Destroying a brain tumor through an intact skull and intact blood-brain barrier
The blood-brain barrier (BBB) is the single greatest obstacle in neuro-oncology drug delivery — it blocks over 98% of small-molecule drugs from reaching brain tissue. Sonodynamic therapy sidesteps this problem entirely by using a sensitizer that crosses passively through the tumor's own leaky vasculature, and an energy source (ultrasound) that needs no barrier crossing at all.
The BBB is formed by brain endothelial cells joined by tight junctions, wrapped by pericytes and astrocyte end-feet, that together restrict paracellular diffusion and actively efflux most xenobiotics back into the blood (via P-glycoprotein and other ABC transporters). It evolved to protect the central nervous system from toxins and pathogens — but it is equally effective at excluding therapeutic drugs, including nearly all chemotherapy agents and antibodies.
Only small, highly lipophilic molecules below ~400–500 Da can cross the intact BBB by passive diffusion in meaningful quantities. This single constraint has stalled decades of otherwise-promising glioblastoma drug candidates at the barrier itself, long before target engagement becomes the limiting factor.
Glioblastoma partially disrupts its own local vasculature as it grows — new tumor vessels are structurally abnormal, with fenestrations and disrupted tight junctions, making the blood-tumor barrier (BTB) substantially leakier than the surrounding normal BBB. This is precisely the leak that oral 5-ALA exploits.
Because the tumor's own vasculature is already compromised, no separate BBB-opening step is required for 5-ALA to reach the tumor — only the intact BBB elsewhere in the brain remains a barrier, which is exactly where sonosensitizer accumulation is undesired.
5-ALA is not itself photo- or sono-active. It is a natural precursor in the heme biosynthesis pathway, administered orally at 20 mg/kg roughly 4 hours before treatment. Once absorbed, it is taken up by cells throughout the body (including tumor cells preferentially, due to their leaky vasculature and altered transporter expression) and enters the mitochondrial heme synthesis cascade.
5-ALA is already an FDA- and EMA-approved agent — marketed as Gliolan for fluorescence-guided glioma resection, where surgeons use blue light to visualize PpIX-loaded tumor tissue intraoperatively (pink fluorescence) against normal brain (blue-violet). Sonodynamic therapy repurposes this same, clinically validated pharmacokinetic pathway, only swapping the activation energy from visible light (which cannot penetrate the skull) for focused ultrasound (which can).
Normally, 5-ALA is converted step-by-step to protoporphyrin IX (PpIX), and then ferrochelatase inserts an iron atom into PpIX to produce heme, which is not sono- or photo-active. In many cancer cells — including glioma — ferrochelatase activity is relatively downregulated while porphobilinogen deaminase (upstream) is upregulated, so PpIX is produced faster than it can be converted to heme.
The result is a metabolic bottleneck: PpIX accumulates intracellularly, specifically in the tumor cells with this altered enzyme balance, while normal brain cells convert PpIX to heme efficiently and never build up a significant sonosensitizer pool. This selective accumulation — roughly 5 to 10-fold higher in glioma versus adjacent normal brain — is what allows a whole-brain-penetrating ultrasound field to activate cytotoxicity only where the tumor is.
The human skull is not acoustically transparent. It varies in thickness (roughly 4–10 mm), density, and internal trabecular structure across its surface, scattering and delaying sound waves unevenly. Before a single therapeutic pulse is fired, the exact acoustic fingerprint of the patient's own skull must be mapped and corrected for — element by element.
Ultrasound propagates at different speeds through different tissues. Soft tissue transmits sound at roughly 1540 m/s, but cortical bone transmits it much faster (~2800–3200 m/s) and attenuates it heavily due to its porous, heterogeneous trabecular structure. A beam that would converge cleanly to a millimeter-scale focus in soft tissue instead gets refracted, phase-shifted, and partially absorbed as it crosses the skull — smearing the intended focal spot and reducing peak pressure, potentially by an order of magnitude if left uncorrected.
Skull thickness and density are also highly non-uniform across an individual's head and vary substantially between patients, so a generic correction cannot be reused — every patient requires an individualized acoustic model built from their own imaging before treatment.
A pre-treatment CT scan of the skull is registered to the treatment coordinate system. Each voxel's Hounsfield unit value (a measure of X-ray attenuation, correlated with bone density) is converted into estimates of local speed-of-sound and acoustic attenuation using validated skull-tissue models.
