Feridex Preloading Permits Tracking of CNS-Resident Macrophages after Transient Middle Cerebral Artery Occlusion

Abstract

At this time, the pathophysiology of macrophage involvement and their role in stroke progression are poorly understood. Recently, T₂- and T₂*-weighted magnetic resonance imaging (MRI), after intravenous administration of iron-oxide particles, have been used to understand the inflammatory cascade. Earlier studies report that image enhancement after stroke is from iron-laden macrophages; however, they do not account for potential blood-brain barrier disruption and nonspecific contrast enhancement. In this study, spontaneously hypertensive rats were preloaded with Feridex 7 days before stroke, permitting the labeling of bone-marrow-derived macrophages. Three-dimensional gradient-echo imaging showed average signal decreases of 13% to 23% in preloaded animals, concentrated on the lesion periphery and reaching a maximum on days 2 to 4 after stroke. Immunohistochemistry showed ED-2⁺, PB⁺, MHC-II⁺ and TNF-α⁺ perivascular macrophages (PVM), meningeal macrophages (MM), and choroid plexus macrophages (CPM). ED-1⁺ and IBA⁺ tissue macrophages and/or activated microglia were located throughout the lesion, but were PB⁻. These findings indicate the following: (1) Feridex preloading permits tracking of the central nervous system (CNS)-resident macrophages (PVM, MM, and CPM) and (2) CNS-resident macrophages likely play an integral role in the inflammatory cascade through antigen presentation and expression of proinflammatory cytokines. Further refinement of this method should permit noninvasive monitoring of inflammation and potential evaluation of antiinflammatory therapies in preclinical models of stroke.

Keywords

Feridex inflammation macrophages MRI stroke

Introduction

Ischemia-reperfusion (I-R) triggers the production of inflammatory cytokines, resulting in an increased expression of cellular adhesion molecules, microglia activation, and leukocyte infiltration (Wang et al, 2007). Although microglia play an integral role in the hyperacute time frame (minutes), macrophages are recruited heavily in the subacute to chronic phase of stroke (days) (Flaris et al, 1993; Hallenbeck et al, 2005; Kochanek and Hallenbeck, 1992). The importance of macrophage activity to overall stroke progression, from the hyperacute phase to the chronic phase, has largely been underappreciated. Recently, it has been suggested that they may play a larger role in the initiation and continuation of the inflammatory cascade (Becher et al, 2006; Greter et al, 2005).

Macrophages of the brain may be divided into two major subtypes, namely tissue macrophages and central nervous system (CNS)-resident macrophages (Lassmann et al, 1993; McMenamin, 1999). Tissue macrophages are derived from activated microglia that have become phagocytic. The CNS-resident macrophages include perivascular macrophages (PVM), meningeal macrophages (MM), and macrophages within the choroid plexus (CPM). Given their strategic location within the CNS, these macrophages (PVM, MM, and CPM) may be considered as the ‘gatekeepers’ of the brain. It has been speculated that PVM and MM may serve as antigen-presenting cells, expressing a major histocompatibility complex (MHC) class II and mediating the immune response through an interaction with T cells (Angelov et al, 1998; Becher et al, 2006; Hickey and Kimura, 1988; Polfliet et al, 2002). They have also been reported to express proinflammatory cytokines, suggesting a direct involvement in the progression of disease (Bauer et al, 1996; Becher et al, 2006; Lehrmann et al, 1998). Given this duality, understanding the role(s) of CNS-resident macrophages in stroke is vital.

The development of iron-oxide contrast agents, such as superparamagnetic iron oxides (SPIOs, size 60 to 150 nm) and ultrasmall paramagnetic iron oxides (USPIOs, size 10 to 40 nm), prompted an interest in tracking inflammation through macrophage activity (Thorek et al, 2006; Wang et al, 2001; Weissleder et al, 1990). SPIOs and USPIOs may be internalized by macrophages through receptor-mediated endocytosis (SR-A types I/II) or phagocytosis (Raynal et al, 2004), making them visible on T₂-weighted imaging (T₂WI) and gradient-echo imaging (GRE). Earlier work in experimental (Rausch et al, 2001; Saleh et al, 2004b; Wiart et al, 2007) and clinical (Manninger et al, 2005; Saleh et al, 2004a) stroke has shown promise in the ability to monitor iron-laden macrophages; however, these studies do not account for potential blood-brain barrier disruption and pooling of contrast on the lesion periphery before macrophage uptake. As iron-oxide administration has generally involved systemic injection shortly after insult, the question arises as to whether nonspecific contrast enhancement is a problem.

In this study, we preloaded SHR (spontaneously hypertensive) rats with the SPIO Feridex (Bayer HealthCare Pharmaceuticals, Wayne, NJ, USA) 7 days before transient middle cerebral artery occlusion (tMCAO) (Figure 1). We hypothesize that this will permit the iron labeling of bone-marrow-derived macrophages and the imaging of their activity after tMCAO. We also hypothesize that this particular population of macrophages, termed ‘CNS-resident macrophages,’ is essential for the initiation and maintenance of the inflammatory response through antigen presentation and cytokine expression. Although this is an accepted concept in experimental and clinical multiple sclerosis, it is new to stroke and has yet to be investigated thoroughly. The results presented herein provide a basis for potential investigation of antiinflammatory therapies in preclinical models of stroke.

Figure 1

Hypothesized method of Feridex accumulation with 7-day preloading. Feridex has a blood half-life of 5 to 6h, the majority of which is cleared by the reticuloendothelial system (RES). The remainder accumulates in the bone marrow before ischemic insult. After ischemia–reperfusion (I–R), these iron-laden macrophages are recruited to the stroke region. Using noninvasive magnetic resonance imaging (MRI), their involvement in the acute and subacute/chronic phase of the inflammatory cascade may be monitored. iv, intravenous; tMCAO, transient middle cerebral artery occlusion.

