Opportunities and challenges for biomarker discovery using electronic health record data.

Electronic health records (EHRs) have become increasingly relied upon as a source for biomedical research. One important research application of EHRs is the identification of biomarkers associated with specific patient states, especially within complex conditions. However, using EHRs for biomarker identification can be challenging because the EHR was not designed with research as the primary focus. Despite this challenge, the EHR offers huge potential for biomarker discovery research to transform our understanding of disease etiology and treatment and generate biological insights informing precision medicine initiatives. This review paper provides an in-depth analysis of how EHR data is currently used for phenotyping and identifying molecular biomarkers, current challenges and limitations, and strategies we can take to mitigate challenges going forward.

Design of a novel flow-and-shoot microbeam.

Presented here is a novel microbeam technology--the Flow-And-ShooT (FAST) microbeam--under development at RARAF. In this system, cells undergo controlled fluidic transport along a microfluidic channel intersecting the microbeam path. They are imaged and tracked in real-time, using a high-speed camera and dynamically targeted, using a magnetic Point and Shoot system. With the proposed FAST system, RARAF expects to reach a throughput of 100,000 cells per hour, which will allow increasing the throughput of experiments by at least one order of magnitude. The implementation of FAST will also allow the irradiation of non-adherent cells (e.g. lymphocytes), which is of great interest to many of the RARAF users. This study presents the design of a FAST microbeam and results of first tests of imaging and tracking as well as a discussion of the achievable throughput.

Single-camera method to determine the optical axis position of ellipsoidal drops.

The sizing of droplets by optical imaging typically requires a small depth of field so that variations in the magnification ratio are minimized. However, if the location of the drop along the optical axis can be determined, a variable magnification ratio can be imposed on each imaged drop, and the depth of field can be increased. Previous research suggested that droplet location can be determined with a characteristic of droplet images that is obtained when the droplet is illuminated from behind. In this prior research, the method was demonstrated with spherical glass objects to simulate raindrops. Raindrop are known to deviate significantly from a spherical shape, especially when the drop size is large. We demonstrate the ability to locate the position of objects that deviate from sphericity. Deformed water drops and glass ellipsoids are tested, along with glass spheres. The role of refractive index is also discussed.

Metabolism of 2-(4-methylsulphonyl-2-nitrobenzoyl)-1,3-cyclohexanedione (mesotrione) in rat and mouse.

1. The metabolic fate of [14C]-2-(4-methylsulphonyl-2-nitrobenzoyl)-1,3-cyclohexanedione (mesotrione) has been determined in the male and female rat and mouse following a single oral dose of either 1 or 100 mg kg(-1), in rat given 14 consecutive oral doses of 1 mg kg(-1), and in the surgically prepared, bile duct-cannulated rat following a single oral dose of 50 mg kg(-1). The excretion of a single i.v,. dose of 1 mg kg(-1) in the male and female rat was also investigated. 2. Mesotrione was extensively absorbed and rapidly excreted via urine in both rat and mouse. The absorbed dose was not well metabolized in either species. Unabsorbed material was subject to metabolic action by the gut microflora. 3. The major metabolic pathway was hydroxylation of the aromatic ring. There was evidence for cleavage of the dione and aromatic rings followed by reduction of the nitro group in the gastrointestinal tract. 4. There were no species differences in the metabolism and excretion of mesotrione, which could explain the species differences in toxicity reported for this class of compounds.

Deletion of a nuclease-sensitive region between the Igf2 and H19 genes leads to Igf2 misregulation and increased adiposity.

The insulin-like growth factor 2 gene (Igf2) is imprinted in most somatic tissues of the mouse with the exception of the choroid plexus and leptomeninges of the brain, where it is expressed from both alleles. The imprinting of Igf2 is dependent upon an imprinting control region (ICR) that lies 90 kb 3' of the gene and acts as a chromatin insulator to block enhancers that lie further 3' on the chromosome. Based on this model we would expect that enhancers of brain-specific expression of Igf2 would lie 5' of the ICR, and thus be insensitive to its action. Here we describe a 12 kb deletion of a region 5' of the ICR that is hypersensitive to nuclease digestion in chromatin. Its deletion results in a biallelic decrease in expression of Igf2, but not H19, in the brain, consistent with the proposal that it encodes a positive regulatory element. In addition, the deletion results in a minor relaxation of Igf2 imprinting in skeletal muscle and tongue. Lastly, the reduction in IGFII expression in the adult is accompanied by increased fat deposition and occasional obesity. Overweight animals are hypophagic, suggesting that IGFII affects fat metabolism rather than feeding behavior in adult mice.