Ray-tracing or full-wave acoustic simulation then computes, for every individual transducer element, exactly how much time delay and amplitude adjustment is needed so that its contribution arrives at the target focus in phase with every other element's contribution — despite each element's ultrasound path crossing a different thickness and density of bone.
This computation produces a per-element phase-and-amplitude correction table that is loaded into the array driver electronics immediately before treatment, and can be recomputed in real time if the target shifts.
This CT-based aberration-correction approach — built on foundational work by Kullervo Hynynen, Jean-François Aubry, and Mickael Tanter in the late 1990s and 2000s — is what converted transcranial focused ultrasound from a laboratory curiosity into a clinically viable, non-invasive neurosurgical tool.
Path planning also accounts for the trajectory of each of the (typically) 1024 individual beam paths through the skull, avoiding trajectories that would concentrate too much energy in any single region of bone (a heating risk) and confirming that the union of all corrected beams converges tightly on the tumor volume identified on the diagnostic MRI, registered into the same coordinate frame.
Only once this plan is validated — confirming an achievable focal pressure at the target with acceptable predicted skull heating — does the system proceed to firing the array.
With the correction table computed, treatment begins: a helmet-shaped array of over a thousand individual transducer elements fires in a precisely choreographed sequence. Each element's signal is delayed by its own pre-computed amount, so that despite traveling through different thicknesses of skull, every wavefront arrives at the tumor focus in phase — and interferes constructively into one sharp focal spot.
Systems such as InSightec's ExAblate Neuro arrange over a thousand individually-addressable piezoelectric elements across a hemispherical helmet that surrounds the entire head, coupled to the scalp with degassed water for acoustic transmission. Distributing elements over a large hemispherical surface — rather than a small flat panel — means each individual element only needs to contribute a small fraction of the total energy, keeping the intensity at the skull surface (and therefore skull heating) low while still achieving a therapeutically effective pressure at the deep focal point.
Each element emits a spherical wavefront. Where the phase-corrected wavefronts from many elements arrive simultaneously and in phase, their pressure amplitudes sum constructively — producing a focal peak pressure many times higher than any single element could produce alone. Everywhere else in the field, the many wavefronts arrive out of phase and largely cancel, leaving surrounding tissue exposed to only a small fraction of the focal energy.
This is the same beamforming principle used in radar and phased-array radio antennas, applied to mechanical (acoustic) waves instead of electromagnetic ones — and it is what allows the system to reach a target deep in the brain without any incision, using a beam that would otherwise be scattered into uselessness by the skull.
Because the array corrects for the skull individually per element, the same physical hardware can re-target a new focal point purely electronically — by recomputing delays — without physically moving the helmet, enabling rapid, precise coverage of an irregular tumor volume.
Higher focal pressure improves sonodynamic activation efficiency but also increases near-field intensity at the skull, where absorption generates heat. Treatment protocols therefore use pulsed, low-duty-cycle sonication (brief bursts separated by cooling intervals) rather than continuous exposure, and interleave real-time skull-surface temperature estimates from MR thermometry to keep bone heating within a safe margin throughout the multi-minute sonication.
At the focal spot, mechanical ultrasound energy is converted into a chemical cytotoxic event. Cavitation and sonoluminescence effects at the focus excite PpIX molecules sitting inside glioma cells, driving them into reactive states that generate cytotoxic reactive oxygen species — while normal brain tissue, largely devoid of accumulated PpIX, is left essentially untouched by the identical acoustic field.
Sonodynamic therapy is mechanistically related to photodynamic therapy, but substitutes ultrasound for light as the activating energy — a critical difference, since light cannot penetrate more than a few millimeters of tissue while focused ultrasound reaches centimeters deep through bone. Two overlapping mechanisms are thought to activate PpIX acoustically: inertial cavitation, where microscopic gas bubbles nucleate and collapse under the pressure field producing localized sonoluminescence (brief flashes of light) capable of exciting nearby PpIX molecules directly, and direct mechanochemical excitation of the porphyrin ring by the acoustic field itself.
Excited PpIX transfers energy either to molecular oxygen (Type II pathway, producing cytotoxic singlet oxygen) or reacts directly with substrates to form radical species (Type I pathway). Both pathways cause lipid peroxidation of cell membranes, protein oxidation, and DNA damage concentrated specifically within the PpIX-loaded cell.