Materials and methods

Animal Preparation

This study was approved by the IACUC (Institutional Animal Care and Use Committee) of the National Institute of Neurological Disorders and Stroke, National Institutes of Health (IACUC protocol #1225-05). A total of 16 SHR rats (SHR/NCrl: male, 3 to 4 months, 350 ± 20 g, Charles River Laboratories, White Plains, NY, USA) were anesthetized with isoflurane (5% induction, 2.5% maintenance) delivered in a 2:2:1 mixture of breathing quality air, nitrogen, and oxygen. Twelve animals were subjected to 30 mins of the right middle cerebral artery occlusion, followed by reperfusion (tMCAO) using the intraluminal suture model (Koizumi et al, 1986; Longa et al, 1989). The remaining four animals underwent sham surgery (no suture insertion). The rats were monitored for 2 h after recovery, and were then returned to their cages.

Feridex Administration

The animals were divided into four groups (n = 4 per group). Group 1 (Preloaded tMCAO) animals were injected with 0.5 mL of intravenous Feridex (16 mg/kg of 11.2 mg/mL iron) 7 days before tMCAO; group 2 (Postloaded tMCAO) animals were injected with Feridex 5 mins after tMCAO; group 3 (Control tMCAO) animals had tMCAO without Feridex administration; and group 4 (Preloaded SHAM) animals were injected with Feridex 7 days before sham surgery.

Magnetic Resonance Imaging

Imaging was performed using a 7.0T/30 cm imaging spectrometer (Bruker Biospin, Billerica, MA, USA) equipped with ± 45 G/cm self-shielded gradients (Resonance Research, Billerica, MA, USA). Diffusion-weighted echo-planar imaging (DW-EPI), T₂WI, and three-dimensional (3D) GRE were performed on days 1, 2, 3, 4, and 7. Twelve 1.5-mm-thick slices were acquired using a field of view (FOV) = 25.6 × 25.6 mm². For DW-EPI, the matrix size = 96 × 96; TR/TE = 6,200/46 ms; Δ = 13 ms; δ = 4 ms; b-values = 0, 250, 500, 750, and 1,000 s/mm² (acquired for all three principle directions, x, y, and z); and NEX = 4. For T₂WI, the matrix size = 128 × 128; TR = 2,000 ms; TE = 15, 30, 45, 60, 75, 90, 105, and 120 ms; and NEX = 1. For 3D-GRE, the matrix size = 256 × 256 × 256; TR/TE = 30/8 ms; flip angle = 30°; and NEX = 1. These parameters yielded an in-plane resolution of 267 μm for DWI and 200 μm for T₂WI. Three-dimensional GRE scans had an isotropic resolution of 100 μm. Total scan time was ~53 mins per animal.

Inclusion/Exclusion Criteria

Animals (n = 3) were excluded for lesion volumes <50 mm³, considered to be an incomplete middle cerebral artery occlusion (Gerriets et al, 2004). All remaining animals (n = 13) were included in the study, yielding a success rate of 13 of 16 (81%).

Immunohistochemistry

After the final magnetic resonance imaging (MRI) on day 7, the rats were killed by means of transcardial perfusion fixation after an overdose of isoflurane anesthesia. Brains were extracted and placed in 4% paraformaldehyde for 1 to 2 days and then transferred overnight to 20% sucrose. They were then snap frozen in isopentane and stored at −73°C. Serial 16-μm-thick sections were collected, beginning at the olfactory bulb (+ 5.0 from the bregma) and extending to the start of the cerebellum (−13.0 from the bregma). For every set of 10 slides, the cryostat was advanced 500 μm.

ED-1, ED-2, and ionized calcium-binding adaptor molecule (IBA) staining were performed for the identification of macrophages and microglia (ED-1: nonspecific macrophage marker (monocytes, macrophages, and activated/phagocytic microglia); ED-2: specific to well-differentiated macrophages; and IBA: specific to activated microglia). The sections were incubated for 10 mins in methanol containing 0.6% hydrogen peroxide, permeabilized with 0.3% Tween (Mallincrodt Baker, Phillipsburg, NJ, USA) for 1h, and blocked with 5% NDS (Jackson Laboratories, West Grove, PA, USA) for 1 h. Any of three primary antibodies were used on consecutive slides, namely ED-1 (1:200, CD68, Abcam, Cambridge, MA, USA), ED-2 (1:100, CD163, AbD Serotec, Raleigh, NC, USA), or IBA (1:150, Iba1, Wako, Richmond, VA, USA). Negative control sections were prepared with mouse IgG (1:1,500, Jackson Laboratories, West Grove, PA, USA) for ED-1 and ED-2, and with rabbit IgG (1:1,500, Jackson Laboratories, West Grove, PA, USA) for IBA. The slides were incubated in a humidity chamber at 4°C overnight. Biotinylated secondary antibodies (donkey antimouse or antirabbit, 1:1,500, Jackson Laboratories, West Grove, PA, USA) were applied to all sections and incubated for 1 h, followed by the Avidin–biotin–peroxidase complex (ABC-Elite Kit, Vector Laboratories, Burlingame, CA, USA) for 1 h. The binding between antibody and antigen was visualized with diaminobenzidine (DAB Kit, Sigma, St Louis, MO) for 2 to 5 mins. All slides were counterstained with nuclear fast red for nuclear identification.