The Dlk1 and Gtl2 genes are linked and reciprocally imprinted.

Genes subject to genomic imprinting exist in large chromosomal domains, probably reflecting coordinate regulation of the genes within a cluster. Such regulation has been demonstrated for the H19, Igf2, and Ins2 genes that share a bifunctional imprinting control region. We have identified the Dlk1 gene as a new imprinted gene that is paternally expressed. Furthermore, we show that Dlk1 is tightly linked to the maternally expressed Gtl2 gene. Dlk1 and Gtl2 are coexpressed and respond in a reciprocal manner to loss of DNA methylation. These genes are likely to represent a new example of coordinated imprinting of linked genes.

Igf2 imprinting does not require its own DNA methylation or H19 RNA.

Three models have been proposed to explain the imprinting of the mouse Igf2 gene on the maternal chromosome. We ruled out the importance of DNA methylation at Igf2 by showing that silencing of Igf2 accompanying the loss of DNA methylation could be overcome by a mutation at the neighboring H19 gene that activates Igf2. By replacing the H19 structural gene with a protein-coding gene, we have ruled out a role for H19 RNA in the imprinting of Igf2. This replacement resulted in sporadic activation of the H19 promoter on the paternal chromosome without affecting the level of expression of Igf2, a finding that is inconsistent with strict promoter competition between the genes. We conclude that a transcriptional model involving access to a common set of enhancers shared between Igf2 and H19 is the most likely explanation for Igf2 imprinting.

Design, synthesis, and in vitro evaluation of cyclic nitrones as free radical traps for the treatment of stroke.

Analogs of the cyclic nitrone free radical trap 1 (3,3-dimethyl-3,4-dihydroisoquinoline N-oxide, a cyclic analog of phenyl-tert-butylnitrone (PBN)) were prepared in which (1) the fused phenyl ring was replaced with a naphthalene ring, an electron rich heterocycle, or a dimethylphenol, (2) the nitrone-containing ring comprised five, six, or seven atoms, and (3) the gem-dimethyl group was replaced with spirocyclic groups. The most active antioxidant, which bears a dimethylphenol fused to a 7-membered ring nitrone (compound 6h), inhibited lipid peroxidation in vitro with an IC50 of 22 microM, a 75-fold improvement over that of 1. The previously observed correlation between lipophilicity and activity vs lipid peroxidation in vitro has been further substantiated and refined by this study. Moreover, certain classes of compounds (namely, dimethylphenols 6g,h and furan 6j) have now been found which are considerably more active in vitro than expected on the basis of their log k'(w) values.

Cyclic nitrone free radical traps: isolation, identification, and synthesis of 3,3-dimethyl-3,4-dihydroisoquinolin-4-ol N-oxide, a metabolite with reduced side effects.

A C-4 hydroxylated metabolite (2, 3,3-dimethyl-3,4-dihydroisoquinolin-4-ol N-oxide) of the previously described cyclic nitrone free radical trap 1 (3,3-dimethyl-3,4-dihydroisoquinoline N-oxide, a cyclic analog of phenyl-tert-butylnitrone (PBN)) was isolated, identified, and synthesized. The metabolite (2), though a less potent antioxidant than 1 in an in vitro lipid peroxidation assay, showed greatly reduced acute toxicity and sedative properties. Several analogs of 2 were prepared in attempts to improve on its weak antioxidant activity while retaining the desirable side effect profile. Effective structural changes included replacement of the gem-dimethyl moiety with spirocycloalkane groups and/or oxidation of the alcohols to the corresponding ketones. All of the analogs were more lipophilic (log k'(w)) and more active in the standard lipid peroxidation assay than 2. In addition, some of the compounds were able to protect cerebellar granule cells against oxidative damage (an in vitro model of oxidative brain injury) with IC50 values well below the value of the lead compound 1. The ketones, as predicted, were much more potent than 2 (and 1) in both of the above assays (up to ca. 200-fold). However, only compounds with a hydroxyl or an acetate group at C-4 displayed significantly reduced acute toxicity and sedative properties relative to those of 1. Importantly, the diminishment of toxicity and sedation were not the result of a lack of brain penetration as both 2 and the corresponding ketone (3,3-dimethyl-3,4-dihydro-3H-isoquinolin-4-one N-oxide) achieved equal or greater brain levels than those of 1 when administered to rats i.p.

