Context matters: using reinforcement learning to develop human-readable, state-dependent outbreak response policies.

The number of all possible epidemics of a given infectious disease that could occur on a given landscape is large for systems of real-world complexity. Furthermore, there is no guarantee that the control actions that are optimal, on average, over all possible epidemics are also best for each possible epidemic. Reinforcement learning (RL) and Monte Carlo control have been used to develop machine-readable context-dependent solutions for complex problems with many possible realizations ranging from video-games to the game of Go. RL could be a valuable tool to generate context-dependent policies for outbreak response, though translating the resulting policies into simple rules that can be read and interpreted by human decision-makers remains a challenge. Here we illustrate the application of RL to the development of context-dependent outbreak response policies to minimize outbreaks of foot-and-mouth disease. We show that control based on the resulting context-dependent policies, which adapt interventions to the specific outbreak, result in smaller outbreaks than static policies. We further illustrate two approaches for translating the complex machine-readable policies into simple heuristics that can be evaluated by human decision-makers. This article is part of the theme issue 'Modelling infectious disease outbreaks in humans, animals and plants: epidemic forecasting and control'. This theme issue is linked with the earlier issue 'Modelling infectious disease outbreaks in humans, animals and plants: approaches and important themes'.

Isolation of toxigenic Hafnia alvei from a probable case of hemolytic uremic syndrome.

An 11-year-old girl presented to a central California children's hospital with a 3-day history of erythematous lesions on her forehead, neck, and trunk, abdominal pain, persistent emesis, and decreased urinary output. One day prior to admission she had a mild bout of diarrhea with a small amount of blood in her stool. Upon admission her condition rapidly worsened with acute renal failure, anemia, and thrombocytopenia. One of the possible causes of this condition included hemolytic uremic syndrome. Stool cultures of this patient tested at the children's hospital and at a state reference laboratory were repeatedly negative for Escherichia coli O157:H7. However, the state reference laboratory detected a toxigenic strain of Hafnia alvei active on Vero cells from two consecutive stool cultures during the acute phase of her illness.

Mapping the ligand-binding region of Borrelia burgdorferi fibronectin-binding protein BBK32.

The cellular attachment and entry of pathogenic microorganisms can be facilitated by the expression of microbial adhesins that bind fibronectin. We have previously described a Borrelia burgdorferi gene, bbk32, that encodes a 47-kDa fibronectin-binding protein. In this study, the ligand-binding region of BBK32 from B. burgdorferi isolate B31 was localized to 32 amino acids. The bbk32 gene was cloned and sequenced from three additional B. burgdorferi isolates representing different genospecies of B. burgdorferi sensu lato. All four bbk32 genes encoded proteins having fibronectin-binding activity when expressed in Escherichia coli, and the deduced proteins shared 81 to 91% amino acid sequence identity within the ligand-binding domain. In addition, the ligand-binding region of BBK32 was found to share sequence homology with a fibronectin-binding peptide defined for protein F1 of Streptococcus pyogenes. The structural and functional similarity between the ligand-binding region of BBK32 and the UR region of protein F1 suggests a common mechanism of cellular adhesion and entry for B. burgdorferi and S. pyogenes.

Biochemical identification and characterization of DNA groups within the Proteus vulgaris complex.

We placed 43 isolates belonging to the Proteus vulgaris complex into proposed DNA groups (genomovars) using five previously recommended tests (tests for esculin hydrolysis, production of acid from salicin, L-rhamnose fermentation, and elaboration of DNase and lipase). On the basis of the results of these five tests, 49% of the isolates fell into DNA groups 5 and 6, 37% fell into DNA group 2, and the remaining 14% fell into DNA groups 3 and 4. Sequencing of the 16S rRNA genes of 11 members of DNA groups 5 and 6 indicated that 10 of these isolates (91%) could be unambiguously assigned to one of these two genomospecies. Overall expression of selected enzymatic and virulence-associated characteristics did not differ significantly among DNA groups.

Amino acid sequences of proteins from Leptospira serovar pomona.

