Alum administration

The dose of alum-containing vaccine administered to mice was calibrated to mimic the mean number of doses received by MMF patients. One dose of commercially available anti-hepatitis B vaccine contains 0.5 mg Al according to the product data sheet. Based on an average of human body weight of 60 kg (most patients being women), the amount received for each immunization is 8.33 μg/kg. The allometric conversion from human to mouse (FDA Guidance 5541) gives a final amount of approximately 100 μg/kg. A dose of 36 μL vaccine, which corresponds to 18 μg Al, was injected to mimic the cumulative effect induced by 5.2 human doses to 35 g mice (the mean weight at the d180 midtime of brain analysis). This dose represents an equivalent 6.8 human doses in the youngest animal (27 g body weight, 11 weeks of age at sacrifice) and 4.3 in the oldest one (42 g at 62 weeks).

Furnace atomic absorption spectrometry

Al concentrations were determined in whole tibialis anterior (TA) muscles and brains dried at 37°C and digested with concentrated HNO₃ (14 mol/L). Digests were allowed to cool before dilution to 10% HNO₃ with ultra-pure water. The total aluminum in each digest was measured by transversely heated graphite atomizer graphite furnace atomic absorption spectrometry (TH GFAAS) and results were expressed as Al mg/g tissue dry weight.

PIXE

As in normal conditions Al may be detected with marked interindividual variations in tissues, de novo incorporation of aluminum in too low doses does not cause easily detectable changes when global conventional approaches are used [10]. Here we used particle induced X-ray emission (PIXE), a procedure analyzing radiation emitted from the interaction of a proton beam with the matter [19], to detect areas enclosing small Al spots. Sections (20 μm-thick) carefully protected from environmental Al were mounted on fresh formvar films, kept in the cryostat for 6 hours and stored under Al-free silica gel. Mineral and metal ions were detected using the nuclear microprobe of the Centre d’Etudes Nucléaires de Bordeaux-Gradignan. A 1 MeV proton beam focused down to a 2 μm spot was randomly scanned over multiple 500 × 500 μm fields of tissue sections. In the case of an Al signal, a re-test of 100 × 100 μm areas of interest was performed. PIXE and Rutherford backscattering spectrometry analyses were employed simultaneously and quantitative results were computed, as previously described [19]. Al spots were considered eligible on three criteria: a size of more than 3 pixels (that is, above the background noise), a depot not colocalized with Si, and a depot surrounded by a rounded halo of decreased intensity (both characteristics limiting confusion with contamination by external dust overcoming the protection procedures).

Synthesis of Al-Rho particles

Gadolinium oxide nanohybrids with Al(OH)₃ coating were obtained in three steps: (i) gadolinium oxide nanoparticles were first synthesized; (ii) polysiloxane shell growth was then induced by hydrolysis-condensation of convenient silane precursors in the presence of the nanoparticles; and (iii) the nanohybrids were coated by the addition of aluminum nitrate and soda in stoichiometric conditions.

Peripheral injections of fluorescent nanomaterials

Two types of fluorescent nanomaterials were used: exploratory polychromatic fluorescent latex beads (FLBs) (500 nm fluorospheres, Polysciences, Warrington, PA, USA) and confirmatory Al-Rho nanohybrids constructed with a rhodamine containing core and an Al(OH)₃ shell. FLBs were used first because they offer several characteristics that facilitate their detection in tissues, including strong fluorescence, spheric appearance and homogeneous size. This allowed us to get a clear picture of what was happening in terms of biodistribution of these avidly phagocytosed particles. Al Rho particles were less fluorescent and more heterogeneous in shape and size than FLBs but better represented alum adjuvant surrogates. Almost all biodistribution experiments performed with FLBs in wild type mice were also done with Al-Rho. In contrast, FLBs and Al Rho were differentially used in mutated/genetically-modified mice: FLBs were preferred to study particle biodistribution in mdx mice with BBB alterations and when the GFP marker was used (that is, CX3CR1^GFP/+ mice with fluorescent microglia, GFP + BMT studies); Al-Rho particles were preferred in gain/loss of CCL2/MCP-1 function studies designed on the basis of preliminary results on the CCL2 status of alum-intolerant humans.