Prussian blue (PB) staining was performed for the identification of iron-labeled cells (PB: specific to ferric iron (Fe³⁺)). Rehydrated sections were immersed in a stock solution of 10% potassium ferrocyanide for 5 mins, followed by a working solution composed of 3% concentrated hydrochloric acid and 7% potassium ferrocyanide for 30 mins. All slides were counterstained with nuclear fast red for nuclear identification.

ED-1, ED-2, and PB double staining was performed to verify which population(s) of macrophages contain iron. Single sections were initially run through the standard immunohistochemistry protocol, with staining of contiguous sections for ED-1 or ED-2. After the DAB step, PB staining was performed as per protocol, followed by counterstaining with nuclear fast red for nuclear identification.

ED-2, MHC-II, tumor necrosis factor-α (TNF-α), and interleukin-1 (IL-1) expressions were used for a rough assessment of antigen presentation and cytokine expression within the well-differentiated macrophage population. Double immunofluorescence for ED-2:TNF-α, MHC-II:TNF-α, ED-2:IL-1, and MHC-ILIL-1 was performed as follows: incubations with TNF-α (1:100, YC-032, Yanaihara Institute, Shizuoka, Japan) and IL-1 (1:100, AF-501-NA, R&D Systems, Minneapolis, MN, USA) were followed by incubation with fluorescein-conjugated secondary antibodies (1:200, Jackson Laboratories, West Grove, PA, USA); incubations with ED-2 (1:100) and MHC-II (1:250, OX6, Abcam, Cambridge, MA, USA) were followed by incubation with rhodamine-conjugated secondary antibodies (1:200, Jackson Laboratories, West Grove, PA, USA). All slides were mounted using Prolong Antifade (Invitrogen, Carlsbad, CA, USA).

Image Analysis: Magnetic Resonance Imaging

Image reconstruction was carried out using Paravision's Image Processing and Display Software (PV3.0.2, Billerica, MA, USA). Image analysis and parameter map production (apparent diffusion coefficient (ADC), T₂) were performed using MIPAV Software (BIRSS, NIH, Bethesda, MD, USA), and routines were written in IDL (Research Systems Inc., Boulder, CO, USA). Lesion volumes were collected manually by two readers using T₂ parameter maps. ADC and T₂ values were determined for volumes of interest in both healthy and ischemic tissues. To correct for edema, the ipsilateral (right) and contralateral (left) hemispheres were compared (edema = (ipsi—contra)/contra). Interreader variability was determined and the mean volumes (uncorrected and corrected) of the two readers were used for statistical analysis.

A separate reader blinded to lesion volume determined Feridex-positive and Feridex-negative regions on 3D-GRE images. Feridex-positive volumes were collected manually, with the equivalent contralateral region serving as a Feridex-negative control. Percentage signal intensity reduction was calculated as: [% SI decrease = (SI_{Feridex-positive} −SI_{Feridex-negative})/SI_{Feridex-negative}) × 100].

Image Analysis: Immunohistochemistry

For ED-1, ED-2, IBA, and PB staining, photos (20 x) were taken using an Axioplan light microscope (Carl Zeiss Microimaging GmbH, Berlin, Germany) in 10 locations along the entire lesion periphery (interior and exterior). Additional photos were taken in the lateral ventricles (left and right) for assessment of the choroid plexus. These 12 images were acquired for the central portion of the lesion, corresponding to the central three imaging slices (Figure 2). Cell counts were performed using the ImageJ software (BIRSS, NIH, Bethesda, MD, USA) by a single reader blinded to all animal identifiers.

Figure 2

Schematic representation of imaging slice selection and the corresponding immunohistochemistry. Top left: sagittal T₂-weighted image with slice package (dotted lines) for T₂-weighted and diffusion-weighted imaging. Bottom left (A-C): T₂-weighted images corresponding to the central three imaging slices (shaded area). For ED-1, ED-2, IBA, and PB staining, photos were taken at three locations with regard to the bregma (+ 0.25, −1.25, −2.75). For each location, 12 regions were investigated (numbered 1 to 12), 10 along the lesion periphery and 2 within the lateral ventricles.

For double staining and double immunofluorescence, photos (40 x) were taken in multiple locations along the lesion periphery, with particular attention to perivascular regions and the meninges along the exterior edge. For double immunofluorescence, additional photos were taken in the lateral ventricles (left and right) for assessment of the choroid plexus.

Statistical Analysis

Data are presented as mean (s.d.). Normality was tested using the Shapiro–Wilk and Kolmogorov–Smirnov tests. For the temporal evolution of lesion volume, edema, ADC, and T₂ values, one-way analysis of variance (ANOVA) with post hoc Tukey's tests were performed. For between-group comparisons (i.e., lesion versus healthy tissue), two-way between-groups ANOVA and Kruskal–Wallis tests were used. For Feridex-positive volumes, the comparison of %SI decrease was performed using the two-way between-groups ANOVA and one-way ANOVA with post hoc Tukey's tests. For immunohistochemistry, the comparison of cell counts per staining method and region was carried out using Kruskal–Wallis tests, with post hoc testing using the Mann-Whitney test. P-values < 0.05 were considered statistically significant.

Results

Lesion Progression After Ischemia–Reperfusion

Interreader variability was 20mm² (20) (r = 0.961%, 96.2% within ± 2 s.d.). There was a significant increase in lesion volume from day 1 (150 mm³ (50)) to day 2 (220mm³ (40), P < 0.01), which persisted up to day 4 (210mm³ (50), P = 0.967). Concomitant with this volume increase was a significant increase in vasogenic edema (10% (7), P = 0.025), increased T₂ (90 ms (5), P <0.01), and a persistent reduction in ADC (0.62 × 10⁻³ mm²/s (0.06), P < 0.01) when compared with those in the healthy tissue. By day 7, this edema had subsided (2% (2), P = 0.334), and lesion volumes had returned to approximately day 1 values (140 mm³ (40), P = 0.987). ADC values remained significantly reduced until day 2, with renormalization by day 4 (0.75 × 10⁻³ mm²/s (0.07), P = 0.951). T₂ values remained elevated through day 7 (75 ms (4), P < 0.01), indicative of permanent damage.