The human growth hormone gene is regulated by a multicomponent locus control region.

The five-member human growth hormone (hGH)/chorionic somatomammotropin (hCS) gene cluster encodes the pituitary-specific hGH-N gene and four highly related genes (hGH-V, hCS-A, hCS-B, and hCS-L) that are expressed only in the placenta. When the hGH-N or hCS-A gene, together with all previously identified cis-acting regulatory sequences, was integrated into the mouse genome, it was expressed only sporadically and at low levels in the transgenic target organs. DNase I mapping of chromatin from expressing and nonexpressing cell types was used to identify a pituitary-specific set of DNase I-hypersensitive sites (HS) and a set of HS common to both the pituitary and placenta, centered approximately 15 and 30 kb 5' of hGH-N, respectively. When contained on a cosmid insert in their native genomic configuration, these HS consistently directed high-level, pituitary-specific expression of hGH-N in transgenic mice and appeared to define a locus control region required for hGH-N expression. Individually, each set of HS was able to mediate position-independent hGH-N expression in the pituitary but demonstrated loss of physiologic control and loss of tissue specificity. The gene-proximal set of HS contained a potent enhancer activity in the pituitary, while the more distal set appeared to function primarily to establish site-of-integration independence. These data indicate that synergistic interactions among multiple elements are required to restrict hGH-N transcription to the pituitary and generate appropriate levels of expression. In addition, these results suggest a role for both shared and unique regulatory sequences in locus control region-mediated expression of the hGH/hCS gene cluster in the pituitary and possibly the placenta.

Physical linkage of the human growth hormone gene cluster and the skeletal muscle sodium channel alpha-subunit gene (SCN4A) on chromosome 17.

The human growth hormone (GH) locus, a cluster of five genes, spans 47 kb on chromosome 17q22-q24. The skeletal muscle sodium channel alpha-subunit locus (SCN4A), a 32.5-kb gene, has previously been mapped to 17q23.1-q25.3. We demonstrate that both the GH gene cluster and the SCN4A gene colocalize to a single 525-kb yeast artificial chromosome (YAC) containing DNA derived from human chromosome 17. Restriction maps of two cosmids encompassing the 5' terminus of the GH locus and including up to 40 kb of 5'-flanking sequences demonstrate a perfect 20-kb overlap with previously published maps of the SCN4A gene. A 720-bp DNA segment, encompassing sequences 32.3 to 31.6 kb 5' to GH, was sequenced and found to be identical to exon 14 of SCN4A. These data demonstrate that the SCN4A gene and the entire GH gene cluster are contained within 100 kb on chromosome 17 and are separated by only 21.5 kb. Remarkably, this physical linkage between GH and SCN4A also reveals that multiple elements critical to tissue-specific transcriptional activation of the GH gene lie within the SCN4A gene.

Predictors of beef tenderness among carcasses produced under commercial conditions.

Beef carcasses (n = 240), processed using conventional commercial procedures and selected to differ in weight and s.c. fat thickness, were used to evaluate marbling score, s.c. fat thickness, 3-h pH (pH3) of the longissimus muscle (LM), and early-postmortem measurements of LM temperature as predictors of rib steak tenderness. Of the carcass traits evaluated, marbling score was the best single predictor of shear force (WBS) and panel ratings for myofibrillar tenderness (MFT). However, marbling, used alone, accounted for only 9.0 and 5.1% of the variation in WBS and MFT, respectively, and was not associated with panel ratings for connective tissue amount (CTA). Including pH3 in the prediction equation for WBS increased the R2 to .115, and inclusion of s.c. fat thickness in the equation for MFT increased the R2 to .062. Ratings for CTA were most effectively predicted using a regression equation that included 9-h LM temperature, pH3, and s.c. fat thickness (R2 = .063). Marbling score was the most effective factor evaluated for classifying carcasses into tenderness groups. Use of a minimum fat thickness constraint of .5 cm was effective for identifying tenderness differences among Select grade carcasses but was less effective within the Choice grade. Compared with marbling and s.c. fat thickness, pH3 was less effective for use in classifying carcasses into tenderness groups; however, pH3 values below 6.2 were associated with a reduction in tenderness variation. Measurements of early-postmortem LM temperature were not effective for use in identifying differences in tenderness.

Further validation of an in vitro method to reduce the need for in vivo studies for measuring the absorption of chemicals through rat skin.

