Wireless magnetothermal deep brain stimulation

See allHide authors and affiliations

Science  27 Mar 2015:
Vol. 347, Issue 6229, pp. 1477-1480
DOI: 10.1126/science.1261821

Exciting nerve cells deep inside the brain

Current techniques to stimulate regions inside the brain need a permanently implanted wire or an optical fiber. Working in mice, Chen et al. developed a method to overcome this problem (see the Perspective by Temel and Jahanshahi). They introduced heat-sensitive capsaicin receptors into nerve cells and then injected magnetic nanoparticles into specific brain regions. The nanoparticles could be heated by external alternating magnetic fields, which activated the ion channel–expressing neurons. Thus, cellular signaling deep inside the brain can be controlled remotely without permanent implants.

Science, this issue p. 1477; see also p. 1418


Wireless deep brain stimulation of well-defined neuronal populations could facilitate the study of intact brain circuits and the treatment of neurological disorders. Here, we demonstrate minimally invasive and remote neural excitation through the activation of the heat-sensitive capsaicin receptor TRPV1 by magnetic nanoparticles. When exposed to alternating magnetic fields, the nanoparticles dissipate heat generated by hysteresis, triggering widespread and reversible firing of TRPV1+ neurons. Wireless magnetothermal stimulation in the ventral tegmental area of mice evoked excitation in subpopulations of neurons in the targeted brain region and in structures receiving excitatory projections. The nanoparticles persisted in the brain for over a month, allowing for chronic stimulation without the need for implants and connectors.

Stimulation of deep brain structures affected by treatment-resistant psychiatric and neurological disorders can ameliorate associated symptoms but is currently only achieved by permanently implanted electrodes (1). Second-generation neuromodulation technologies rely on acoustic (2), electromagnetic induction (3), or optical (4) signals. These fields are largely absorbed and scattered by tissue and similarly require a conduit for deep brain stimulation. In contrast, low-radiofrequency alternating magnetic fields (100 kHz to 1 MHz) can penetrate into the body without substantial attenuation and thus enable signal delivery into deep brain regions (5). Alternating magnetic fields can be converted into biological stimuli by magnetic nanoparticles (MNPs) that dissipate heat via hysteretic power loss (6). Although MNP heating has been investigated as a cell-destructive therapy in magnetic hyperthermia for 50 years (7), this effect has only recently been exploited for control of cell membrane depolarization and gene expression in engineered xenografts and invertebrates (8, 9). Magnetothermal control of neural activity in vivo in a mammalian system remains to be demonstrated.

To achieve reversible neuronal activation with alternating magnetic fields, we developed an intracellular calcium control scheme by sensitizing cells to heat generated from MNPs (Fig. 1A). Earlier experiments relied on synthetic transgenes to target MNPs to the cell membrane and required tens to thousands of seconds to observe increased calcium ion (Ca2+) influx, which exceeded temporal dynamics of neuronal firing by orders of magnitude. Recent studies suggest that MNP functionalization with proteins induces cell internalization and the formation of protein coronas that may reduce the effectiveness of targeting and heat dissipation in vivo (10, 11). We reasoned that untargeted Fe3O4 MNPs optimized for efficient heat dissipation at clinically relevant alternating magnetic field conditions can (i) reduce the latency period for neural excitation, (ii) eliminate exogenous targeting transgenes, and (iii) have chronic utility in vivo because MNPs exhibit minimal cytotoxicity and remain intact several months after injection (12, 13). Spherical Fe3O4 MNPs 22 nm in diameter possess some of the highest heating rates per gram, or specific loss power, measured for a synthetic material at a therapeutically relevant frequency ƒ = 500 kHz and field amplitude Ho = 15 kA/m (14). We prepared these monodisperse MNPs via the thermal decomposition of an environmentally benign iron-oleate precursor (15) and dispersed them in water through high-temperature ligand exchange with poly(acrylic acid) (PAA) (Fig. 1B) (14). Grafting poly(ethylene glycol) (PEG) chains onto PAA-coated MNPs resulted in their steric dispersion, which improved colloidal stability (Fig. 1, C and D) and biocompatibility, as indicated by the increased viability of human embryonic kidney (HEK) 293FT cells over prolonged exposure (fig. S1) (16). These MNPs exhibited specific loss of power of 660 ± 50 W/g, which is sixfold greater than that of hyperthermia agents currently used in clinical settings (fig. S2). Magnetic fields were generated by a resonant coil custom designed for fluorescence imaging during stimulation (fig. S3, A to E). Although transient receptor potential cation channel subfamily V member 1 (TRPV1) is naturally expressed across the mammalian nervous system (17), we designed a transgene to establish sustained and uniform levels of TRPV1 expression for magnetothermal membrane depolarization across different cell lines (18). The TRPV1 transgene was placed under the excitatory neuronal promoter calmodulin kinase II α-subunit along with mCherry separated from TRPV1 by the posttranscriptional cleavage linker p2A (CamKIIα::TRPV1-p2A-mCherry) (19) and packed into the lentiviral vector so as to enable long-term in vitro and in vivo neural transfection (20). Cells were additionally transfected with the adeno-associated virus serotype 9 (AAV9) carrying GCaMP6s under the neuronal promoter human synapsin (hSyn::GCaMP6s) for measurement of intracellular Ca2+ changes as a proxy for membrane depolarization (21). Functionality of the two genes was confirmed by observing increased fluorescence intensity in response to capsaicin, a TRPV1 agonist, and temperature increase above 43°C in nonexcitable HEK293FT cells (fig. S4, A to C).