Feridex Visualization of the Inflammatory Response

Figure 3 shows 3D-GRE images for group 1: Preloaded tMCAO on days 1, 4, and 7 (top row); and for groups 2 to 4: Postloaded tMCAO, Control tMCAO, Preloaded SHAM on day 4. Within the Preloaded tMCAO group, vasodilation and blood-vessel darkening were evident on day 1 GRE images (Figure 3, top left, arrows). Total lesion hypointensity was significantly greater in magnitude than that in the Preloaded SHAM group for day 1 (P < 0.01), day 4 (P < 0.05), and day 7 (P < 0.05), with a trend toward significance for day 3 (P = 0.107). Marked hypointensity developed along the lesion periphery by day 2 (−21%), reaching a maximum by day 4 (−22% (8)), and remaining constant through day 7 (−23% (3)). The GRE signal decrease along the lesion periphery (arrows) was significantly greater in magnitude than that in the Preloaded SHAM group for all days after I–R injury (P < 0.01 for days 1, 2, 4, and 7; P < 0.05 for day 3). There seemed to be no major GRE signal changes in the Postloaded tMCAO group (Figure 3, bottom left). There was a slight GRE signal decrease in the Control tMCAO group (~3% to 10%) (Figure 3, bottom middle). Three-dimensional GREs for the Preloaded SHAM (Figure 3, bottom right) group were considered negative.

Figure 3

Feridex visualization of the inflammatory response. Day 1, 4, and 7 coronal GRE images for group 1: Preloaded transient middle cerebral artery occlusion (tMCAO) (top row). Vasodilation and blood-vessel darkening were evident on day 1 (top left). Marked hypointensity developed along the lesion periphery, reached a maximum by day 4 (top middle), and remained constant through day 7 (top right). Day 4 coronal GRE images for group 2: Postloaded tMCAO (bottom left), for group 3: Control tMCAO (bottom middle), and for group 4: Preloaded SHAM (bottom right). The GRE signal decrease along the lesion periphery (top middle) was significantly greater in magnitude than that in the Preloaded SHAM group (bottom right). The GRE signal decrease within the lesion core was not statistically different than that in the Preloaded SHAM group (P ≥ 0.183).

Table 1 provides a detailed, quantitative summary of 3D-GRE signal evolutions for individual groups after tMCAO. For the Preloaded tMCAO group, time-averaged signal changes for total lesion (−15% (4)) and the lesion periphery (−22% (2)) were significantly different from those for Postloaded tMCAO, Control tMCAO, and Preloaded SHAM groups (P < 0.05). Signal changes within the infarct core (−11% (3%)) were significantly different from those in the Postloaded tMCAO and Preloaded SHAM groups (P < 0.05), but were similar to those in the Control tMCAO group (−7% (3), P = 0.09).

Table 1

3D-GRE signal changes after 30 mins transient middle cerebral artery occlusion (tMCAO)

Time (days)	30 mins tMCAO					SHAM

	Group 1: 7 days Preloaded			Group 2: Postloaded	Group 3:Control	Group 4: 7 days Preloaded

	Total	Core	Periphery
1	−13 (3)**	−15 (20)	−13 (4)**	−3 (3)	−4	4 (4)
2	−20	−10 (10)	−21	2 (2)	−10	2 (1)
3	−20 (10)	−10 (20)	−20 (8)*	8 (5)	−10	2 (5)
4	−9 (7)*	−8 (2)	−22 (8)**	−2 (5)	−3	−3 (3)
7	−15 (2)*	−10 (9)	−23 (3)**	−10 (10)>	−7	−2 (5)
Average	−15 (4)^#	−11 (3)*	−22 (2)**^{# †}	0 (1)	−7 (3)	1 (3)

3D-GRE, three-dimensional gradient-echo imaging

Significance with P < 0.01, versus SHAM.

Significance with P < 0.05, versus SHAM.

Significance with P < 0.05, versus Postloaded tMCAO, Control tMCAO.

†

Significance with P < 0.05, Periphery versus Core.

ED-1 Correlates with IBA and ED-2 Correlates with Prussian Blue

Figure 4 shows the correlations between ED-1, IBA, ED-2, and PB immunohistochemical staining within the Preloaded tMCAO group. ED-1 and IBA staining were well-correlated on a region-by-region basis (r = 0.521, P < 0.01), with no significant difference in cell counts (P = 0.33, averaged across R1–R12). ED-2 and PB staining were well-correlated visually; however, although the two had the same magnitude of cell counts, ED-2 counts (20 (20)) were significantly higher than were PB counts (10 (20)) by an approximate factor of two (P < 0.01).

Figure 4

Correlation between ED-1/IBA and ED-2/PB staining. ED-1⁺ staining visually correlated with ionized calcium-binding adaptor molecule (IBA⁺) staining and there was no significant difference in cell counts. ED-2⁺ staining (arrows, inset, brown) visually correlated with Prussian blue (PB⁺) staining (arrows, inset, blue). However, although the two had the same magnitude of cell counts, ED-2⁺ was significantly higher than PB⁺ by an approximate factor of two (P < 0.01). Colocalized images (20 x), scale bar = 50 μm. NS, nonsignificant.