FLB suspension diluted at 1:1 in PBS contained 1.8 × 10¹¹ particles per mL. A total volume of 40 μL (20 μL in each TA muscle) was injected, corresponding to a total amount of 7.2 × 10⁹ particles. The same volume of Al-Rho suspension was injected in TA muscles. PBS-injected mice were used as controls. Tissues, including popliteal and inguinal DLNs, spleen, brain and blood, were collected at various time points post injection. Three mice were included per group at each time point for both injected materials and their controls. Other administration routes were compared to the standard i.m. injection, including s.c. injection of 20 μL FLBs in each hindlimb, and i.v. injection of 40 μL FLBs in the tail vein.

Stereotactic cerebral injections

Mice were anaesthetized with ketamine and xylazine. Al-Rho suspension (0.5 μL) was stereotactically injected in the striatum using a 1 μL Hamilton syringe. Biodistribution of i.c. injected Al-Rho to cervical DLNs, assessed by serial sectioning of the whole cervical region, and spleen, was compared to biodistribution to the popliteal DLN and spleen of the same amount of Al-Rho injected in the TA muscle.

Pharmacological and physical migration blockade

The prostaglandin analog BW245C, an agonist of the PGD2 receptor, was used to inhibit APC migration as previously reported [20]. Since BW245C is active for two days after injection, BW245C (100 nM, Cat.no.12050, Cayman Chemical, Ann Arbor, MI, USA) was injected twice in the TA muscle: it was first co-injected with FLBs at d0 and a second time alone at d2, and DLNs were removed for examination at d4. Untreated FLB-injected mice were used as controls. In another set of experiments DLNs were surgically ablated and mice immediately injected with FLBs in the TA muscle.

Loss and gain of CCL2 function experiments

Exploratory analyses performed in MMF patients with ASIA [see Additional file 1: supplementary information section] yielded a CCL2 signal in the form of: (1) a selective increase of CCL2 in the serum of MMF patients compared to healthy controls; and (2) a given haplotype in the CCL2 gene tending to be more frequent in MMF patients than in the general population. These results led us to use mouse models to explore the role of CCL2 in the biodisposition of particulate materials. Loss of CCL2 function studies were done using CCL2^−/− mice injected i.m. with 40 μL Al-Rho. Gain of CCL2 function experiments consisted first of i.m. co-injection of 10 μL murine rCCL2 (100 μg/ml; R&D, Minneapolis, MN, USA) with 40 μL Al-Rho. DLNs were removed at d4, spleen, brain and blood at d21. In other experiments murine rCCL2 was infused into the brain through a catheter stereotactically inserted into the striatum at d7 post-Al-Rho, fed by a subcutaneously implanted osmotic micropump fixed into the neck (0.25 μL/hour Alzet brain infusion kit, Charles River, L’Arbresle, France). rCCL2 was infused for 14 days (diffusion rate 180 pg/day), with or without rCCL2 i.m. injection concurrent with Al-Rho injection. At d21 post Al-Rho injection, animals were sacrificed, and blood and tissues were collected. For controls, osmotic pumps filled with PBS were used.

Tissue preparation and particle counting

Mice under terminal anesthesia were transcardially perfused with PBS followed by ice-cold 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer. Tissues and organs were removed, post-fixed in PFA for 4 hours at 4°C, immersed overnight at 4°C in a 30% sucrose solution, and quickly frozen. Whole brains were serially cut into coronal cryosections of 40 μm, spleen and muscle into 20 μm, and DLNs into 10 μm, and stored at −20°C until particle counting or treatment. Brain sections were successively deposited on 10 different Superfrost® slides in order to obtain 10 identical series, thus allowing determination of total particle content by multiplying by 10 the number of particles found in one series. A similar approach was used for DLNs and spleen. Blood was collected by heart puncture and 100 μL were smeared for particle counting.

