in regenerating bone. Our imaging test can track the location, estimate the quantity, and provide information about the phagocytic activity of bone marrow cells. For example, we previously noted that normal bone marrow cells of young patients take up more iron compared with bone marrow cells of older patients (42, 43). Our imaging test cannot determine the efficacy of the transplanted cells to regenerate bone. However, in case our imaging studies suggest low quantities of iron labeled cells in the delivered bone marrow aspirates, we could increase the number of transplanted cells and/or add “off the shelf” MSC products or banked donor MSCs. We previously established ferumoxytol labeling procedures for ex vivo labeling of donor MSCs (21, 44).

In summary, we provide proof-of-concept for a new imaging test, which allows to track autologous bone marrow–derived cell transplants in patients with MRI, after a simple intravenous injection of an FDA-approved iron supplement. The ability to directly detect and track iron-labeled therapeutic cells in vivo, in patients, could help us to recognize inter-individual differences in the delivered quantity and location of transplanted cells and correlate results with tissue repair outcomes. This could ultimately improve our ability to develop more successful cell therapies. This new imaging test could become a powerful new tool to monitor the delivery and engraftment of bone marrow–derived therapeutic cells noninvasively in patients with cancer with the ability to directly impact patient management. Future studies under this clinical trial will correlate MRI signal characteristics of-iron labeled cells with clinical outcomes.

No potential conflicts of interest were disclosed.

Conception and design: A.J. Theruvath, H. Nejadnik, A.M. Muehe, S.B. Goodman, H.E. Daldrup-Link

Development of methodology: A.J. Theruvath, H. Nejadnik, S.B. Goodman, H.E. Daldrup-Link

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.J. Theruvath, H. Nejadnik, A.M. Muehe, F. Gassert, N.J. Lacayo, S.B. Goodman, H.E. Daldrup-Link

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.J. Theruvath, H. Nejadnik, A.M. Muehe, F. Gassert, H.E. Daldrup-Link

Writing, review, and/or revision of the manuscript: A.J. Theruvath, H. Nejadnik, A.M. Muehe, F. Gassert, N.J. Lacayo, S.B. Goodman, H.E. Daldrup-Link

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

Study supervision: A.J. Theruvath, A.M. Muehe, S.B. Goodman, H.E. Daldrup-Link

We thank members of the Daldrup-Link lab for their helpful reviews and discussions of our study design and research results. We thank Jin Long from the Quantitative Sciences Unit at Stanford University and Ketan Yerneni for their excellent statistical consulting. Part of this work was performed at the Richard Lucas Center for MR Imaging and at the Stanford Nano Shared Facilities (SNSF) at Stanford University. This work was supported by research grant 2R01AR054458 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases.

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.