On the inner edge of the peri-infarct area (R1–R5), ED-2⁺ and PB⁺ staining were observed primarily in perivascular locations and in a limited quantity in the parenchyma. On the outer edge of the peri-infarct area (R6–R10), ED-2⁺ and PB⁺ staining were observed primarily within the meninges. Within the infarct core, ED-2⁺ and PB⁺ staining were extremely limited, if not absent. ED-1⁺ and IBA⁺ staining were present throughout the infarct. Similar to ED-2⁺ and PB⁺ staining, ED-1⁺ and IBA⁺ staining were greater in the peri-infarct area than that in the core.

Table 2 provides a detailed, quantitative summary of cell counts for ED-1, IBA, ED-2, and PB immunohistochemical staining for individual groups on day 7 after tMCAO. In Postloaded tMCAO and Control tMCAO groups, although the IBA and ED-1 cell counts were of the same magnitude, the IBA cell counts were significantly higher than the ED-1 cell counts (P < 0.01). ED-2 counts and PB counts were significantly lower than ED-1 counts and IBA counts across all groups (P < 0.01). ED-2 counts were similar for all tMCAO animals, but differences were present for PB. The Preloaded tMCAO group had significantly higher PB counts than all other groups (P < 0.01). The PB cell counts for the Postloaded tMCAO and Control tMCAO groups were similar to that for the Preloaded SHAM group (P = 0.782, 0.347), with means close to zero. Within the Preloaded SHAM group, there were no differences between ED-1, ED-2, and PB counts (P ≥ 0.319), and IBA counts were extremely low.

Table 2

Cell counts on a group-by-group basis (per 20 × field)

Staining Method	30 mins tMCAO			SHAM
				SHAM
	Group 1: 7 days Preloaded	Group 2: Postloaded	Group 3: Control	Group 4: 7 days Preloaded
ED-1	460 (290) [0-1127]	320 (250) [0-945]	230 (160) [0-544]	1 (1) [0-10]
IBA	520 (290) [0-1294]	590 (290) [37-1101]	460 (230) [11-906]	30 (20) [0-216]
ED-2	20 (20) [0-92]*	8 (1) [0-76]*	10 (20) [0-67]*	1 (4) [0-45]
PB	10 (20) [0-85]*^{# †}	2 (9) [0-85]*^#	1 (2) [0-11]*^#	1 (2) [0-10]

IBA, ionized calcium-binding adaptor molecule; tMCAO, transient middle cerebral artery occlusion; PB, Prussian blue.

Versus IBA and ED-1 (within group), significance with P < 0.01.

Versus ED-2 (within group), significance with P < 0.01.

†

Versus all groups for PB staining, significance with P < 0.01.

Verification of ED-2:PB Double Staining

Figure 5 shows double staining for ED-2:PB (top and middle) and ED1:PB (bottom) for CNS-resident macrophages located in the meninges. ED-2⁺ and PB⁺ macrophages were identified (top and middle row), validating the presence of iron in well-differentiated macrophages. Although some of the ED-2⁺/PB⁺ macrophages stained brown with an intense blue color (middle right), the majority stained brown and light blue (top). ED-1⁺ macrophages were PB negative (bottom row), confirming the absence of iron in monocytes/macrophages, and in activated microglia (bottom right).

Figure 5

Confirmation of ED-2:PB double staining. ED-2⁺ and Prussian blue (PB⁺) macrophages were identified (top, middle row), validating the presence of iron in well-differentiated macrophages. ED-1⁺ and PB⁻ macrophages were identified (bottom row), confirming the absence of iron in monocytes/macrophages and in activated microglia. Brown = ED-2⁺ (top, middle) or ED-1⁺ (bottom), Blue = PB⁺, background = nuclear fast red counterstain. Colocalized images (40 x), scale bar = 25 μm.

Central Nervous System-Resident Macrophages Express MHC-II and TNF-α

Figure 6 shows the colocalization of ED-2⁺ − and PB⁺-stained macrophages and the colocalization of ED-2⁺ macrophages expressing MHC-II complex and TNF-α. ED-2⁺ macrophages were isolated to three distinct locations, the meninges, choroid plexus, and perivascular regions. These macrophages stained positive for the MHC-II complex and TNF-α. An additional small number of ED-2⁻, PB⁻, and MHC-II⁺ cells was observed scattered throughout the lesion area (not shown). It is likely that these correspond to ED-1⁺ − and IBA⁺-activated microglia. Overall, IL-1 staining was weak, with high background fluorescence and nonspecific staining. For these reasons, IL-1 results were excluded from this study.

Figure 6

CNS-resident macrophages express major histocompatibility complex class II (MHC-II) complex and tumor necrosis factor-α (TNF-α). Immunohistochemistry (Set A, contiguous sections): ED-2⁺ and PB⁺ macrophages (arrows) were identified within the meninges, choroid plexus, and perivascular locations (colocalized images, (20 x), scale bar = 25 μm). Brown = ED-2⁺, blue = PB⁺, background = nuclear fast red counterstain. Immunofluorescence (Set B, contiguous sections): These ED-2⁺ and PB⁺ macrophages, located at the interface between the blood-brain barrier and the extravascular space, were found to express both MHC-II complex and TNF-α (colocalized images (40 x), scale bar = 25 μm). Blue = DAPI for nuclei. Red = Rhodamine for ED-2/MHC-II. Green = fluorescein for TNF-α.

Discussion

In this study, we monitored CNS-resident macrophages using Feridex preloading and noninvasive MRI after tMCAO. In addition, we verified that these CNS-resident macrophages likely play an integral role in the inflammatory response, using immunohistochemistry and immunofluorescence techniques.