Immunohistochemistry and Morin staining

Immunostaining was done using commercial primary antibodies routinely used in the lab, raised against CD11b (1/200, AbD Serotec, Oxford, UK), F4/80 (1/50, AbCam, Cambridge, UK), GFAP (1/200, DakoCytomation, Trappes, France), vimentin (1/500 DakoCytomation), collagen IV (1/100 Millipore, Temecula, CA, USA), NG2 (1/200, Millipore, Molsheim, France), MAP2 (1/100, Sigma-Aldrich, Lyon, France), and IL1β (1/100, AbCam, Paris, France) or nonspecific mouse IgG (Jackson ImmunoResearch, Suffolk, UK). Then, biotinylated anti-rat and anti-rabbit antibodies (1/200, Vector Laboratories, Paris, France) were used accordingly and were revealed using Alexa fluor 488-conjugated streptavidin (1/200 Invitrogen, Cergy-Pontoise, France). Neuron labeling was done using NeuroTrace® blue fluorescent Nissl Stain according to the manufacturer” instructions (Invitrogen). Al was stained with Morin (M4008-2 G, Sigma-Aldrich) used as 0.2 g dissolved in a solution consisting of 0.5% acetic acid in 85% ethanol [21]. Formation of a fluorescent complex with Al was detected under a 420 nm excitation wavelength as an intense green fluorescence with a characteristic 520 nm emission. Notably, nanohybrids (Gd₂O₃) core encapsulated by polysiloxane shell were not positively stained by Morin. In contrast, when coated with Al(OH)₃, these particles were strongly positive for Morin. Fluorescence microscopy and spectral analyses were done using Carl Zeiss light and confocal microscopes.

Cell isolation from blood and tissues and flow cytometry

For blood cell immunophenotyping, 100 μL blood was treated with ethylenediaminetetraacetic acid (EDTA) and stained with fluorescein isothiocyanate (FITC)-conjugated antibodies. Erythrocytes were lysed using hypotonic lysis solution, and then cells were washed with (D)MEM and sorted using a MoFlo cell sorter (Beckman Coulter, Villepinte, France). Cells were extracted from tissues of exsanguinated mice perfused with PBS. Tissues were removed and freshly dissociated in (D)MEM. DLNs and spleen were dissociated in (D)MEM containing 0.2% collagenase-B (Roche Diagnostics, Meylan, France) and 0.2% trypsine-EDTA at 37°C for 45 minutes twice. Brain tissue was dissociated in 1% Trypsin-HBSS (Thermo Scientific HyClone, South Logan, UK) containing 100 U/mL DNase (Roche Diagnostics). Cell suspensions were filtered and counted. CD45⁺ or CD11b⁺ cells were isolated using magnetic cell sorting (MACS, Miltenyi Biotec, Paris, France) and stained with one of the following antibodies and their isotypes: FITC-conjugated anti-CD11b, FITC–conjugated anti–Ly-6C (GR1), FITC–conjugated anti-CD11c (BD-Pharmingen Bioscience, San Diego, CA, USA). Cells were sorted using a cell sorter. Populations presenting >90% purity were used. Sorted cells were cytospinned and stained with Hoechst-33342 for nucleus. Particle loaded cells were counted under a fluorescence microscope.

Bone marrow transplantation experiments

GFP⁺ bone marrow (BM) cells were obtained by flushing the femurs of adult CAG-GFP mice and were injected retroorbitally (1 × 10⁷ cells per mouse) to four-week-old C57BL/6 mice, as previously described [15]. Recipient mice were irradiated at 9.0 Gy on d1 before transplantation, and were treated with 10 mg/kg/day ciprofloxacin for 10 days. Blood chimerism of >90% was controlled at three to four weeks post-transplantation.

Fluorospheres injected into mouse muscle undergo lymphatic and systemic biodistribution

To examine if particles translocate to distant sites, we next injected polychromatic FLBs. A size of 500 nm was chosen as an approximation of the average size of alum agglomerates observed in vivo, allowing FLB visualization as individual spheres by confocal and fluorescence microscopes (resolution >200 nm). After i.m. injection of 20 μL suspensions, FLBs transiently peaked in free form in blood (1,200 + 400 FLBs per 100 μL) at hour 1. As early as 1 hour post-injection, some FLBs had also reached DLNs. I.m. injection of GFP⁺CD45⁺ cells, either pre-loaded with FLBs or coinjected with FLBs, showed no GFP⁺ cells translocation to DLNs at hour 1 (data not shown), indicating early cell-independent particle translocation to DLNs by lymphatic drainage of the muscle interstitial fluid [24]. In DLNs, however, most FLBs were cell-associated suggesting rapid capture by DLN resident cells. Within 24 hours, FLBs were phagocytosed by muscle CD11b⁺ MO/MPs. Phagocytes progressively cleared the particles away from the interstitium to form collections (Figure ​(Figure2a),2a), mainly located in muscle envelopes at d4.

