Variational autoencoder: An unsupervised model for encoding and decoding fMRI activity in visual cortex.

Goal-driven and feedforward-only convolutional neural networks (CNN) have been shown to be able to predict and decode cortical responses to natural images or videos. Here, we explored an alternative deep neural network, variational auto-encoder (VAE), as a computational model of the visual cortex. We trained a VAE with a five-layer encoder and a five-layer decoder to learn visual representations from a diverse set of unlabeled images. Using the trained VAE, we predicted and decoded cortical activity observed with functional magnetic resonance imaging (fMRI) from three human subjects passively watching natural videos. Compared to CNN, VAE could predict the video-evoked cortical responses with comparable accuracy in early visual areas, but relatively lower accuracy in higher-order visual areas. The distinction between CNN and VAE in terms of encoding performance was primarily attributed to their different learning objectives, rather than their different model architecture or number of parameters. Despite lower encoding accuracies, VAE offered a more convenient strategy for decoding the fMRI activity to reconstruct the video input, by first converting the fMRI activity to the VAE's latent variables, and then converting the latent variables to the reconstructed video frames through the VAE's decoder. This strategy was more advantageous than alternative decoding methods, e.g. partial least squares regression, for being able to reconstruct both the spatial structure and color of the visual input. Such findings highlight VAE as an unsupervised model for learning visual representation, as well as its potential and limitations for explaining cortical responses and reconstructing naturalistic and diverse visual experiences.

Neural Encoding and Decoding with Deep Learning for Dynamic Natural Vision.

Convolutional neural network (CNN) driven by image recognition has been shown to be able to explain cortical responses to static pictures at ventral-stream areas. Here, we further showed that such CNN could reliably predict and decode functional magnetic resonance imaging data from humans watching natural movies, despite its lack of any mechanism to account for temporal dynamics or feedback processing. Using separate data, encoding and decoding models were developed and evaluated for describing the bi-directional relationships between the CNN and the brain. Through the encoding models, the CNN-predicted areas covered not only the ventral stream, but also the dorsal stream, albeit to a lesser degree; single-voxel response was visualized as the specific pixel pattern that drove the response, revealing the distinct representation of individual cortical location; cortical activation was synthesized from natural images with high-throughput to map category representation, contrast, and selectivity. Through the decoding models, fMRI signals were directly decoded to estimate the feature representations in both visual and semantic spaces, for direct visual reconstruction and semantic categorization, respectively. These results corroborate, generalize, and extend previous findings, and highlight the value of using deep learning, as an all-in-one model of the visual cortex, to understand and decode natural vision.

Transferring and generalizing deep-learning-based neural encoding models across subjects.

Recent studies have shown the value of using deep learning models for mapping and characterizing how the brain represents and organizes information for natural vision. However, modeling the relationship between deep learning models and the brain (or encoding models), requires measuring cortical responses to large and diverse sets of natural visual stimuli from single subjects. This requirement limits prior studies to few subjects, making it difficult to generalize findings across subjects or for a population. In this study, we developed new methods to transfer and generalize encoding models across subjects. To train encoding models specific to a target subject, the models trained for other subjects were used as the prior models and were refined efficiently using Bayesian inference with a limited amount of data from the target subject. To train encoding models for a population, the models were progressively trained and updated with incremental data from different subjects. For the proof of principle, we applied these methods to functional magnetic resonance imaging (fMRI) data from three subjects watching tens of hours of naturalistic videos, while a deep residual neural network driven by image recognition was used to model visual cortical processing. Results demonstrate that the methods developed herein provide an efficient and effective strategy to establish both subject-specific and population-wide predictive models of cortical representations of high-dimensional and hierarchical visual features.

Deep recurrent neural network reveals a hierarchy of process memory during dynamic natural vision.