We observed an ~20% signal decrease along the lesion periphery between days 2 and 7 after I–R injury, corresponding to ED-2⁺, PB⁺, MHC-II⁺, and TNF-α⁺ macrophages. These findings, based on Feridex preloading, are in contrast to earlier work using iron oxides after stroke (Farr et al, 2008; Rausch et al, 2002; Schroeter et al, 2006). This difference may largely be attributed to the time frame of macrophage labeling. According to Raynal et al (2004), a 3-day incubation resulted in a 1.1% to 3.0% uptake of Fe-injected dose. Without adequate time for macrophage labeling, injection of iron oxide after I–R injury merely serves as a negative contrast method for the perfusion status and assessment of the blood-brain barrier permeability, and not as a marker of macrophage activity. The choice of 7-day preloading was based on Feridex blood-pool kinetics and liver/spleen clearance rates (Wang et al, 2001) and earlier testing in traumatic brain injury (Foley et al, 2006), ensuring that the MRI signal is from macrophages of systemic origin. Given the difficulty in isolating bone-marrow-derived macrophages, we do not know the precise labeling efficiency in our study. However, we do know that the signal decrease observed on 3D-GRE images is consistent with the 24% (Vande Berg et al, 1999) and 27.8% (Hundt et al, 2000) signal decreases in MRI studies directly labeling the bone marrow. As a further support, the pattern of signal evolution from the subacute to the chronic time frame after tMCAO mirrors the direct quantification of GFP-labeled bone-marrow-derived macrophages in radiation-induced chimeras after stroke (Tanaka et al, 2003).

There was a small decrease in the GRE signal over time (~3% to 10%) in the Control tMCAO group. This decrease could potentially be from endogenous iron deposition during the process of tissue necrosis, but it is more likely that this signal change is from the formation of edema and the concomitant change in T₁ related to increased water content. In this animal model, edema reached a maximum at 2 to 3 days after stroke, with some resolution, but not full resolution, by day 7. To use Feridex preloading or other USPIO preloading to monitor the inflammatory cascade, a correction would need to be applied to account for the presence of vasogenic edema. An alternative approach would be to carry out a separate set of experiments with appropriate controls (i.e., nonpreloaded).

We found quantitative and qualitative correlations between ED-1 and IBA staining (i.e., activated microglia), as well as between ED-2 and PB staining (i.e., iron-labeled, well-differentiated macrophages), validating the signal changes observed on 3D-GRE images. Additional double staining for ED-2:PB and ED-1:PB verified both the presence of iron within ED-2⁺ macrophages and the absence of iron in ED-1⁺ macrophages. These findings are in contrast to earlier work in permanent ischemia, in which there exists an inherent disconnect between the findings on MRI and those on histology/immunohistochemistry (Rausch et al, 2001; Saleh et al, 2004b). We propose that this discordance is a result of two factors. First, iron-oxide administration several hours after permanent ischemia is potentially troublesome. Depending on their size, iron oxides can remain in the blood pool for anywhere from several minutes to several hours. If blood-brain barrier disruption is present, they may extravasate, travel along the Virchow-Robin spaces, and diffuse into the brain (Thorek et al, 2006). This increases the potential for nonspecific accumulation of contrast when smaller iron oxides, such as combidex and mion particles, are being used. Second, ED-1 is a nonspecific marker of macrophage activity (Beelen et al, 1987). ED-1 has been shown to be coincident with OX-42 staining, typically used for the identification of activated microglia (Lassmann et al, 1993). Therefore, it is not surprising that the number of ED-1⁺ cells far exceeded the number of PB⁺ cells in these earlier studies.

To circumvent these problems, we used methods known to accurately differentiate between various populations of macrophages and brain microglia (Beelen et al, 1987; Ito et al, 1998; Lassmann et al, 1993). OX-42 staining is more widely used in the identification of microglia; however, we chose to use IBA, a recently identified marker specific to a 17kDa calcium-binding protein expressed only in activated microglia (Ito et al, 1998). ED-1 is a marker that recognizes the rat homolog of human CD68, a 110 kDa cell-surface glycoprotein readily expressed on the membrane surface of monocytes, macrophages, dendritic cells, and microglia. ED-2 is a marker that recognizes the rat homolog of human CD163, a 175 kDa cell-surface glycoprotein expressed only by resident or well-differentiated macrophages, not by monocytes or microglia. Prussian blue staining, most commonly known as Perl's method, targets the release of ferric iron (Fe³⁺) from binding proteins and the production of an insoluble blue compound. Therefore, if a macrophage contained ferric iron, it is labeled blue.

To reiterate, we found a significant correlation between IBA⁺ and ED-1⁺ cells. IBA⁺ counts were either equivalent to or slightly higher than ED-1⁺ counts, indicating that the majority of ED-1⁺ cells were in fact activated microglia. The number of ED-2⁺ cells in our study was significantly higher than the number of PB⁺ cells; however, the quantitative results were of similar magnitude and were significantly different from both IBA⁺ and ED-1⁺ cell counts. This difference between ED-2⁺ and PB⁺ counts is easily explained by the reported turnover of macrophages every 1 to 2 weeks (Whitelaw, 1966). We preloaded animals with Feridex 7 days before I–R injury, and cell counts were not performed until 7 days after I–R injury. This means that on insult, macrophages will begin to turnover, leading to two populations of macrophages—one that is Feridex labeled and a newly recruited nonlabeled population. We do not know the temporal dynamics of Feridex-labeled versus nonlabeled macrophages, and admittedly, further study is warranted to determine the effect of macrophage turnover on MRI signal variability. Investigation into the kinetics of iron uptake by the bone marrow, spleen, lymph organs, and the CNS after Feridex administration, but before insult, as well as at serial time points after insult, would be required. Regardless of this shortcoming, the correlations between ED-1⁺ and IBA⁺ macrophages, ED-2⁺ and PB⁺ macrophages, and the observed 3D-GRE signal changes confirm our primary hypothesis—Feridex preloading permits labeling of well-differentiated macrophages and monitoring of their activity after tMCAO.