At d4, FLBs had dramatically increased in DLNs, forming intracellular agglomerates in the interfollicular area (Figure ​(Figure2b-e).2b-e). Particle-loaded cells extracted from DLNs at d4 were CD45⁺, CD11b⁺, and more often GR1⁺/Ly6C⁺ (69% to 81%), and CD11c⁺, with either intermediate (46%) or high (22%) intensity (Figure ​(Figure2a,c,d),2a,c,d), thus corresponding to MO-derived inflammatory DCs and MPs [25]. Co-injection of FLBs with the synthetic prostaglandin analog BW245C, a compound known to inhibit DC migration [20], inhibited FLB translocation to DLNs at d4 by 32% in the popliteal and 69% in the inguinal DLNs, respectively (Figure ​(Figure2f).2f). This indicated prominent particle transport within phagocytic cells, at least downstream to popliteal DLN. At later time points, both the number of particle-loaded cells and the individual cell load markedly decreased in DLNs (Figure ​(Figure2e).2e). While decreasing in DLNs, FLBs dramatically increased in spleen from d4 to d21 (Figure ​(Figure3a,b).3a,b). As spleen is unplugged to lymphatic vessels, the particle transfer from DLNs to spleen implicated exit from the lymphatic system through the thoracic duct and circulation in the blood stream. Consistently, smears showed a similar d21 peak of FLB-loaded CD11b⁺cells in the circulation (Figure ​(Figure3c,d).3c,d). From d4, circulating FLBs were cell-associated (Figure ​(Figure3d).3d). Most FLB-loaded cells in blood, DLNs and spleen exhibited a few particles and were GR1⁺/Ly6C⁺ (Figure ​(Figure3e,f).3e,f). However, 22% to 33% were GR1^–/Ly6C^– in spleen and had frequently incorporated >5FLBs, suggesting phagocytosis-associated maturation of inflammatory MO-derived cells [20,25,26]. FLB-loaded cells had markedly decreased in spleen at d90. Although declining after d21, FLB-loaded cells were still detected in blood at d45 and d90.

Fluorosphere incorporation into brain is delayed and depends on prior cell loading in peripheral and lymphoid tissues

Particles were detected in brain mainly from d21 post-injection. After d21 post-i.m. injection, FLBs gradually increased in brain until the d90 endpoint in the C57Bl6 mouse (Figure ​(Figure4a,b)4a,b) and until the d180 endpoint in the CX3CR1^GFP/+ mouse conventionally used to study resident microglia (Figures  4a and ​and5a).5a). FLBs were predominantly found in the grey matter (82% to 95%), regardless of the amount of injected FLBs (4, 10, 20 μL), vaccine co-injection (36 μL), or post-injection time from d21 to d365. Some FLBs were detected in leptomeninges (9%) and in the white matter (9%) at d21, but these locations became rare at later time points. FLBs were <5% in choroid plexus (Table  1). Comparative FLB distribution at month 3, month 6 and month 12 showed no prominent accumulations of particles at any neuroanatomical location (Figure ​(Figure4c).4c). FLBs were usually detected in brain as single particles located within or at the surface of cells; 37% to 62% of particles could be reliably assigned to a given cell subset by immunohistochemical screening. At d21, particles were mainly associated with perivascular CD11b⁺ MPs, but at d90 they were also found in deep ramified CX3CR1⁺microglia (Figure ​(Figure5a).5a). Particles were also detected in GFAP⁺ astrocytes, MAP2⁺ or Neurotrace-stained neurons, and vimentin⁺ leptomeningeal cells (Figure ​(Figure5b-e),5b-e), and in NG2⁺ oligodendroglial progenitors/pericytes (not shown). FLB incorporation into GFP⁺ resident ramified microglia of CX3CR1^GFP/+ mice increased by up to 26-fold the d21 value at d180.