The human visual cortex extracts both spatial and temporal visual features to support perception and guide behavior. Deep convolutional neural networks (CNNs) provide a computational framework to model cortical representation and organization for spatial visual processing, but unable to explain how the brain processes temporal information. To overcome this limitation, we extended a CNN by adding recurrent connections to different layers of the CNN to allow spatial representations to be remembered and accumulated over time. The extended model, or the recurrent neural network (RNN), embodied a hierarchical and distributed model of process memory as an integral part of visual processing. Unlike the CNN, the RNN learned spatiotemporal features from videos to enable action recognition. The RNN better predicted cortical responses to natural movie stimuli than the CNN, at all visual areas, especially those along the dorsal stream. As a fully observable model of visual processing, the RNN also revealed a cortical hierarchy of temporal receptive window, dynamics of process memory, and spatiotemporal representations. These results support the hypothesis of process memory, and demonstrate the potential of using the RNN for in-depth computational understanding of dynamic natural vision.

Synthesis and evaluation of non-dimeric HCV NS5A inhibitors.

Based on the symmetrical bidentate structure of the NS5A inhibitor BMS-790052, a series of new monodentate molecules were designed. The synthesis of 36 new non-dimeric NS5A inhibitors is reported along with their ability to block HCV replication in an HCV 1b replicon system. Among them compound 5a showed picomolar range activity along with an excellent selectivity index (SI > 90,000).

Synthesis and evaluation of novel potent HCV NS5A inhibitors.

Judicious modifications to the structure of the previously reported HCV NS5A inhibitor 1, resulted in more potent anti-HCV compounds with similar and in some cases improved toxicity profiles. The synthesis of 19 new NS5A inhibitors is reported along with their ability to block HCV replication in an HCV 1b replicon system. For the most potent compounds chemical stability, stability in liver microsomes and inhibition of relevant CYP450 enzymes is also presented.

Synthesis and biological evaluation of new potent and selective HCV NS5A inhibitors.

NS5A inhibitors are a new class of direct-acting antiviral agents which display very potent anti-HCV activity in vitro and in humans. Rationally designed modifications to the central biphenyl linkage of a known NS5A series led to selection of several compounds that were synthesized and evaluated in a HCV genotype 1b replicon. The straight triphenyl linked compound 11a showed similar anti-HCV activity to the clinical compound BMS-790052 and a superior cytotoxicity profile in three different cell lines, with an EC(50) value of 26 pM and a therapeutic index of over four million in an HCV replicon assay. This triphenyl analog warrants further preclinical evaluation as an anti-HCV agent.

Synthesis and antiviral activity of 2'-deoxy-2'-fluoro-2'-C-methyl-7-deazapurine nucleosides, their phosphoramidate prodrugs and 5'-triphosphates.

Thirty novel α- and ß-d-2'-deoxy-2'-fluoro-2'-C-methyl-7-deazapurine nucleoside analogs were synthesized and evaluated for in vitro antiviral activity. Several α- and ß-7-deazapurine nucleoside analogs exhibited modest anti-HCV activity and cytotoxicity. Four synthesized 7-deazapurine nucleoside phosphoramidate prodrugs (18-21) showed no anti-HCV activity, whereas the nucleoside triphosphates (22-24) demonstrated potent inhibitory effects against both wild-type and S282T mutant HCV polymerases. Cellular pharmacology studies in Huh-7 cells revealed that the 5'-triphosphates were not formed at significant levels from either the nucleoside or the phosphoramidate prodrugs, indicating that insufficient phosphorylation was responsible for the lack of anti-HCV activity. Evaluation of anti-HIV-1 activity revealed that an unusual α-form of 7-carbomethoxyvinyl substituted nucleoside (10) had good anti-HIV-1 activity (EC(50)=0.71±0.25 µM; EC(90)=9.5±3.3 µM) with no observed cytotoxicity up to 100 µM in four different cell lines.

An efficient and diastereoselective synthesis of PSI-6130: a clinically efficacious inhibitor of HCV NS5B polymerase.