This report describes a partial amino acid sequences from three putative outer envelope proteins from Leptospira serovar pomona. In order to obtain internal fragments for protein sequencing, enzymatic and chemical digestion was performed. The enzyme clostripain was used to digest the proteins 32 and 45 kDa. In situ digestion of 40 kDa molecular weight protein was accomplished using cyanogen bromide. The 32 kDa protein generated two fragments, one of 21 kDa and another of 10 kDa that yielded five residues. A fragment of 24 kDa that yielded nineteen residues of amino acids was obtained from 45 kDa protein. A fragment with a molecular weight of 20 kDa, yielding a twenty amino acids sequence from the 40 kDa protein.

Identification of a 47 kDa fibronectin-binding protein expressed by Borrelia burgdorferi isolate B31.

The attachment of pathogenic microorganisms to host cells and tissues is often mediated through the expression of surface receptors recognizing components of the extracellular matrix (ECM). Here, we investigate the ability of Borrelia spirochaetes to bind the ECM constituent, fibronectin. Borrelia lysates were separated by SDS-PAGE, transferred to nitrocellulose and probed with alkaline phosphatase-labelled fibronectin (fibronectin-AP). Five of six Borrelia species and four of eight B. burgdorferi sensu lato isolates expressed one or more fibronectin-binding proteins. Borrelia burgdorferi isolate B31 expressed a 47 kDa (P47) fibronectin-binding protein that was localized to the outer envelope based on susceptibility to proteinase K. The interaction of P47 with fibronectin was specific, and the region of fibronectin bound by P47 mapped to the gelatin/collagen binding domain. P47 was purified by affinity chromatography, digested with endoproteinase Lys-C, and the peptide fragments analysed by liquid chromatography/tandem mass spectroscopy. A search of protein databases disclosed that the P47 peptide mass profile matched that predicted for the bbk32 gene product of B. burgdorferi isolate B31. The bbk32 gene was cloned into Escherichia coli, and the ability of recombinant BBK32 to bind fibronectin and inhibit the attachment of B. burgdorferi was demonstrated. The identification of BBK32 as a receptor for fibronectin binding may enhance our understanding of the pathogenesis and chronic nature of Lyme disease.

Immunization with outer surface protein (Osp) A, but not OspC, provides cross-protection of mice challenged with North American isolates of Borrelia burgdorferi.

The identification of antigens with the capacity to induce a broad spectrum of protective immunity is an important consideration in the design of a Lyme disease vaccine. In this study, the range of protection provided by outer surface protein (Osp) A or OspC vaccination was compared. Mice actively immunized with OspA or OspC were challenged with 3 North American isolates of Borrelia burgdorferi. OspA-immunized mice were fully protected from infection with each of the isolates, whereas mice immunized with OspC were protected from infection with the homologous isolate but not with 2 heterologous isolates. Sequence analysis revealed that the ospA genes from these 3 isolates were >99% homologous, whereas the ospC genes shared only 81%-85% homology. Western blot analysis suggested antigenic heterogeneity associated with OspC but not OspA. These results indicate that genetic and antigenic heterogeneity may limit the usefulness of OspC as a vaccine constituent.

Antibodies to OspB prevent infection of C3H mice challenged with Borrelia burgdorferi isolates expressing truncated OspB antigens.

Truncation of outer surface protein B (OspB) of the Lyme disease agent, Borrelia burgdorferi, may allow the organism to escape immunological destruction and render an OspB-based vaccine ineffective. To investigate this possibility, we have identified two isolates, 297 and CA4, which predominantly express a truncated form of the OspB antigen. In each case, nucleic acid sequencing revealed that truncation of the OspB antigen resulted from a nonsense mutation within the 3', end of the ospB gene. Growth inhibition and protection studies demonstrated that both isolates were neutralized by an anti-OspB serum. Our results indicate that truncated forms of the OspB antigen possess epitopes that may represent important targets for neutralizing antibodies and thus, support the inclusion of OspB as a component of a subunit vaccine.

Identification and characterization of a surface-exposed, 66-kilodalton protein from Borrelia burgdorferi.