The overall safety margin of sonodynamic ablation comes from two independent, multiplicative selectivity mechanisms rather than one:
• Spatial selectivity — the phased array concentrates acoustic energy almost entirely within a 2–3 mm focal spot; tissue even a short distance away receives only a small fraction of focal intensity.
• Biochemical selectivity — even within tissue that does experience meaningful acoustic exposure (e.g., at the margins of the focus), only cells that have accumulated PpIX (predominantly glioma cells, at 5–10× normal brain levels) generate significant ROS.
Together, these two independent filters mean that normal brain tissue is protected twice over: first by rarely receiving high acoustic energy, and second by lacking the sonosensitizer even where energy does reach it.
This dual selectivity — acoustic focus plus metabolic sensitizer accumulation — is what distinguishes sonodynamic ablation from purely thermal focused ultrasound ablation, which relies on spatial focusing alone and must be more conservative near critical structures.
SONALA-001, an oral 5-ALA formulation developed specifically for sonodynamic therapy, has been evaluated in early-phase clinical trials combined with low-intensity transcranial focused ultrasound in patients with recurrent glioblastoma — a population with essentially no remaining effective options after surgery, radiation, and chemotherapy have failed. These trials build directly on the decades of clinical experience with 5-ALA fluorescence-guided surgery and the maturing transcranial MRgFUS platforms already approved for other indications (essential tremor, Parkinsonian tremor), repurposing both toward a fundamentally new, non-invasive tumor-ablation modality.
Throughout treatment, real-time MRI monitors the focal region — confirming both that the ablation zone conforms to the tumor boundary and that surrounding brain and skull remain within safe thermal limits. The entire procedure, from sensitizer administration to confirmed ablation, is completed without an incision, without a craniotomy, and without disturbing the blood-brain barrier anywhere outside the tumor itself.
The proton resonance frequency (PRF) shift method exploits the fact that the resonance frequency of water protons shifts predictably and linearly with temperature. By continuously acquiring phase images during sonication, the MRI scanner reconstructs a live temperature map of the focal region and surrounding tissue with roughly ±1 °C accuracy and sub-second update rates.
This serves two purposes simultaneously: it verifies that the intended focal zone is actually receiving the planned thermal/mechanical dose (confirming treatment efficacy in real time, not just after the fact), and it acts as an automatic safety interlock — if unexpected heating appears outside the planned focus, the system can immediately reduce power or halt sonication before injury occurs.
Because the correction, the focusing, and the verification are all closed-loop and image-guided, the treating team can watch the ablation zone grow to match the tumor boundary on-screen in real time — a level of intraoperative feedback that open neurosurgery cannot match.
The same hemispherical phased-array, CT-corrected, MRI-guided hardware underlying sonodynamic ablation is already validated clinically for a widening set of neurological applications, making this a platform technology rather than a single-purpose device:
• Focal BBB opening — low-intensity ultrasound combined with intravenously injected microbubbles transiently and reversibly opens the BBB at a chosen location, enabling delivery of chemotherapy, antibodies, or gene therapy vectors that otherwise could not reach the brain (first-in-human demonstrated by Carpentier et al., 2016).
• Thermal ablation — higher-intensity continuous sonication raises focal tissue temperature above the coagulation threshold (~55–60 °C), already FDA-approved for essential tremor and tremor-dominant Parkinson's disease via thalamotomy.
• Neuromodulation — very low-intensity pulsed ultrasound can transiently excite or suppress neural activity at a target without any tissue destruction, an active area of research for psychiatric and movement disorders.
Sonodynamic ablation adds a fourth, chemically-selective mechanism to this toolkit — combining the precision of focusing with the biological selectivity of a sensitizer, rather than relying on heat or mechanical disruption alone.
Glioblastoma remains one of the most lethal solid tumors, with a median survival after standard therapy (surgical resection plus radiochemotherapy) of roughly 12–15 months, and recurrence is nearly universal — often at margins where further open surgery carries high risk to eloquent brain tissue or is simply not offered to frail or elderly patients. A fully non-invasive, repeatable, image-guided ablation option — deliverable without a craniotomy, under conscious sedation, and re-targetable purely electronically for multifocal or recurrent disease — would fill a gap that no current therapy addresses, particularly for patients who are poor candidates for repeat open surgery.