Immunoglobulin germline gene polymorphisms influence the function of SARS-CoV-2 neutralizing antibodies.

The human immunoglobulin heavy-chain (IGH) locus is exceptionally polymorphic, with high levels of allelic and structural variation. Thus, germline IGH genotypes are personal, which may influence responses to infection and vaccination. For an improved understanding of inter-individual differences in antibody responses, we isolated SARS-CoV-2 spike-specific monoclonal antibodies from convalescent health care workers, focusing on the IGHV1-69 gene, which has the highest level of allelic variation of all IGHV genes. The IGHV1-69∗20-using CAB-I47 antibody and two similar antibodies isolated from an independent donor were critically dependent on allele usage. Neutralization was retained when reverting the V region to the germline IGHV1-69∗20 allele but lost when reverting to other IGHV1-69 alleles. Structural data confirmed that two germline-encoded polymorphisms, R50 and F55, in the IGHV1-69 gene were required for high-affinity receptor-binding domain interaction. These results demonstrate that polymorphisms in IGH genes can influence the function of SARS-CoV-2 neutralizing antibodies.

(p)ppGpp controls stringent factors by exploiting antagonistic allosteric coupling between catalytic domains.

Amino acid starvation is sensed by Escherichia coli RelA and Bacillus subtilis Rel through monitoring the aminoacylation status of ribosomal A-site tRNA. These enzymes are positively regulated by their product-the alarmone nucleotide (p)ppGpp-through an unknown mechanism. The (p)ppGpp-synthetic activity of Rel/RelA is controlled via auto-inhibition by the hydrolase/pseudo-hydrolase (HD/pseudo-HD) domain within the enzymatic N-terminal domain region (NTD). We localize the allosteric pppGpp site to the interface between the SYNTH and pseudo-HD/HD domains, with the alarmone stimulating Rel/RelA by exploiting intra-NTD autoinhibition dynamics. We show that without stimulation by pppGpp, starved ribosomes cannot efficiently activate Rel/RelA. Compromised activation by pppGpp ablates Rel/RelA function in vivo, suggesting that regulation by the second messenger (p)ppGpp is necessary for mounting an acute starvation response via coordinated enzymatic activity of individual Rel/RelA molecules. Control by (p)ppGpp is lacking in the E. coli (p)ppGpp synthetase SpoT, thus explaining its weak synthetase activity.

Multimodal Mass Spectrometry Identifies a Conserved Protective Epitope in S. pyogenes Streptolysin O.

An important element of antibody-guided vaccine design is the use of neutralizing or opsonic monoclonal antibodies to define protective epitopes in their native three-dimensional conformation. Here, we demonstrate a multimodal mass spectrometry-based strategy for in-depth characterization of antigen-antibody complexes to enable the identification of protective epitopes using the cytolytic exotoxin Streptolysin O (SLO) from Streptococcus pyogenes as a showcase. We first discovered a monoclonal antibody with an undisclosed sequence capable of neutralizing SLO-mediated cytolysis. The amino acid sequence of both the antibody light and the heavy chain was determined using mass-spectrometry-based de novo sequencing, followed by chemical cross-linking mass spectrometry to generate distance constraints between the antibody fragment antigen-binding region and SLO. Subsequent integrative computational modeling revealed a discontinuous epitope located in domain 3 of SLO that was experimentally validated by hydrogen-deuterium exchange mass spectrometry and reverse engineering of the targeted epitope. The results show that the antibody inhibits SLO-mediated cytolysis by binding to a discontinuous epitope in domain 3, likely preventing oligomerization and subsequent secondary structure transitions critical for pore-formation. The epitope is highly conserved across >98% of the characterized S. pyogenes isolates, making it an attractive target for antibody-based therapy and vaccine design against severe streptococcal infections.

Structure dynamics of ApoA-I amyloidogenic variants in small HDL increase their ability to mediate cholesterol efflux.