R7128 is the prodrug of 2'-deoxy-2'-fluoro-2'-C-methylcytidine (PSI-6130), a potent and selective inhibitor of HCV NS5B polymerase. Currently, R7128 is in clinical trials for the treatment of HCV infection. To support clinical development efforts, we needed an efficient and scalable synthesis of PSI-6130. We describe an improved, diastereoselective synthetic route starting with protected d-glyceraldehyde. No chiral reagents or catalysts were used to produce the three new contiguous stereocenters. Introduction of fluorine at the C-2 tertiary carbon was accomplished in a highly regio- and stereoselective manner through nucleophilic substitution on a cyclic sulfate. Scale-limiting chromatographic purifications were eliminated through the use of crystalline intermediates.

Deep Residual Network Predicts Cortical Representation and Organization of Visual Features for Rapid Categorization.

The brain represents visual objects with topographic cortical patterns. To address how distributed visual representations enable object categorization, we established predictive encoding models based on a deep residual network, and trained them to predict cortical responses to natural movies. Using this predictive model, we mapped human cortical representations to 64,000 visual objects from 80 categories with high throughput and accuracy. Such representations covered both the ventral and dorsal pathways, reflected multiple levels of object features, and preserved semantic relationships between categories. In the entire visual cortex, object representations were organized into three clusters of categories: biological objects, non-biological objects, and background scenes. In a finer scale specific to each cluster, object representations revealed sub-clusters for further categorization. Such hierarchical clustering of category representations was mostly contributed by cortical representations of object features from middle to high levels. In summary, this study demonstrates a useful computational strategy to characterize the cortical organization and representations of visual features for rapid categorization.

Modulation of 5-fluorouracil host-toxicity and chemotherapeutic efficacy against human colon tumors by 5-(Phenylthio)acyclouridine, a uridine phosphorylase inhibitor.

PURPOSE: The purpose of this investigation was to evaluate the effectiveness of oral 5-(phenylthio)acyclouridine (PTAU) in reducing 5-fluorouracil (FUra) host-toxicity and enhancing its chemotherapeutic efficacy against human colon tumors. PTAU is a potent and specific inhibitor of uridine phosphorylase (UrdPase, EC 2.4.2.3), the enzyme responsible for uridine catabolism. METHODS: SCID mice bearing human colon DLD-1 or HCT-15 tumors were injected intraperitoneally with FUra (50, 200 or 300 mg/kg) on days 17, 24 and 31 after tumor cell inoculation. PTAU (120 mg/kg), uridine (1,320 mg/kg) or their combination was administered orally 2 or 4 h after FUra injection. Another four administrations of PTAU+uridine were given every 8 h after the first treatment with PTAU plus uridine. Survival and body weight were used to evaluate host toxicity. Tumor weight was used to evaluate the efficacy of the drugs on tumor growth. The mice were monitored for 38 days. RESULTS: Administration of the maximum tolerated dose (50 mg/kg) of FUra reduced DLD-1 and HCT-15 tumor weights by 48 and 59%, respectively, at day 38 post implantation. Administration of 200 mg/kg FUra resulted in 100% mortality. Oral administration of uridine (1,320 mg/kg) alone, 2 h following the administration of 200 mg/kg FUra, did not alleviate FUra host-toxicity as all the mice died. Administration of 120 mg/kg PTAUresulted in partial rescue from this lethal dose of FUra as 63% of mice survived and tumor weights were reduced by approximately 60%. Coadministration of PTAU plus uridine resulted in complete rescue from the toxicity of FUra as 100% of the mice survived and tumor weights were reduced by 81-82%. Delaying the administration of the combination of PTAU plus uridine to 4 h post FUra treatment was less effective in rescuing from FUra toxicity as only 88% of the mice survived and tumor weights were reduced by only 62%. Administration of PTAU alone, under the same conditions, resulted in a 38% survival rate while the tumor weights were reduced by 47%. Treatment with uridine alone did not protect from FUra toxicity at the dose of 200 mg/kg as all mice died. At the higher dose of 300 mg/kg FUra, neither uridine nor PTAU alone, administered 2 h following the treatment with FUra, had any rescuing effect. On the other hand, the use of the PTAU plus uridine combination reduced the tumor weight by 79%, although this reduction in the tumor weight was accompanied by 37% mortality. There was no significant difference between DLD-1 and HCT-15 in their response to the different regimens employed in this study despite the fact that the tumors have different levels of UrdPase. CONCLUSIONS: The present results demonstrate that the combination of PTAU plus uridine represents an exceptionally efficient method in increasing FUra chemotherapeutic efficacy while minimizing its host-toxicity. The efficiency of the PTAU plus uridine combination can be attributed to the extraordinary effectiveness of this combinationin raising and maintaining higher levels of uridine in vivo (Al Safarjalani et al., Cancer Chemo Pharmacol 55:541-551, 2005). Therefore, the combination of PTAU plus uridine can provide a better substitute for the large doses of uridine necessary to rescue or protect from FUra host-toxicities, without the toxic side-effects associated with such doses of uridine. This combination may also allow for the escalation of FUra doses for better chemotherapeutic efficacy against human colon carcinoma while avoiding FUra host-toxicities. Alternatively, the combination of PTAU and uridine may be useful as an antidote in the few cases when cancer patients receive a lethal overdose of FUra.