Fig. 1 Wireless ON switch for controlled magnetothermal membrane depolarization of TRPV1+ cells.

(A) Experimental scheme. Magnetic field stimulation (“Field ON”) of TRPV1 from MNP heating is visualized by gCaMP6s fluorescence changes. (B and C) Transmission electron micrographs of MNPs: (B) as-synthesized and (C) after surface modification with a 2-nm PEG shell. (D) Size distribution plot for PAA- and PEG-coated MNPs observed by dynamic light scattering. Aggregation in physiological fluids is observed for PAA-coated MNPs but not for PEG-coated MNPs. (E) Color maps of fluorescence intensity changes for TRPV1 and TRPV1+ HEK293FT cells before and during magnetic field stimulus. Scale bar, 50 μm. (F) Normalized fluorescence intensity change (ΔF/F0) as a function of time (solid lines indicate the mean, and shaded gray areas indicate standard error). Dashed line corresponds to the crossing of TRPV1 activation threshold temperature. Fluorescence increase was observed only in TRPV1+ cells upon magnetic field application. (Inset) Temperature profile without (gray) and with (red) magnetic field application. In all experiments, field amplitude is Ho = 15 kA/m, and frequency is ƒ = 500 kHz.

We first demonstrated magnetothermal control of intracellular Ca2+ influx in HEK293FT cells. Fluorescence intensity maps indicated that only cells expressing TRPV1 (TRPV1+) responded to the field stimulus (ƒ = 500kHz, Ho = 15 kA/m) when incubated in MNP solutions (2 mg/mL), whereas cells not expressing TRPV1 (TRPV1) as well as TRPV1+ and TRPV1 cells without field stimulus did not exhibit changes in intracellular Ca2+ concentration (Fig. 1E). A field-induced temperature increase in excess of 43°C in MNP solutions triggered a GCaMP6s fluorescence increase of ΔF/F0 > 50% in 36.1 ± 4.3% (mean ± SD) of TRPV1+ cells, whereas only 1.7 ± 1.6% (mean ± SD) of TRPV1 cells exhibited a similar response (Fig. 1F and fig. S5, A to D).

Magnetothermal membrane depolarization was sufficient to evoke trains of action potentials in primary hippocampal neurons expressing TRPV1 when exposed to 10-s field pulses at 60-s intervals. Viral transfection with AAV9-hSyn::GCaMP6s, which allows for fluorescence detection of single action potential events (21), and Lenti-CamKIIα::TRPV1-p2A-mCherry (TRPV1+) or Lenti-CamKIIα::mCherry (TRPV1) yielded a coexpression efficiency of 57% after 5 days (Fig. 2A). In MNP solutions (10 mg/ml), 85 ± 14% of TRPV1+ neurons exhibited synchronized firing within 5 s after stimulus, whereas only sporadic activity was observed in TRPV1 neurons (Fig. 2, B to H). This implies that the temperature increase (Fig. 2D) in MNP solutions exposed to alternating magnetic field was sufficient to trigger TRPV1 (Fig. 2H) while avoiding nonspecific thermal effects such as changes in membrane capacitance (Fig. 2F) (22). In the absence of MNPs, magnetic field did not induce appreciable solution heating (Fig. 2C), and no correlated response was observed in TRPV1+ or TRPV1 neurons (Fig. 2, B, E, and G). We recorded neural activity from GCaMP6s temporal fluorescence traces (fig. S6, A to D, and movie S1) (23). Waves of Ca2+ spikes were repeatedly induced by field pulses only in TRPV1+ neurons in the presence of MNPs (Fig. 2, I to P). The observed 5-s latency between the field application and the onset of neural activity is fivefold faster than previously described (8).