Apolipoprotein A-I (ApoA-I) of high density lipoproteins (HDLs) is essential for the transportation of cholesterol between peripheral tissues and the liver. However, specific mutations in ApoA-I of HDLs are responsible for a late-onset systemic amyloidosis, the pathological accumulation of protein fibrils in tissues and organs. Carriers of these mutations do not exhibit increased cardiovascular disease risk despite displaying reduced levels of ApoA-I/HDL cholesterol. To explain this paradox, we show that the HDL particle profiles of patients carrying either L75P or L174S ApoA-I amyloidogenic variants show a higher relative abundance of the 8.4-nm versus 9.6-nm particles and that serum from patients, as well as reconstituted 8.4- and 9.6-nm HDL particles (rHDL), possess increased capacity to catalyze cholesterol efflux from macrophages. Synchrotron radiation circular dichroism and hydrogen-deuterium exchange revealed that the variants in 8.4-nm rHDL have altered secondary structure composition and display a more flexible binding to lipids than their native counterpart. The reduced HDL cholesterol levels of patients carrying ApoA-I amyloidogenic variants are thus balanced by higher proportion of small, dense HDL particles, and better cholesterol efflux due to altered, region-specific protein structure dynamics.

Molecularly imprinted polymers synthesized via template immobilization on fumed silica nanoparticles for the enrichment of phosphopeptides.

Phosphorylation is a protein post-translational modification (PTM) that plays an important role in cell signaling, cell differentiation, and metabolism. The hyper phosphorylated forms of certain proteins have been appointed as biomarkers for neurodegenerative diseases, and phosphorylation-related mutations are important for detecting cancer pathways. Due to the low abundance of phosphorylated proteins in biological fluids, sample enrichment is beneficial prior to detection. Thus, a need to find new strategies for enriching phosphopeptides has emerged. Molecularly imprinted polymers (MIPs) are synthetic polymeric materials manufactured to exhibit affinity for a target molecule. In this study, MIPs have been synthesized using a new approach based on the use of fumed silica as sacrificial support acting as solid porogen with the template (phosphotyrosine) immobilized on its surface. Phosphotyrosine MIPs were tested against a mixture of peptides and phosphopeptides by performing micro-solid phase extraction using MIPs (µMISPE) packed in a pipette tip. First, the capability of the materials to preferentially enrich phosphopeptides was evaluated. In a next step, the enrichment of phosphopeptides from a whole-cell lysate of human embryonic kidney (HEK) 293T cells was performed. The eluates were analyzed using MALDI-MS in the first case and with nano-HPLC-ESI-MS/MS in the second case. The results showed that the MIPs provided affinity for phosphopeptides, binding preferentially to multi-site phosphorylated peptides. The MIPs could enrich phosphopeptides in over 10-fold compared with the number of phosphopeptides found in a cell lysate without enrichment.

Mass-tag enhanced immuno-laser desorption/ionization mass spectrometry for sensitive detection of intact protein antigens.

A new read-out method for antibody arrays using laser desorption/ionization-mass spectrometry (LDI-MS) is presented. Small, photocleavable reporter molecules with a defined mass called "mass-tags" are used for detection of immunocaptured proteins from human plasma. Using prostate specific antigen (PSA), a biomarker for prostate cancer, as a model antigen, a high sensitivity generic detection methodology based immunocapture with a primary antibody and with a biotin labeled secondary antibody coupled to mass-tagged avidin is demonstrated. As each secondary antibody can bind several avidin molecules, each having a large number of mass-tags, signal amplification can be achieved. The developed PSA sandwich mass-tag analysis method provided a limit of detection below 200 pg/mL (6 pM) for a 10 µL plasma sample, well below the clinically relevant cutoff value of 3-4 ng/mL. This brings the limit of detection (LOD) for detection of intact antigens with matrix-assisted laser desorption/ionization-mass spectrometry (MALDI-MS) down to levels comparable to capture by anti-peptide antibodies selected reaction monitoring (SISCAPA SRM) and enzyme linked immunosorbent assay (ELISA), as 6 pM corresponds to a maximal amount of 60 amol PSA captured on-spot. We propose the potential use of LDI (laser desorption/ionization) with mass-tag read-out implemented in a sandwich assay format for low abundant and/or early disease biomarker detection.

Acoustic trapping for bacteria identification in positive blood cultures with MALDI-TOF MS.