We discovered that these Feridex-labeled macrophages (ED-2⁺ and PB⁺) were confined to three distinct locations, including perivascular regions (PVM), the meninges (MM), and the choroid plexus (CPM). These ED-2⁺ and PB⁺ macrophages were also MHC-II⁺ and TNF-α⁺. Unfortunately, we were unable to draw any conclusions about IL-1. As we used a polyclonal antibody, the IL-1 staining results had an extremely high background, making true IL-1 expression difficult to differentiate from nonspecific IL-1 staining. Despite our lack of conclusive evidence on IL-1 expression, the strategic location of these ED-2⁺ and PB⁺ macrophages, and their dual expression of MHC-II⁺ and TNF-α⁺, leads us to believe that CNS-resident macrophages play an active role in the continuation of the inflammatory response after tMCAO. The best way to confirm their role in reperfusion injury is to perform a selective depletion before stroke induction. These experiments are extremely difficult even in healthy animals, so it would be more challenging to carry them out in hypertensive animals with stroke. In addition, it would be important to investigate these CNS-resident macrophages, their expression of MHC-II⁺, and their interaction with T cells. The MHC-II⁺ expression is not synonymous with, but is required for, T-cell activation. Although this only permits us to state that CNS-resident macrophages expressing MHC-II may mediate T-cell activation and the immune response after tMCAO, earlier work in experimental allergic encephalitis supports this contention (Becher et al, 2006; Greter et al, 2005; McMahon et al, 2005).

Extensive investigation of this specific population of macrophages has been underway since the 1990s in experimental allergic encephalitis and clinical multiple sclerosis. These macrophages actively express MHC-II antigen after administration of interferon-γ and TNF in bone-marrow chimeras (Lassmann et al, 1991). They participate in the pathogenesis of experimental allergic encephalitis through the production and secretion of products, such as proteases (matrix metalloproteinases) and inflammatory cytokines (IL-1, IL-6, and TNF-α), and may interact with leukocytes through intercellular adhesion molecule 1 expression (Bauer et al, 1996). More recently, and probably most importantly, it was shown that their selective depletion results in the suppression of disease (Polfliet et al, 2002). Given that this population of macrophages forms a first line of defense at the interface between the blood and the brain, their manipulation could provide attenuation of reperfusion injury in both the acute and subacute to chronic phases of stroke.

In conclusion, we have shown that preloading with Feridex 7 days before tMCAO permits the selective monitoring of CNS-resident macrophages using 3D-GRE imaging, a unique population of cells that seem to be playing an active rather than a passive role in inflammation and disease progression. Therefore, Feridex preloading provides an attractive avenue for the noninvasive assessment of antiinflammatory therapies in preclinical models of stroke and experimental allergic encephalitis.

Footnotes

Acknowledgements

The authors thank Christina A Rabinak, BS and Narendra C Panapitiya, BS, for their preliminary work with respect to MRI analysis and immunohistochemistry.

The authors declare no conflict of interest.

References

Angelov

Walther

Streppel

Guntinas-Lichius

van Dam

Stennert

Neiss

(1998) ED2-positive perivascular phagocytes produce interleukin-1beta during delayed neuronal loss in the facial nucleus of the rat. J Neurosci Res 54:820–7

Bauer

Ruuls

Huitinga

Dijkstra

(1996) The role of macrophage subpopulations in autoimmune disease of the central nervous system. Histochem J 28:83–97

Becher

Bechmann

Greter

(2006) Antigen presentation in autoimmunity and CNS inflammation: how T lymphocytes recognize the brain. J Mol Med 84:532–43

Beelen

RHJ

Eestermans

Döpp

Dijkstra

(1987) Monoclonal antibodies ED1, ED2 and ED3 against rat macrophages: expression of recognized antigens in different stages of differentiation. Transpl Proc 19:3166–70

Farr

Hoehn

Seehafer

(2008) Imaging macrophage infiltration of ischemic tissue is not possible following transient middle cerebral artery occlusion. ISMRM Proceedings

Flaris

Densmore

Molleston

Hickey

(1993) Characterization of microglia and macrophages in the central nervous system of rats: definition of the differential expression of molecules using standard and novel monoclonal antibodies in normal CNS and in four models of parenchymal reaction. Glia 7:34–40

Foley

Hitchens

Kochanek

Melick

(2006) Visualization of phagocytic cells in a mouse brain after traumatic brain injury with micron-sized iron-oxide particles: a preliminary report. ISMRM Proceedings

Gerriets

Stolz

Walberer

Muller

Rottger

Kluge

Kaps

Fisher

Bachmann

(2004) Complications and pitfalls in rat stroke models for middle cerebral artery occlusion: a comparison between the suture and the macrosphere model using magnetic resonance angiography. Stroke 35:2372–7

Greter

Heppner

Lemos

Odermatt

Goebels

Laufer

Noelle

Becher

(2005) Dendritic cells permit immune invasion of the CNS in an animal model of multiple sclerosis. Nat Med 11:328–34

10.

Hallenbeck

Hansson

Becker

(2005) Immunology of ischemic vascular disease: plaque to attack. Trends Immunol 26:550–6

11.

Hickey

Kimura

(1988) Perivascular microglial cells of the CNS are bone marrow-derived and present antigen in vivo. Science 239:290–2

12.