5-(Phenylthio)acyclouridine: a powerful enhancer of oral uridine bioavailability: relevance to chemotherapy with 5-fluorouracil and other uridine rescue regimens.

PURPOSE: The purpose of this investigation was to evaluate the effectiveness of oral 5-(phenylthio)acyclouridine (PTAU) in improving the pharmacokinetics and bioavailability of oral uridine. PTAU is a potent and specific inhibitor of uridine phosphorylase (UrdPase, EC 2.4.2.3), the enzyme responsible for uridine catabolism. This compound was designed as a lipophilic inhibitor in order to facilitate its access to the liver and intestine, the main organs involved in uridine catabolism. PTAU is fully absorbed after oral administration with 100% oral bioavailability. METHODS: Uridine (330, 660 or 1320 mg/kg) and/or PTAU (30, 45, 60, 120, 240 or 480 mg/kg) were orally administered to mice. The plasma levels of uridine, its catabolite uracil, and PTAU were measured using HPLC, and pharmacokinetic analysis was performed. RESULTS: Oral PTAU up to 480 mg/kg per day is not toxic to mice. Oral PTAU at 30, 45, 60, 120 and 240 mg/kg has a prolonged plasma half-life of 2-3 h, and peak plasma PTAU concentrations (C(max)) of 41, 51, 74, 126 and 161 microM with AUCs of 70, 99, 122, 173 and 225 micromol h/l, respectively. Coadministration of uridine with PTAU did not have a significant effect on the pharmacokinetic parameters of plasma PTAU at any of the doses tested. Coadministration of PTAU (30, 45, 60 and 120 or 240 mg/kg) with uridine (330, 660 or 1320 mg/kg) elevated the concentration of plasma uridine over that following the same dose of uridine alone, a result of reduced metabolic clearance of uridine as evidenced by decreased plasma exposure (C(max) and AUC) to uracil. Plasma uridine was elevated with the increase of uridine dose at each PTAU dose tested and no plateau was reached. Coadministration of PTAU at 30, 45, 60, 120 and 240 mg/kg improved the low oral bioavailability (7.7%) of uridine administered at 1320 mg/kg by 4.3-, 5.9-, 9.9-, 11.7- and 12.5-fold, respectively, and reduced the AUC of plasma uracil (1227.8 micromol h/l) by 5.7-, 6.8-, 8.2-, 6.3-, and 6.9-fold, respectively. Similar results were observed when PTAU was coadministered with lower doses of uridine. Oral PTAU at 30, 45, 60, 120 and 240 mg/kg improved the oral bioavailability of 330 mg/kg uridine by 1.7-, 2.4-, 2.6-, 5.2- and 4.3- fold, and that of 660 mg/kg uridine by 2.3-, 2.7-, 3.3-, 4.6- and 6.7-fold, respectively. CONCLUSION: The excellent pharmacokinetic properties of PTAU, and its extraordinary effectiveness in improving the oral bioavailability of uridine, could be useful to rescue or protect from host toxicities of 5-fluorouracil and various chemotherapeutic pyrimidine analogues used in the treatment of cancer and AIDS, as well as in the management of medical disorders that are remedied by the administration of uridine including CNS disorders (e.g. Huntington's disease, bipolar disorder), liver diseases, diabetic neuropathy, cardiac damage, various autoimmune diseases, and transplant rejection.