Matrix assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) is currently changing the clinical routine for identification of microbial pathogens. One important application is the rapid identification of bacteria for the diagnosis of bloodstream infections (BSI). A novel approach based on acoustic trapping and an integrated selective enrichment target (ISET) microchip that improves the sample preparation step for this type of analysis is presented. The method is evaluated on clinically relevant samples in the form of Escherichia coli infected blood cultures. It is shown that noncontact acoustic trapping enables capture, enrichment, and washing of bacteria directly from the complex background of crude blood cultures. The technology replaces centrifugation-based separation with a faster and highly automated sample preparation method that minimizes manual handling of hazardous pathogens. The presented method includes a solid phase extraction step that was optimized for enrichment of the bacterial proteins and peptides that are used for bacterial identification. The acoustic trapping-based assay provided correct identification in 12 out 12 cases of E. coli positive blood cultures with an average score of 2.19 ± 0.09 compared to 1.98 ± 0.08 when using the standard assay. This new technology opens up the possibility to automate and speed up an important and widely used diagnostic assay for bloodstream infections.

Aptamer/ISET-MS: a new affinity-based MALDI technique for improved detection of biomarkers.

With the rapid progress in the development of new clinical biomarkers there is an unmet need of fast and sensitive multiplex analysis methods for disease specific protein monitoring. Immunoaffinity extraction integrated with matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) analysis offers a route to rapid and sensitive protein analysis and potentially multiplex biomarker analysis. In this study, the previously reported integrated selective enrichment target (ISET)-MALDI-MS analysis was implemented with ssDNA aptamer functionalized microbeads to address the specific capturing of thrombin in complex samples. The main objective for using an aptamer as the capturing ligand was to avoid the inherently high background components, which are produced during the digestion step following the target extraction when antibodies are used. By applying a thrombin specific aptamer linked to ISET-MALDI-MS detection, a proof of concept of antibody fragment background reduction in the ISET-MALDI-MS readout is presented. Detection sensitivity was significantly increased compared to the corresponding system based on antibody-specific binding as the aptamer ligand does not induce any interfering background residues from the antibodies. The limit of detection for thrombin was 10 fmol in buffer using the aptamer/ISET-MALDI-MS configuration as confirmed by MS/MS fragmentation. The aptamer/ISET-MALDI-MS platform also displayed a limit of detection of 10 fmol for thrombin in five different human serum samples (1/10 diluted), demonstrating the applicability of the aptamer/ISET-MALDI-MS analysis in clinical samples.

Nucleotide binding to the ATP-cone in anaerobic ribonucleotide reductases allosterically regulates activity by modulating substrate binding.

A small, nucleotide-binding domain, the ATP-cone, is found at the N-terminus of most ribonucleotide reductase (RNR) catalytic subunits. By binding adenosine triphosphate (ATP) or deoxyadenosine triphosphate (dATP) it regulates the enzyme activity of all classes of RNR. Functional and structural work on aerobic RNRs has revealed a plethora of ways in which dATP inhibits activity by inducing oligomerisation and preventing a productive radical transfer from one subunit to the active site in the other. Anaerobic RNRs, on the other hand, store a stable glycyl radical next to the active site and the basis for their dATP-dependent inhibition is completely unknown. We present biochemical, biophysical, and structural information on the effects of ATP and dATP binding to the anaerobic RNR from Prevotella copri. The enzyme exists in a dimer-tetramer equilibrium biased towards dimers when two ATP molecules are bound to the ATP-cone and tetramers when two dATP molecules are bound. In the presence of ATP, P. copri NrdD is active and has a fully ordered glycyl radical domain (GRD) in one monomer of the dimer. Binding of dATP to the ATP-cone results in loss of activity and increased dynamics of the GRD, such that it cannot be detected in the cryo-EM structures. The glycyl radical is formed even in the dATP-bound form, but the substrate does not bind. The structures implicate a complex network of interactions in activity regulation that involve the GRD more than 30 Å away from the dATP molecules, the allosteric substrate specificity site and a conserved but previously unseen flap over the active site. Taken together, the results suggest that dATP inhibition in anaerobic RNRs acts by increasing the flexibility of the flap and GRD, thereby preventing both substrate binding and radical mobilisation.

New insights into complex formation by SARS-CoV-2 nsp10 and nsp14.