Fig. 2 Alternating magnetic field stimulus evokes correlated and repeated trains of action potentials.

(A) Confocal fluorescent images of cotransfected hippocampal neurons. Scale bar, 25 μm. (B) Population study of 100 neurons from three trials counting the number of neurons that spike within a 5-s bin after magnetic field stimulus. (C and D) Temperature profiles during magnetic field application in Tyrode’s solution (C) without and (D) with MNPs. Shaded area is the SD with average value overlaid (black). (E to H) Example fluorescence traces of 10 individual neurons with average overlaid (black). (I to L) Raster plots of 100 randomly selected neurons from three trials. Calcium spikes were counted according to an automated algorithm. (M to P) Peristimulus time histograms of the raster plots binned at 2 s. Color scheme for (E) to (P): TRPV1 neurons in Tyrode’s solution without MNPs, gray; TRPV1 neurons in Tyrode’s solution with MNPs, red; TRPV1+ neurons in Tyrode’s solution without MNPs, blue; TRPV1+ neurons in Tyrode’s with MNPs, orange. Shaded blue bars represent alternating magnetic field pulses (Ho = 15 kA/m, ƒ = 500 kHz).

We next tested whether alternating magnetic field could activate a subpopulation of neurons in deep brain tissue in mice. We used finite element modeling corroborated with temperature recordings in brain phantoms to predict local temperature changes in response to field stimulus (fig. S7). Injections (2.5 μL) of MNP solution (100 mg/mL) delivered temperature gradients sufficient to reach the TRPV1 activation threshold within 5 s and cool back to 37°C over 60-s cycles (fig. S7, B to F), thus avoiding prolonged exposure to noxious heat (fig. S7G) (24).

With low endogenous expression of TRPV1 (25) and well-characterized projections (26), the ventral tegmental area (VTA) was an attractive deep brain target for initial demonstration of magnetothermal stimulation. Furthermore, phasic excitation in the VTA has therapeutic implications in the treatment of major depression (27). We sensitized excitatory neurons in the VTA to heat through the lentiviral delivery of TRPV1, which was followed by MNP injection into the same region 4 weeks later (Fig. 3, A and B, and fig. S8A). The anesthetized mice were exposed to the magnetic field conditions described above (fig. S8, B and C). Neuronal excitation was quantified by the extent of activity-dependent expression of the immediate early gene c-fos within a 250-μm vicinity of the MNP injection (Fig. 3, C to F) (28). Neural activity was only triggered by magnetic field in the VTA of mice transfected with TRPV1 in the presence of MNPs, resulting in a significantly higher proportion of c-fos–positive (c-fos+) cells, as revealed by a two-way analysis of variance (ANOVA) with a Bonferroni post hoc test (F1,13=47.5, P < 0.0001) (Fig. 3G). Control subjects testing whether the MNP injection, heat dissipation with field stimulus, or TRPV1 expression alone can result in neural stimulation showed no significant c-fos expression (Fig. 3, C to E and G). Furthermore, the spatial extent of neuronal activation was largely collocated with TRPV1 expression in the VTA (Fig. 3, H and I).

Fig. 3 Wireless magnetothermal stimulation in vivo.

(A) In vivo experimental scheme. (B) Confocal image of a coronal slice representative of the TRPV1-p2A-mCherry expression profile in the VTA. (C to F) 4′,6-diamidino-2-phenylindole (DAPI) (blue), mCherry (red), and c-fos (green) and overlay confocal images of regions used for quantification of neural stimulation. Scale bar, 25 μm. All animals were injected with MNPs. Experimental conditions were (C) without (OFF) and (D) with (ON) magnetic field stimulation in TRPV1 VTA, and (E) OFF and (F) ON stimulation in TRPV1+ VTA. (G) Percentage of mCherry-positive and c-fos–positive neurons within cell population indicated by DAPI corresponding to the four conditions presented in (C) to (F). Significance is confirmed by two-way ANOVA with Bonferroni post hoc test (n = 4 mice, F1,13 = 47.5, P < 0.0001). (H and I) Confocal images of the VTA after acute magnetothermal stimulation. c-fos expression is largely confined to the VTA in regions where TRPV1 is expressed. Scale bar, 100 μm. (J to L) Confocal images of the (J) VTA, (K) mPFC, and (L) NAc 1 month after MNP injection without (OFF) and with (ON) field treatment. Scale bar, 100 μm. (M) Percentage of c-fos+ neurons in the VTA among DAPI-labeled cells with and without magnetic field stimulation. Increased c-fos expression is observed after field treatment (ON) as compared with unstimulated (OFF) controls (n = 3 mice OFF/ON; Student’s t test, P < 0.02). (N and O) Similarly, up-regulation is observed in (N) the mPFC and (O) in the NAc with alternating magnetic field (ON) as compared with the same regions without (OFF) the field stimulus (n = 3 mice OFF/ON; Student’s t test *P< 0.02, **P < 0.002).

