Tay-Sachs Disease (TSD) is an inherited neurological disorder caused by deficiency of hexosaminidase A (HexA). Preclinical work demonstrated safety and efficacy of CNS gene therapy using AAVrh8-HEXA/HEXB. Here we describe an expanded access trial in two patients with infantile TSD (IND 18225).Case TSD-001 demonstrated neurodevelopmental regression by 8 months of age and severe seizures by 1 year was treated at 30 months. An equimolar mix of AAVrh8-HEXA and AAVrh8-HEXB (now AXO-AAV-GM2) was administered intrathecally (IT), with 75% of the dose (1x10 14 vg) delivered to the cisterna magna and 25% at the thoraco-lumbar junction. The second patient (TSD-002) was treated at 7 months of age with 4•2x10 13 vg by a combination of bilateral thalamic (0•18 mL; 1•5x10 12 vg per thalamus), and IT infusion (3•9x10 13 vg). Both patients underwent immunosuppression with sirolimus, corticosteroids, and rituximab.Injection procedures were well tolerated and have shown no vector-related adverse events to date. CSF HexA activity nearly doubled from baseline and remained stable. In TSD-002 (now 16 months of age), MRI showed stabilization of disease by 3 months post-injection; there now appeared to temporarily deviate from the natural history of infantile TSD but declined again 6 months post-treatment. TSD-001 (now 4.5 years of age remains seizure-free on the same anticonvulsant therapy as pre-therapy, but TSD-002 developed seizures between 13 and 17 months post-treatment (by 2 years of age).Administration of AXO-AAV-GM2 by IT and thalamic injections was safe, HexA activity increased in CSF and ongoing myelination was apparent in the younger patient treated at an early symptomatic stage. This study provides early safety and proof-of-concept in humans for treatment of TSD patients by AAV gene therapy.

Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.

Vascular Ehlers-Danlos Syndrome (previously Ehlers-Danlos IV) is a rare autosomal dominant collagen vascular disorder caused by a 2q31 COL3A1 gene mutation encoding pro-alpha1 chain of type III collagen (in contrast to classic Ehlers-Danlos, caused by a COL5A1 mutation). The vascular type accounts for less than 4% of all Ehlers-Danlos cases and usually has a poor prognosis due to life threatening vascular ruptures and difficult, frequently unsuccessful surgical and vascular interventions. In 70% of cases, vascular rupture or dissection, gastrointestinal perforation, or organ rupture is a presenting sign. We present a case of genetically proven vascular Ehlers-Danlos with fatal recurrent retroperitoneal hemorrhages secondary to a ruptured right common iliac artery dissection in a 30-year-old male. This case highlights the need to suspect collagen vascular disorders when a young adult presents with unexplained retroperitoneal hemorrhage, even without family history of such diseases.

Thalamic infusion of adeno-associated viral (AAV) vectors has been shown to have therapeutic effects in neuronopathic lysosomal storage diseases. Preclinical studies in sheep model of Tay-Sachs disease demonstrated that bilateral thalamic injections of AAV gene therapy are required for maximal benefit. Translation of thalamic injection to patients carries risks in that (1) it has never been done in humans, and (2) dosing scale-up based on brain weight from animals to humans requires injection of larger volumes. To increase the safety margin of this infusion, a flexible cannula was selected to enable simultaneous bilateral thalamic infusion in infants while monitoring by imaging and/or to enable awake infusions for injection of large volumes at low infusion rates. In this study, we tested various infusion volumes (200-800 lL) and rates (0.5-5 lL/min) to determine the maximum tolerated combination of injection parameters. Animals were followed for *1 month postinjection with magnetic resonance imaging (MRI) performed at 14 and 28 days. T1-weighted MRI was used to quantify thalamic damage followed by histopathological assessment of the brain. Trends in data show that infusion volumes of 800 lL (2 • the volume required in sheep based on thalamic size) resulted in larger lesions than lower volumes, where the long infusion times (between 13 and 26 h) could have contributed to the generation of larger lesions. The target volume (400 lL, projected to be sufficient to cover most of the sheep thalamus) created the smallest lesion size. Cannula placement alone did result in damage, but this is likely associated with an inherent limitation of its use in a small brain due to the length of the distal rigid portion and lack of stable fixation. An injection rate of 5 lL/min at a volume *1/3 of the thalamus (400-600 lL) appears to be well tolerated in sheep both clinically and histopathologically.

No abstract