SARS-CoV-2 non-structural protein 10 (nsp10) is essential for the stimulation of enzymatic activities of nsp14 and nsp16, acting as both an activator and scaffolding protein. Nsp14 is a bifunctional enzyme with the N-terminus containing a 3'-5' exoribonuclease (ExoN) domain that allows the excision of nucleotide mismatches at the virus RNA 3'-end, and a C-terminal N7-methyltransferase (N7-MTase) domain. Nsp10 is required for stimulating both ExoN proofreading and the nsp16 2'-O-methyltransferase activities. This makes nsp10 a central player in both viral resistance to nucleoside-based drugs and the RNA cap methylation machinery that helps the virus evade innate immunity. We characterised the interactions between full-length nsp10 (139 residues), N- and C-termini truncated nsp10 (residues 10-133), and nsp10 with a C-terminal truncation (residues 1-133) with nsp14 using microscale thermophoresis, multi-detection SEC, and hydrogen-deuterium (H/D) exchange mass spectrometry. We describe the functional role of the C-terminal region of nsp10 for binding to nsp14 and show that full N- and C-termini of nsp10 are important for optimal binding. In addition, our H/D exchange experiments suggest an intermediary interaction of nsp10 with the N7-MTase domain of nsp14. In summary, our results suggest intermediary steps in the process of association or dissociation of the nsp10-nsp14 complex, involving contacts between the two proteins in regions not identifiable by X-ray crystallography alone.

Frequent use of IGHV3-30-3 in SARS-CoV-2 neutralizing antibody responses.

The antibody response to SARS-CoV-2 shows biased immunoglobulin heavy chain variable (IGHV) gene usage, allowing definition of genetic signatures for some classes of neutralizing antibodies. We investigated IGHV gene usage frequencies by sorting spike-specific single memory B cells from individuals infected with SARS-CoV-2 early in the pandemic. From two study participants and 703 spikespecific B cells, the most used genes were IGHV1-69, IGHV3-30-3, and IGHV3-30. Here, we focused on the IGHV3-30 group of genes and an IGHV3-30-3-using ultrapotent neutralizing monoclonal antibody, CAB-F52, which displayed broad neutralizing activity also in its germline-reverted form. IGHV3-30-3 is encoded by a region of the IGH locus that is highly variable at both the allelic and structural levels. Using personalized IG genotyping, we found that 4 of 14 study participants lacked the IGHV3-30-3 gene on both chromosomes, raising the question if other, highly similar IGHV genes could substitute for IGHV3-30-3 in persons lacking this gene. In the context of CAB-F52, we found that none of the tested IGHV3-33 alleles, but several IGHV3-30 alleles could substitute for IGHV3-30-3, suggesting functional redundancy between the highly homologous IGHV3-30 and IGHV3-30-3 genes for this antibody.

Targeting Toll-like receptor-driven systemic inflammation by engineering an innate structural fold into drugs.

There is a clinical need for conceptually new treatments that target the excessive activation of inflammatory pathways during systemic infection. Thrombin-derived C-terminal peptides (TCPs) are endogenous anti-infective immunomodulators interfering with CD14-mediated TLR-dependent immune responses. Here we describe the development of a peptide-based compound for systemic use, sHVF18, expressing the evolutionarily conserved innate structural fold of natural TCPs. Using a combination of structure- and in silico-based design, nuclear magnetic resonance spectroscopy, biophysics, mass spectrometry, cellular, and in vivo studies, we here elucidate the structure, CD14 interactions, protease stability, transcriptome profiling, and therapeutic efficacy of sHVF18. The designed peptide displays a conformationally stabilized, protease resistant active innate fold and targets the LPS-binding groove of CD14. In vivo, it shows therapeutic efficacy in experimental models of endotoxin shock in mice and pigs and increases survival in mouse models of systemic polymicrobial infection. The results provide a drug class based on Nature´s own anti-infective principles.

Miniaturized and automated high-throughput verification of proteins in the ISET platform with MALDI MS.

A major bottleneck in high-throughput protein production is the validation step, which is why parallel and automated sample processing methods are highly desirable. Also, a miniaturized sample preparation format is preferred, as the reduction of reagent volumes significantly decreases the analysis cost per sample. We have developed an automated and miniaturized protein sequence verification protocol for recombinant proteins utilizing peptide mass fingerprinting and MS/MS analysis. The integrated selective enrichment target (ISET) platform, previously developed in our group, with its dual functionality, being both a sample preparation platform and a MALDI target plate, is employed. All steps including immobilized metal ion affinity chromatography of protein on cobalt-loaded beads, tryptic digestion, and MALDI MS analysis are performed in an array format, without any sample transfers, on the same ISET chip. The automated configuration reduced the sample preparation time significantly. Starting with crude lysate, a full plate of 48 purified, digested samples prepared for MALDI-MS can be generated in 4 h, with only 30 min of operator involvement. This paper demonstrates the utility of the method by parallel analysis of 45 His-tagged human recombinant proteins.