We next investigated whether neurons in the VTA can be activated 1 month after MNP injection so as to explore its chronic utility (Fig. 3, J to O). We again observed increased c-fos expression in the VTA only in mice transfected with TRPV1 in the presence of MNPs and exposed to the magnetic field protocol described above (Fig. 3, J and M, “ON”) (Student’s t test, P < 0.02). In these mice, we also found evidence of field-evoked up-regulation of c-fos in the medial prefrontal cortex (mPFC) (Fig. 3, K and N, “ON”) (Student’s t test, P < 0.02) and nucleus accumbens (NAc) (Fig. 3, L and O, “ON”) (Student’s t test, P < 0.002), which are known to receive excitatory inputs from VTA neurons (26, 29). In the absence of stimulation, neurons in the VTA near the MNP injection site and the neurons in the mPFC and NAc did not exhibit increased c-fos expression (Fig. 3, J to O, “OFF”).

We compared the biocompatibility of the MNP injection with a similarly sized stainless steel implant (fig. S9). The interface between the MNP injection and the tissue exhibited significantly lower glial activation and macrophage accumulation and higher proportion of neurons, as compared with that of the steel implant 1 week and 1 month after surgery (fig. S9, A to F). The improved tissue compatibility can likely be attributed to the mechanically pliable nature of the MNP injection and sequestration via endocytosis (12, 13). No difference in neuronal or glial density was observed between brain tissue of stimulated and unstimulated mice, suggesting that the rapidly dissipated magnetothermal cycles cause minimal thermal damage to the surrounding tissue (fig. S9G).

We demonstrated widespread and repeatable control of cellular signaling in nonexcitable and electroactive cells using wireless magnetothermal stimulation in vitro and in vivo. Finer control over stimulation intensity to facilitate applications of this approach to problems in systems neuroscience can be achieved by further reducing the latency between field onset and evoked neural firing by developing MNPs with high specific loss powers (30) and by introducing heat-sensitive ion channels with lower thermal thresholds (31). Mechanosensitive potassium and chloride channels may serve as potential mediators of magnetothermal inhibition (32). Although demonstrated for chronic stimulation of targeted neural circuits, this magnetothermal paradigm may be formulated to trigger thermosensitive ion channels endogenously expressed in the peripheral nervous system (17), enabling wireless control in deep tissue regions that currently pose substantial challenges to bioelectronic medicines (33).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S9

Tables S1 and S2

References (3442)

Movie S1

References and Notes

  1. Acknowledgments: We thank K. Deisseroth, D. Julius, and F. Zheng for generous gifts of plasmids and cell lines; C. Ramakrishnan for molecular biology advice; the GENIE project and Howard Hughes Medical Institute Janelia Farm for AAV9-hSyn::GCaMP6s supplied by the University of Pennsylvania vector core; and D. Irvine and A. Jasanoff for their thoughtful comments on our manuscript. This work was funded in part by a Defense Advanced Research Projects Agency Young Faculty Award (D13AP00043), the McGovern Institute for Brain Research, and the NSF CAREER award (CBET-1253890). This work made use of the MIT Materials Research Science and Engineering Center Shared Experimental Facilities under award DMR-0819762. R.C. and M.G.C. are supported by the NSF Graduate Research Fellowship Program and National Defense Science and Engineering Graduate fellowships, respectively. Methods of analysis and additional data are included in the supplementary materials. P.A., M.G.C., and R.C. have filed a U.S. and international patent (application PCT/US14/67866) describing magnetically multiplexed heating of volumes, which is peripherally related to this work.
View Abstract

Navigate This Article