The surface-exposed antigens of Borrelia burgdorferi represent important targets for the development of a protective immune response. We have identified a proteinase K-accessible, 66-kDa protein from B. burgdorferi and have demonstrated that at least a portion of this protein is surface exposed. The 66-kDa protein was purified by sequential extraction of spirochetes with butanol and Triton X-114 followed by preparative gel electrophoresis. Polyclonal antibodies developed against the purified 66-kDa protein were Borrelia spp. specific, whereas a monoclonal antibody, Route 66, displayed a genospecies-specific pattern of recognition for the 66-kDa protein. N-terminal amino acid sequence was obtained from an internal fragment, a truncated version, and the full-length form of the 66-kDa protein. A search of protein and gene databases for homologous sequences yielded a match with the predicted amino acid sequence from a segment of B. burgdorferi chromosomal DNA (P. A. Rosa, D. Hogan, and T. G. Schwan, J. Clin. Microbiol. 29:524-532, 1991). The construction of primers based on this DNA sequence and the N-terminal amino acid sequence allowed the amplification and cloning of the 66-kDa-protein gene. The identity of the cloned gene was verified by the recognition of the expressed gene product by Route 66. Pulsed-field gel electrophoresis and Southern blot analysis were performed to confirm the chromosomal location of the 66-kDa-protein gene. This study describes the identification and cloning of the first chromosomally encoded, surface-exposed protein from B. burgdoferi.

Protection of C3H/HeN mice from challenge with Borrelia burgdorferi through active immunization with OspA, OspB, or OspC, but not with OspD or the 83-kilodalton antigen.

Recent advances in the development of animal models for Lyme borreliosis have provided means of identifying potential targets for the design of a subunit vaccine to prevent this disease. The C3H/HeN mouse model was used to study several Borrelia burgdorferi antigens from a single isolate for their ability to elicit borreliacidal and protective antibodies. The ospA, ospB, ospC, ospD, and 83-kDa genes from a California isolate, SON 188, were cloned and expressed in Escherichia coli as proteins fused to the C-terminal end of maltose-binding protein. Active immunization of mice with these fusion proteins elicited high titers of antibodies that recognized the homologous SON 188 antigens upon immunoblotting. Antibodies generated to the OspA and OspB fusion proteins, but not to the OspC, OspD, and the 83-kDa fusion proteins, demonstrated in vitro borreliacidal activity. Challenge of all actively immunized mice with 10(7) SON 188 spirochetes resulted in infection in all mice receiving the OspD or 83-kDa immunogens but not in any mice receiving the OspA, OspB, or OspC fusion proteins. These results demonstrate the potential of OspA, OspB, and OspC as components of a subunit vaccine for the prevention of Lyme borreliosis.

Characterization of a chromosomal gene and the antigen it expresses from the Lyme disease agent, Borrelia burgdorferi.

The sequence and characterization of a chromosomal gene from the Lyme disease agent Borrelia burgdorferi and the antigen it encodes are described. The gene was cloned and expressed in transformed Escherichia coli cells. The gene is composed of 597 bases and expresses a predicted protein of 199 amino acids. Antibodies specific for the recombinant antigen reacted with a single B. burgdorferi protein with a molecular mass of approximately 22 kDa. The protein was not susceptible to proteinase digestion but was extracted by n-butanol phase partitioning, suggesting a periplasmic location of the antigen. Sera from humans and canines seropositive for B. burgdorferi reacted with the recombinant antigen. The antigen characterized in this report appears to be immunologically significant in naturally infected hosts.

Sequence requirements for proteolytic processing of glycoprotein B of human cytomegalovirus strain Towne.