Characterization of hepatitis B virus inhibition by novel 2'-fluoro-2',3'-unsaturated beta-D- and L-nucleosides.

The clinical emergence of lamivudine and adefovir resistance mutations on prolonged therapy further necessitates the development of additional drugs for the treatment of hepatitis B virus (HBV) infections. We have evaluated a number of novel 2'-fluoro-2',3'-unsaturated D- and L-nucleosides for their anti-HBV activity in the HepG2-2.2.15 cell system. The most potent nucleosides were beta-L-2'-fluoro-2',3'-dideoxy-2',3'-didehydrocy-tidine (L-2'-Fd4C) and beta-L-2'-fluoro-2',3'-dideoxy-2',3'-didehydro-5-fluorocytidine (L-2'-Fd4FC) with median effective concentrations (EC50) of 0.002 microM and 0.004 microM, respectively. The D-enantiomers of the 2'-fluoro-substituted cytidine analogues in this series showed activity, with the 5-fluorocytidine (D-2'-Fd4FC) being the most potent (EC50 = 0.05 microM). The active compounds were not cytotoxic to a number of cell lines or to bone marrow progenitor cells. Furthermore, mitochondrial DNA synthesis and function were not affected by these nucleosides. L-2'-Fd4C did not affect viral transcription, implying that it does not inhibit cellular RNA polymerase II. Studies with the HBV polymerase in core particles revealed that the 5'-triphosphates of L-2'-Fd4C and D-2'-Fd4FC produced a dose-dependent inhibition of the incorporation of 32P-dCTP into the HBV DNA, indicating that the mechanism of action of these compounds is through specific inhibition of viral DNA synthesis. This class of nucleosides, which exhibit potent antiviral activity and a favourable safety profile, have potential for the treatment of HBV infections and warrant further development.

Unusual olefin formation by PhSe-F trans-elimination.

A new approach to the synthesis of 2',3'-didehydro-2',3-dideoxynucleosides was described in excellent yield through unusual olefin formation by PhSe-F trans-elimination.

N4-hydroxycytosine dioxolane nucleosides and their activity against hepatitis B virus.

Novel racemic, D- and L-beta-dioxolane N4-hydroxycytosine nucleosides have been synthesized and evaluated for their activity against hepatitis B virus. None of the synthesized nucleosides demonstrated selective anti-HBV activity.

Synthesis and in vitro anti-HCV activity of beta-D- and 1-2'-deoxy-2'-fluororibonucleosides.