Hundt

Petsch

Helmberger

Reiser

(2000) Effect of superparamagnetic iron oxide on bone marrow. Eur Radiol 10:1495–500

13.

Ito

Imai

Ohsawa

Nakajima

Fukuuchi

Kohsaka

(1998) Microglia-specific localisation of a novel calcium binding protein, Iba1. Brain Res Mol Brain Res 57:1–9

14.

Kochanek

Hallenbeck

(1992) Polymorphonuclear leukocytes and monocytes/macrophages in the pathogenesis of cerebral ischemia and stroke. Stroke 23:1367–79

15.

Koizumi

Yoshida

Nakazawa

Oonega

(1986) Experimental studies of ischemic brain edema. I. A new experimental model of cerebral embolism in rats in which recirculation can be introduced in the ischemic area, Jpn J Stroke 8:1–8

16.

Lassmann

Schmied

Vass

Hickey

(1993) Bone marrow derived elements and resident microglia in brain inflammation. Glia 7:19–24

17.

Lassmann

Zimprich

Vass

Hickey

(1991) Microglial cells are a component of the perivascular glia limitans. J Neurosci Res 28:236–43

18.

Lehrmann

Kiefer

Christensen

Toyka

Zimmer

Diemer

Hartung

Finsen

(1998) Microglia and macrophages are major sources of locally produced transforming growth factor-beta1 after transient middle cerebral artery occlusion in rats. Glia 24:437–48

19.

Longa

Weinstein

Carlson

Cum mins

(1989) Reversible middle cerebral artery occlusion without craniectomy in rats. Stroke 20:84–91

20.

Manninger

Muldoon

Nesbit

Murillo

Jacobs

Neuwelt

(2005) An exploratory study of ferumoxtran-10 nanoparticles as a blood-brain barrier imaging agent targeting phagocytic cells in CNS inflammatory lesions. AJNR Am J Neuroradiol 26:2290–300

21.

McMahon

Bailey

Castenada

Waldner

Miller

(2005) Epitope spreading initiates in the CNS in two mouse models of multiple sclerosis. Nat Med 11:335–9

22.

McMenamin

(1999) Distribution and phenotype of dendritic cells and resident tissue macrophages in the dura mater, leptomeninges, and choroid plexus of the rat brain as demonstrated in wholemount preparations. J Comp Neurol 405:553–62

23.

Polfliet

van de Veerdonk

Dopp

van Kesteren-Hendrikx

van Rooijen

Dijkstra

van den Berg

(2002) The role of perivascular and meningeal macrophages in experimental allergic encephalomyelitis. J Neuroimmunol 122:1–8

24.

Rausch

Baumann

Neubacher

Rudin

(2002) In-vivo visualization of phagocytotic cells in rat brains after transient ischemia by USPIO. NMR Biomed 15:278–83

25.

Rausch

Sauter

Fröhlich

Neubacher

Radu

Rudin

(2001) Dynamic patterns of USPIO enhancement can be observed in macrophages after ischemic brain damage. Magn Reson Med 46:1018–22

26.

Raynal

Prigent

Peyramaure

Najid

Rebuzzi

Corot

(2004) Macrophage endocytosis of superparamagnetic iron oxide nanoparticles: mechanisms and comparison of ferumoxides and ferumoxtran-10. Invest Radiol 39:56–63

27.

Saleh

Schroeter

Jonkmanns

Hartung

Modder

Jander

(2004a) In vivo MRI of brain inflammation in human ischaemic stroke. Brain 127:1670–7

28.

Saleh

Wiedermann

Schroeter

Jonkmanns

Jander

Hoehn

(2004b) Central nervous system inflammatory response after cerebral infarction as detected by magnetic resonance imaging. NMR Biomed 17:163–9

29.

Schroeter

Oros-Peusquens

Irkens

Celik

Saleh

Shah

Jander

(2006) Spatiotemporal pattern of USPIO enhancement in experimental stroke lesions depends on the type of ischemic injury: a 9.4T MRI study. ISMRM Proceedings

30.

Tanaka

Komine-Kobayashi

Mochizuki

Yamada

Furuya

Migita

Shimada

Mizuno

Urabe

(2003) Migration of enhanced green fluorescent protein expressing bone marrow-derived microglia/macrophage into the mouse brain following permanent focal ischemia. Neuroscience 117:531–9

31.

Thorek

Chen

Czupryna

Tsourkas

(2006) Superparamagnetic iron oxide nanoparticle probes for molecular imaging. Ann Biomed Eng 34:23–38

32.

Vande Berg

Lecouvet

Kanku

Jamart

Van Beers

Maldague

Malghem

(1999) Ferumoxides-enhanced quantitative magnetic resonance imaging of the normal and abnormal bone marrow: preliminary assessment, J Magn Reson Imaging 9:322–8

33.

Wang

Tang

Yenari

(2007) The inflammatory response in stroke. J Neuroimmunol 184:53–68

34.

Wang

Hussain

Krestin

(2001) Superparamagnetic iron oxide contrast agents: physicochemical characteristics and applications in MR imaging. Eur Radiol 11:2319–31

35.

Weissleder

Elizondo

Wittenberg

Rabito

Bengele

Josephson

(1990) Ultrasmail superparamagnetic iron oxide: characterization of a new class of contrast agents for MR imaging. Radiology 175:489–93

36.

Whitelaw

(1966) The intravascular lifespan of monocytes. Blood 28:455–64

37.

Wiart

Davoust

Pialat

Desestret

Moucharrafie

Cho

Mutin

Langlois

Beuf

Honnorat

Nighoghossian

Berthezene

(2007) MRI monitoring of neuroinflammation in mouse focal ischemia. Stroke 38:131–7