Truncated versions of the human cytomegalovirus (CMV) strain Towne glycoprotein B (gB) gene were stably expressed in CHO cell lines. The calcium-specific ionophore A23187 inhibited proteolytic cleavage of C-terminal-truncated gB expressed by cell line 67.77. These inhibition studies also showed that the 93-kilodalton cleavage product most likely represents the N-terminal cleavage fragment of gB. The ionophore carboxyl cyanide m-chlorophenyl-hydrazone was used to show that proteolytic cleavage of gB did not occur in the endoplasmic reticulum. Two-dimensional polyacrylamide gel electrophoresis demonstrated that the N- and C-terminal cleavage products of gB remained associated by disulfide linkages after cleavage. Expression studies using constructs in which 80% or all of the N terminus was deleted demonstrated that the N terminus was required for secretion of the gB molecule. The amino acid sequence at the site of cleavage was shown to be critical for cleavage by a cellular protease. Our results indicate that an arginine-to-threonine change at either amino acid 457 or 460, a lysine-to-glutamine change at amino acid 459, or all three substitutions together block gB cleavage. The effect on proteolysis of the arginine-to-threonine amino acid change at residue 457 (position -4 relative to the cleavage site) demonstrated that a basic pair of amino acids at the endoproteolytic processing site is not the only requirement in cis for gB cleavage.

The human cytomegalovirus strain Towne glycoprotein H gene encodes glycoprotein p86.

The gene encoding the glycoprotein H (gH) homologue of CMV strain Towne was cloned, sequenced, and expressed. The predicted 742 amino acid gH protein had characteristics typical of a membrane glycoprotein including hydrophobic signal and transmembrane domains and six possible N-linked glycosylation sites. The CMV (Towne) gH gene had a 95% nucleotide identity and a 96.6% amino acid identity with the CMV (AD169) gH gene, as described by M. P. Cranage, G. L. Smith, S. E. Bell, H. Hart, C. Brown, A. T. Bankier, P. Tomlinson, B. G. Barrell, and T. C. Minson (1988, J. Virol. 62, 1416-1422). Transcriptional analysis of the gH gene revealed that the 2.9-kilobase (kb) gH transcript was not detected until late after CMV infection, indicating that the kinetics of gH expression were typical of the late class of CMV genes. The gH gene was expressed in COS cells using a vector in which transcription was driven by the SV40 early promoter. The expression of gH was detected by immunofluorescence using the virus neutralizing murine monoclonal antibody 1G6, which is specific for an 86-kilodalton (kDa) CMV virion membrane protein (p86). Amino acid sequence analysis of p86 tryptic peptides revealed sequence identity with peptides from the deduced gH amino acid sequence, confirming that the gH gene encodes p86. These results indicate that CMV gH can induce virus neutralizing antibodies and establishes gH as a candidate antigen for a subunit vaccine against CMV.

Human cytomegalovirus strain Towne glycoprotein B is processed by proteolytic cleavage.

The gene encoding glycoprotein B of human cytomegalovirus (CMV) strain Towne was cloned, sequenced, and expressed in order to study potential targets for viral neutralization. Secondary structure analysis of the 907 amino acid protein predicted a 24 amino acid N-terminal signal sequence and a potential transmembrane region composed of two domains, 34 and 21 amino acids. The CMV (Towne) gB gene had a 94% nucleotide similarity and a 95% amino acid similarity to the CMV (AD169) gB gene [as described by M.P. Cranage et al. (1986, EMBO J. 5, 3057-3063)]. Transcriptional analysis of the CMV (Towne) gB coding strand revealed that the gB message (3.9 kb), was transcribed from this region as early as 4 hr postinfection, and well in advance of gB protein synthesis. Full-length and truncated versions of the gB gene were expressed in COS cells using expression vectors where transcription was driven by the SV40 early promoter or the CMV major immediate early promoter. Expression was detected by immunofluorescence and ELISA using the virus neutralizing murine monoclonal antibody 15D8 (L. Rasmussen, J. Mullenax, R. Nelson, and T.C. Merigan, 1985, J. Virol. 55, 274-280). This antibody had been shown previously to recognize a 55-kDa CMV virion protein and a related 130-kDa intracellular precursor. Amino acid sequence analysis of the N-terminus of the 55-kDa viral glycoprotein (gp55) showed that gp55 is derived from gB (gp130) by proteolytic cleavage and represents the C-terminal region of gp130. The truncated version of gB expressed in COS and CHO cells was also processed by proteolytic cleavage as demonstrated by Western blotting. Our study localizes the epitope recognized by 15D8 to within a 186 amino acid fragment of the gp55 protein. These results indicate that CMV gB is a target for neutralization and establishes gp55 as a candidate component for use in a subunit vaccine.