Current requirements for the registration of agrochemicals, particularly in the U.S.A., often require the provision of dermal absorption data. In this process the rat is often used and complex in vivo studies, using large numbers of animals, are performed. We have compared the data obtained from in vivo and in vitro dermal absorption studies using eight pesticides with a range of physicochemical properties. Measurements were made of the 14C-labeled pesticides which could be washed from the skin, were associated with (on/in) skin, or absorbed through the skin following dermal applications in vivo and in vitro at various time points over a 24-hr exposure period. Good agreement was found between the amounts washed from and associated with the skin in vivo and in vitro. Over the time period 4-24 hr after application the in vitro experiments predicted the in vivo absorption within a factor of 2-3. These results show that, with a range of pesticide molecules, the in vitro method accurately predicted in vivo absorption supporting the utilization of the in vitro method for risk assessment from exposure to pesticides and other chemicals.

DNA base composition determines the specificity of UvrABC endonuclease incision of a psoralen cross-link.

The sequences flanking a psoralen interstrand cross-link may determine how it is repaired. Our comparison of the Escherichia coli UvrABC endonuclease incision of a variety of specific cross-link sequences in a single natural DNA fragment showed that DNA base composition determines which of two cross-linked DNA strands will be incised. G/C enrichment of the region 6-12 bases 5' of the modified T on the furan-side strand results in preferential incision of the furan-side strand. When the G/C-rich region is on the 3' side, or on neither side, incisions occur on either strand. These effects of DNA base composition suggest that UvrAB can bind in two ways to a psoralen cross-link.

The noncovalent complex between DNA and the bifunctional intercalator ditercalinium is a substrate for the UvrABC endonuclease of Escherichia coli.

We have demonstrated that the noncovalent complex formed between DNA and an antitumor bifunctional intercalator, ditercalinium, is recognized in vitro as bulky covalent DNA lesions by the purified Escherichia coli UvrABC endonuclease. It was established that no covalent drug-DNA adduct was formed during the incubation of the drug with DNA or during subsequent incubation with the UvrAB proteins. The nucleoprotein-ditercalinium complexes appear different from those generated by repair of pyrimidine dimers. The UvrA protein is able to form a stable complex with ditercalinium-intercalated DNA in the presence of ATP, whereas both UvrA and UvrB proteins are required to form a stable complex with pyrimidine dimer-containing DNA. The apparent half-life of the UvrA- and UvrAB-ditercalinium-DNA complexes following removal of free ditercalinium is 5 min. However, if the free ditercalinium concentration is maintained to allow the intercalation of one molecule of ditercalinium per 3000 base pairs, the half-life of the UvrA- or UvrAB-ditercalinium-DNA complex is 50 min, comparable to that of the complex of UvrAB proteins formed with pyrimidine dimer-containing DNA. UvrABC endonuclease incises ditercalinium-intercalated DNA as efficiently as pyrimidine dimer-containing DNA. However, unlike repair of pyrimidine dimers, the incision reaction is strongly favored by the supercoiling of the DNA substrate. Because UvrA- or UvrAB-ditercalinium-DNA complexes can be formed with relaxed DNA without leading to a subsequent incision reaction, these apparently dead-end nucleoprotein complexes may become lesions in themselves resulting in the cytotoxicity of ditercalinium. Our results show that binding of excision repair proteins to a noncovalent DNA-ligand complex may lead to cell toxicity.

Glycosylated human growth hormone variant.

The human growth hormone variant (hGH-V) gene is expressed by the syncytiotrophoblastic layer of the human placenta in two forms: hGH-V mRNA encoding a 22 kD protein, and hGH-V2 mRNA which retains intron 4 and is expected to encode a 26 kD protein. There is a predicted N-linked glycosylation site in hGH-V at amino acid 140 that is absent in both hGH-V2 and in the highly homologous normal pituitary GH (hGH-N). Cell lines transfected with the hGH-N gene secrete 22 kD GH and the 20 kD product of an alternatively spliced mRNA, while cell lines transfected with the hGH-V gene secrete three proteins of 22, 24, and 26 kD. To determine whether any of these hGH-V isoforms are glycosylated, the cell lines were grown in the absence and presence of tunicamycin. In addition, conditioned medium from metabolically labelled hGH-V transfected cells was separately digested with peptide:N-glycosidase F and endoglycosidase H. The 26 and 24 kD bands were both absent from the media after tunicamycin treatment and were both sensitive to peptide:N-glycosidase F treatment. Endoglycosidase H digestion resulted in the selective loss of the 24 kD band. These results indicate that hGH-V is partially modified posttranslationally by N-linked glycosylation in a fibroblastic cell line.