Based on the discovery of beta-D-2'-deoxy-2'-fluorocytidine as a potent anti-hepatitis C virus (HCV) agent, a series of beta-D- and L-2'-deoxy-2'-fluoroibonucleosides with modifications at 5 and/or 4 positions were synthesized and evaluated for their in vitro activity against HCV and bovine viral diarrhea virus (BVDV). The introduction of the 2'-fluoro group was achieved by either fluorination of 2,2'-anhydronucleosides with hydrogen fluoride-pyridine or potassium fluoride, or a fluorination of arabinonucleosides with DAST. Among the 27 analogues synthesized, only the 5-fluoro compounds, namely beta-D-2'-deoxy-2',5-difluorocytidine (5), had anti-HCV activity in the subgenomic HCV replicon cell line, and inhibitory activity against ribosomal RNA. As beta-D-N4-hydroxycytidine (NHC) had previously shown potent anti-HCV activity, the two functionalities of the N4-hydroxyl and the 2'-fluoro were combined into one molecule, yielding beta-D-2'-deoxy-2'-fluoro-N4-hydroxycytidine (12). However, this nucleoside showed neither anti-HCV activity nor toxicity. All the L-forms of the analogues were devoid of anti-HCV activity. None of the compounds showed anti-BVDV activity, suggesting that the BVDV system cannot reliably predict anti-HCV activity in vitro.

Interactions of enantiomers of 2',3'-didehydro-2',3'-dideoxy-fluorocytidine with wild type and M184V mutant HIV-1 reverse transcriptase.

Both the beta-D-(+) and beta-L-(-)-enantiomers of 2',3'-didehydro-2',3'-dideoxy-5-fluorocytidine (D4FC) are clinically relevant compounds because of their potent anti-HIV and anti-HBV activities. Cross-resistance to L-D4FC with HBV containing a mutation in the conserved polymerase YMDD region has been observed. In order to better understand the effects of stereochemistry on planar 5-fluorinated cytidine analogs and to gain insight into resistance caused by YMDD mutations in HIV-1 reverse transcriptase (RT), a combination of transient kinetic studies and computer modeling were employed. In contrast to studies with the (+) and (-) isomers of 3TC-TP and FTC-TP, it was found that wild type RT had a high enantiomeric selectivity between the D-(+) and L-(-) isomers of D4FC-TP. While no resistance was conferred by the methionine 184 to valine mutation to D-D4FC-TP, L-D4FC-TP was incorporated 50- to 70-fold less efficiently. The kinetic parameters of incorporation in the presence of L-D4FC-TP by RT(WT) and the mechanism of resistance by RT(M184V) were found to be distinct from those seen with the corresponding L-isomers containing an oxathiolane ring: (-)-3TC-TP and (-)-FTC-TP. Molecular modeling suggests that L- and D-D4FC-TP are positioned in the active site favorably for incorporation by RT(WT) and that L-D4FC-TP, but not D-D4FC-TP, is sterically hindered by the addition of a beta branched amino acid at position 184 of RT(M184V).

Probing the mechanistic consequences of 5-fluorine substitution on cytidine nucleotide analogue incorporation by HIV-1 reverse transcriptase.

Beta-D and beta-L-enantiomers of 2',3'-dideoxycytidine analogues are potent chain-terminators and antimetabolites for viral and cellular replication. Seemingly small modifications markedly alter their antiviral and toxicity patterns. This review discusses previously published and recently obtained data on the effects of 5- and 2'-fluorine substitution on the pre-steady state incorporation of 2'-deoxycytidine-5'-monophosphate analogues by HIV-1 reverse transcriptase (RT) in light of their biological activity. The addition of fluorine at the 5-position of the pyrimidine ring altered the kinetic parameters for all nucleotides tested. Only the 5-fluorine substitution of the clinically relevant nucleosides (-)-beta-L-2',3'-dideoxy-3'-thia-5-fluorocytidine (L-FTC, Emtriva), and (+)-beta-D-2',3'-didehydro-2',3'-dideoxy-5-fluorocytidine (D-D4FC, Reverset), caused a higher overall efficiency of nucleotide incorporation during both DNA- and RNA-directed synthesis. Enhanced incorporation by RT may in part explain the potency of these nucleosides against HIV-1. In other cases, a lack of correlation between RT incorporation in enzymatic assays and antiviral activity in cell culture illustrates the importance of other cellular factors in defining antiviral potency. The substitution of fluorine at the 2' position of the deoxyribose ring negatively affects incorporation by RT indicating the steric gate of RT can detect electrostatic perturbations. Intriguing results pertaining to drug resistance have led to a better understanding of HIV-1 RT resistance mechanisms. These insights serve as a basis for understanding the mechanism of action for nucleoside analogues and, coupled with studies on other key enzymes, may lead to the more effective use of fluorine to enhance the potency and selectivity of antiviral agents.