Optimizing nanovial outlet designs for improved solid-phase extraction in the integrated selective enrichment target--ISET.

The integrated selective enrichment target is a microfluidic platform for SPE sample preparation with integrated nanocolumns, which simultaneously offers direct MALDI MS read-out. Here, we present a study on the importance of different nanocolumn outlet hole geometries and hole areas in relation to MS signal intensity and reproducibility. A design solution that provides the flow characteristics required for robust sample preparation using automated liquid handling is reported.

An image registration method for voxel-wise analysis of whole-body oncological PET-CT.

Whole-body positron emission tomography-computed tomography (PET-CT) imaging in oncology provides comprehensive information of each patient's disease status. However, image interpretation of volumetric data is a complex and time-consuming task. In this work, an image registration method targeted towards computer-aided voxel-wise analysis of whole-body PET-CT data was developed. The method used both CT images and tissue segmentation masks in parallel to spatially align images step-by-step. To evaluate its performance, a set of baseline PET-CT images of 131 classical Hodgkin lymphoma (cHL) patients and longitudinal image series of 135 head and neck cancer (HNC) patients were registered between and within subjects according to the proposed method. Results showed that major organs and anatomical structures generally were registered correctly. Whole-body inverse consistency vector and intensity magnitude errors were on average less than 5 mm and 45 Hounsfield units respectively in both registration tasks. Image registration was feasible in time and the nearly automatic pipeline enabled efficient image processing. Metabolic tumor volumes of the cHL patients and registration-derived therapy-related tissue volume change of the HNC patients mapped to template spaces confirmed proof-of-concept. In conclusion, the method established a robust point-correspondence and enabled quantitative visualization of group-wise image features on voxel level.

Multivariate mining of an alpaca immune repertoire identifies potent cross-neutralizing SARS-CoV-2 nanobodies.

Conventional approaches to isolate and characterize nanobodies are laborious. We combine phage display, multivariate enrichment, next-generation sequencing, and a streamlined screening strategy to identify numerous anti-severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) nanobodies. We characterize their potency and specificity using neutralization assays and hydrogen/deuterium exchange mass spectrometry (HDX-MS). The most potent nanobodies bind to the receptor binding motif of the receptor binding domain (RBD), and we identify two exceptionally potent members of this category (with monomeric half-maximal inhibitory concentrations around 13 and 16 ng/ml). Other nanobodies bind to a more conserved epitope on the side of the RBD and are able to potently neutralize the SARS-CoV-2 founder virus (42 ng/ml), the Beta variant (B.1.351/501Y.V2) (35 ng/ml), and also cross-neutralize the more distantly related SARS-CoV-1 (0.46 µg/ml). The approach presented here is well suited for the screening of phage libraries to identify functional nanobodies for various biomedical and biochemical applications.

Pseudouridine-modified tRNA fragments repress aberrant protein synthesis and predict leukaemic progression in myelodysplastic syndrome.

Transfer RNA-derived fragments (tRFs) are emerging small noncoding RNAs that, although commonly altered in cancer, have poorly defined roles in tumorigenesis1. Here we show that pseudouridylation (Ψ) of a stem cell-enriched tRF subtype2, mini tRFs containing a 5' terminal oligoguanine (mTOG), selectively inhibits aberrant protein synthesis programmes, thereby promoting engraftment and differentiation of haematopoietic stem and progenitor cells (HSPCs) in patients with myelodysplastic syndrome (MDS). Building on evidence that mTOG-Ψ targets polyadenylate-binding protein cytoplasmic 1 (PABPC1), we employed isotope exchange proteomics to reveal critical interactions between mTOG and functional RNA-recognition motif (RRM) domains of PABPC1. Mechanistically, this hinders the recruitment of translational co-activator PABPC1-interacting protein 1 (PAIP1)3 and strongly represses the translation of transcripts sharing pyrimidine-enriched sequences (PES) at the 5' untranslated region (UTR), including 5' terminal oligopyrimidine tracts (TOP) that encode protein machinery components and are frequently altered in cancer4. Significantly, mTOG dysregulation leads to aberrantly increased translation of 5' PES messenger RNA (mRNA) in malignant MDS-HSPCs and is clinically associated with leukaemic transformation and reduced patient survival. These findings define a critical role for tRFs and Ψ in difficult-to-treat subsets of MDS characterized by high risk of progression to acute myeloid leukaemia (AML).