Repair of 4,5',8-trimethylpsoralen monoadducts and cross-links by the Escherichia coli UvrABC endonuclease.

Using an oligonucleotide model substrate, we observed two unusual mechanisms of UvrABC endonuclease in the repair of 4,5',8-trimethylpsoralen monoadducts and crosslinks. (i) UvrABC endonuclease usually incises a psoralen monoadduct only on the damaged strand. However, for one of the monoadducts we studied, incision on the complementary undamaged strand was also observed at a very low frequency, as though the adduct were on the thymine across from the damaged strand. Although the details of the erroneous incision are not yet known, such erroneous incision is potentially mutagenic. (ii) In cross-link repair, we observed that the UvrABC endonuclease incises the cross-linked DNA on either the furan side strand or the pyrone side strand. The incisions are not equally efficient. These data suggest that the structure of a psoralen cross-link, as seen by a repair enzyme, varies with the DNA sequence.

Alkali reversal of psoralen cross-link for the targeted delivery of psoralen monoadduct lesion.

Psoralen intercalates into double-stranded DNA and photoreacts mainly with thymines to form monoadducts and interstrand cross-links. We used an oligonucleotide model to demonstrate a novel mechanism: the reversal of psoralen cross-links by base-catalyzed rearrangement at 90 degrees C (BCR). The BCR reaction is more efficient than the photoreversal reaction. We show that the BCR occurs predominantly on the furan side of a psoralen cross-link. The cleavage does not result in the breaking of the DNA backbone, and the thymine base freed from the cross-link by the cleavage reaction appears to be unmodified. Similarly, BCR of the furan-side monoadduct of psoralen removed the psoralen molecule and regenerated the unaltered native oligonucleotide. The pyrone-side psoralen monoadduct is relatively resistant to BCR. One can use BCR to perform efficient oligonucleotide-directed, site-specific delivery of a psoralen monoadduct. As a demonstration of this approach, we have hybridized a 19 base long oligonucleotide vehicle containing a furan-side psoralen monoadduct to a 56 base long complementary oligonucleotide target strand and formed a specific cross-link at the target site with 365-nm UV. Subsequent BCR released the oligonucleotide vehicle and deposited the psoralen at the target site.

Modifications of guanine bases during oligonucleotide synthesis.

Guanine bases are sensitive to modification during automated DNA synthesis and processing reactions. Methods for the detection of two types of guanine modifications are described. The first method uses the higher reactivity of the modified G base to KMn04 oxidation than T bases, and thus allows detection by chemical DNA sequencing. The second method makes use of the Escherichia coli nucleotide excision repair enzyme UvrABC endonuclease which can detect "bulky" base modifications at each nucleotide in the synthetic DNA. Though the chemical structures of the two modifications are not known, they may be related. Both types of G modifications are often found in oligonucleotides synthesized by the methoxy-diisopropyl-phosphoramidite (MEDP) chemistry but non-detectable in the products of the beta-cyanoethyl-diisopropyl-phosphoramidite (CEDP) chemistry. The Rubin and Schmid pyrimidine-specific chemical DNA sequencing procedure (Rubin, C.M., and Schmid, C.W. (1980) Nucleic Acids Res. 8, 4613-4619) was found to be applicable to oligonucleotides synthesized by the CEDP chemistry, and to oligonucleotides synthesized by the MEDP chemistry if precautionary measures are taken to destroy the signals produced by the highly KMnO4 sensitive modified guanine bases. We also show how chemical DNA sequencing might be useful for diagnosing other chemical modifications in synthetic oligonucleotides.

Photoreactivities and thermal properties of psoralen cross-links.

We have studied the photoreaction of 8-methoxypsoralen (8-MOP), 4,5',8-trimethylpsoralen (TMP), and 4'-(hydroxymethyl)-4,5',8-trimethylpsoralen (HMT) with a pair of 18-base-long oligonucleotides in which a 14-base region is complementary. Only one 5'TpA site, favored for both monoadduct and cross-link formation with psoralen, is present in this oligonucleotide pair. We have used this model system to demonstrate, for the first time, strand specificity in the photoreaction of psoralen with DNA. We found that the two types of cross-links which form at this site have large differences in thermal stabilities. In addition, the denaturation of each cross-link isomer duplex occurred in at least three stages, which can be visualized as three bands in thermal equilibrium under the conditions of a denaturing polyacrylamide gel. This novel observation suggests that there are several domains differing in thermal stability in a psoralen cross-link.