Synthesis of beta-enantiomers of N4-hydroxy-3'-deoxypyrimidine nucleosides and their evaluation against bovine viral diarrhoea virus and hepatitis C virus in cell culture.

N4-Hydroxycytidine (NHC) was recently reported to have anti-pestivirus and anti-hepacivirus activity. It is thought that this nucleoside acts as a weak alternative substrate for the hepatitis C virus (HCV) polymerase. In addition to NHC, 3'-deoxyuridine (3'-dU) was found to inhibit bovine diarrhoea virus (BVDV) production by 1 log10 at 37.2 microM. These initial findings prompted the synthesis of beta-D and beta-L analogues of (i) base-modified 3'-deoxy-NHC; (ii) 3'-deoxyuridine; and 3'-deoxycytidine. The antiviral activity of these 42 nucleosides was evaluated against BVDV and HCV bicistronic replicon in cell culture. Among the NHC analogues, the antiviral activity observed for the beta-L-3'-deoxy-5-fluoro-derivative 1-(3-deoxy-beta-L-erythro-pentofuranosyl)-5-fluoro-4-hydroxyaminopyrimidin-2(1H)-one and the beta-D-3'-deoxy-5-iodo-derivative 1-(3-deoxy-beta-D-erythro-pentofuranosyl)-5-iodocytosine in the replicon system (1 log10 reduction at 100 microM) was due to the concomitant toxicity towards intracellular ribosomal RNA levels (CC90 equal or lower than the EC90). In conclusion, none of the newly synthesized derivatives exhibited enhanced antiviral activity compared to the parent nucleoside NHC.

N4-acyl-modified D-2',3'-dideoxy-5-fluorocytidine nucleoside analogues with improved antiviral activity.

A series of 2',3'-dideoxy (D2) and 2',3'-didehydro-2',3'-dideoxy (D4) 5-fluorocytosine nucleosides modified with substituted benzoyl, heteroaromatic carbonyl, cycloalkylcarbonyl and alkanoyl at the N4-position were synthesized and evaluated for anti-human immunodeficiency virus type 1 (HIV-1) and anti-hepatitis B virus (HBV) activity in vitro. For most D2-nucleosides, N4-substitutions improved the anti-HIV-1 activity markedly without increasing the cytotoxicity. In the D4-nucleosides series, some of the substituents at the N4-position enhanced the anti-HIV-1 activity with a modest increase in the cytotoxicity. The most potent and selective N4-modified nucleoside for the D2-series was N4-p-iodobenzoyl-D2FC, which had a 46-fold increase in anti-HIV-1 potency in MT-2 cells compared to the parent nucleoside D-D2FC. In the D4-series, N4-p-bromobenzoyl-D4FC was 12-fold more potent in MT-2 cells compared to the parent nucleoside D-D4FC. All eight N4-p-halobenzoyl-substituted D2- and D4-nucleosides evaluated against HBV in HepAD38 cells demonstrated equal or greater potency than the two parental compounds, D-D2FC and D-D4FC. The N4-modification especially in the D2-nucleoside series containing the N4-nicotinoyl, o-nitrobenzoyl and n-butyryl showed a significant reduction in mitochondrial toxicity relative to the parent nucleoside analogue. Although the 5'-triphosphate of the parent compound (D-D4FC-TP) was formed from the N4-acyl-D4FC analogues in different cells, the levels of the 5'-triphosphate nucleotide did not correlate with the cell-derived 90% effective antiviral concentrations (EC90), suggesting that a direct interaction of the triphosphates of these N4-acyl nucleosides was involved in the antiviral activity.