Noncovalent antibody immobilization on porous silicon combined with miniaturized solid-phase extraction (SPE) for array based immunoMALDI assays.

This paper presents a new strategy to combine the power of antibody based capturing of target species in complex samples with the benefits of microfluidic reverse phase sample preparation on an integrated sample enrichment target (RP-ISET) and the analysis speed of MALDI MS. The immunoaffinity step is performed on an in-house developed 3D-structured high surface area porous silicon (PSi) matrix, which allows efficient antibody immobilization by surface adsorption without any coupling agents in 30-60 min. The hydrophilic nature of the porous silicon surface at the molecular level displays a low adsorption of background peptides when exposed to complex digests or plasma samples, improving the conditions for the antigen specific extraction and subsequent readout. At the same time, the hydrophobic behavior, due to the nanostructured surface, of the PSi material facilitates liquid confinement during the assay. Using a footprint conforming to the standard for 384 well microplates, direct adaption of the protocol into standard sample handling robots is possible. The performance of the proposed immunoaffinity PSi-ISET immunoMALDI (iMALDI) assay was evaluated by specific detection of angiotensin I at a 10 femtomol level in diluted plasma samples (10 µL, 1 nM).

Binding of the human antioxidation protein α1-microglobulin (A1M) to heparin and heparan sulfate. Mapping of binding site, molecular and functional characterization, and co-localization in vivo and in vitro.

Heparin and heparan sulfate (HS) are linear sulfated disaccharide polymers. Heparin is found mainly in mast cells, while heparan sulfate is found in connective tissue, extracellular matrix and on cell membranes in most tissues. α1-microglobulin (A1M) is a ubiquitous protein with thiol-dependent antioxidant properties, protecting cells and matrix against oxidative damage due to its reductase activities and radical- and heme-binding properties. In this work, it was shown that A1M binds to heparin and HS and can be purified from human plasma by heparin affinity chromatography and size exclusion chromatography. The binding strength is inversely dependent of salt concentration and proportional to the degree of sulfation of heparin and HS. Potential heparin binding sites, located on the outside of the barrel-shaped A1M molecule, were determined using hydrogen deuterium exchange mass spectrometry (HDX-MS). Immunostaining of endothelial cells revealed pericellular co-localization of A1M and HS and the staining of A1M was almost completely abolished after treatment with heparinase. A1M and HS were also found to be co-localized in vivo in the lungs, aorta, kidneys and skin of mice. The redox-active thiol group of A1M was unaffected by the binding to HS, and the cell protection and heme-binding abilities of A1M were slightly affected. The discovery of the binding of A1M to heparin and HS provides new insights into the biological role of A1M and represents the basis for a novel method for purification of A1M from plasma.

Faster dense deformable image registration by utilizing both CPU and GPU.

Purpose: Image registration is an important aspect of medical image analysis and a key component in many analysis concepts. Applications include fusion of multimodal images, multi-atlas segmentation, and whole-body analysis. Deformable image registration is often computationally expensive, and the need for efficient registration methods is highlighted by the emergence of large-scale image databases, e.g., the UK Biobank, providing imaging from 100,000 participants. Approach: We present a heterogeneous computing approach, utilizing both the CPU and the graphics processing unit (GPU), to accelerate a previously proposed image registration method. The parallelizable task of computing the matching criterion is offloaded to the GPU, where it can be computed efficiently, while the more complex optimization task is performed on the CPU. To lessen the impact of data synchronization between the CPU and GPU, we propose a pipeline model, effectively overlapping computational tasks with data synchronization. The performance is evaluated on a brain labeling task and compared with a CPU implementation of the same method and the popular advanced normalization tools (ANTs) software. Results: The proposed method presents a speed-up by factors of 4 and 8 against the CPU implementation and the ANTs software, respectively. A significant improvement in labeling quality was also observed, with measured mean Dice overlaps of 0.712 and 0.701 for our method and ANTs, respectively. Conclusions: We showed that the proposed method compares favorably to the ANTs software yielding both a significant speed-up and an improvement in labeling quality. The registration method together with the proposed parallelization strategy is implemented as an open-source software package, deform.