Importantly, compared to i.m. injection, the same FLB amount injected in the tail vein resulted in virtually no cerebral entry at d21 and d90 in C57Bl6 mice (Figure ​(Figure6a).6a). Moreover, ablation of popliteal and inguinal DLNs before FLB injection in TA muscle resulted in 60% to 80% reduction of FLB incorporation into blood, spleen and brain compartments at d21 (Figure ​(Figure6b).6b). Thus, cell uptake in muscle and DLNs, and subsequent cell traffic to blood crucially contributed to delayed particle translocation to spleen and brain (Figure ​(Figure6a-f).6a-f). Consistently, by injecting FLBs into the muscle of GFP⁺BM chimeric mice obtained by transplanting GFP⁺BM-derived cells to irradiated syngenic C56Bl6 mice [15], we detected FLB-loaded GFP⁺cells in these organs (Figure ​(Figure7a,b,c)7a,b,c) and observed delayed incorporation of donor-derived cells in brain (Figure ​(Figure77d,e).

This BM transplantation model is known to be associated with irradiation-induced BBB alteration. Dystrophin-deficient mdx mice also have chronically altered BBB [27]. As a corollary, compared to age-matched controls, they show significantly more CD31⁺ brain capillaries, and a dramatic increase of perivascular CD11b⁺ macrophages (Figure ​(Figure6c)6c) at the expense of deep ramified microglia. FLB injection in mdx mouse muscle resulted in increased brain incorporation of particles at both d21 and d90, as assessed by both histology and cytospins of CD45⁺/CD11b⁺cells extracted from brain (Figure ​(Figure6d,e,f).6d,e,f). Thus, BBB alteration and/or the associated inflammatory/angiogenic response likely favors brain incorporation of circulating particle-loaded cells.

Fluorescent nanohybrids coated with Al(OH)₃ undergo CCL2-dependent systemic scattering and brain penetration

For confirmatory experiments we constructed fluorescent particles mimicking alum. Rhodamine nanohybrids [28] were covalently coated with a Al(OH)₃ shell. As assessed by the Morin stain for alumimum, these Al-Rho particles were avidly phagocytosed after i.m. injection and formed intracellular agglomerates similar in size to the vaccine adjuvant (Figure ​(Figure8a,b).8a,b). Biodistribution of the alum fluorescent surrogate injected into TA muscle was strikingly similar to that of FLB (Table  2), including d4 peak in DLNs, d21 peak in spleen, delayed entry in brain, and main association with GR1⁺/Ly6C⁺ MOs in tissues (Figure ​(Figure8c-h).8c-h). Compared to i.m. injection, s.c. injection of Al-Rho particles was associated with an even higher rate of diffusion to DLNs (Figure ​(Figure8f),8f), a finding consistent with the presence of abundant migratory DCs in skin.

On the grounds of a human SNP study, we performed CCL2 gain and loss of function experiments to investigate the role of CCL2-responsive cells in particle scattering and neurodelivery. Injection of Al-Rho particles into the TA muscle of CCL2-deficient mice decreased particle incorporation by 35% into popliteal DLN and by 76% in inguinal DLN at d4, and by 71%, 85% and 82% in spleen, blood, and brain, respectively, at d21 (Figure ​(Figure9a).9a). Conversely, Al-Rho particle biodistribution increased in different gain of CCL2 function experiments (Figure ​(Figure9b-d).9b-d). I.m. co-injection of Al-Rho with murine recombinant CCL2 (rCCL2: 1 μg) increased particle incorporation by 47% into popliteal and 163% into inguinal DLN (d4), and by 180% in spleen, 274% in blood, and 341% in brain (d21).

Morever, slow intracerebral (i.c.) infusion of CCL2 by an osmotic pump (180 pg/day during 15 days starting at d7 after Al-Rho i.m. injection) increased particle incorporation into brain by 74% at d21 compared to PBS control. The combination of i.m. injection and i.c. infusion of rCCL2 increased particle incorporation into the brain by 539%. Despite important interindividual variations, a consistent trend of CCL2-dependent increase of Al brain levels was detected 21 days after i.m. injection of 40 μL of alum-containing vaccine (Figure ​(Figure9e).9e). Taken together these results indicate that after i.m. injection, particles associated with inflammatory MOs can enter the brain using a CCL2-dependent mechanism, possibly through a Trojan horse mechanism. Importantly, Al-Rho particles gaining access to the brain after i.m. injection remained intact since they were still coated with Al(OH)₃ as assessed by both Morin stain (Figure ​(Figure10a),10a), and PIXE (Figure ​(Figure10b).10b). Their incorporation in neural cells was consistently associated with the expression of IL-1β (Figure ​(Figure10c),10c), a reliable marker of particle-induced NALP3 inflammasome activation [29].