Role of Naturally Occurring Antisense mRNAs in Neural Development.

SPECIFIC AIMS

            The introduction of artificial antisense RNAs and DNAs into eukaryotic cells has become a versatile tool for manipulating gene expression.  While the experiments have demonstrated a practical value of this approach for silencing eukaryotic gene expression, the molecular mechanisms underlying such regulation remain unclear.  Clues as to potential mechanisms can be elucidated from studies of naturally occurring antisense mRNAs.   The expression of natural antisense mRNAs is well documented in prokaryotes and viruses.  Prokaryotic antisense mRNAs have been shown to regulate mRNA transcription, processing, translation and DNA replication.  The observation of a number of antisense transcripts in eukaryotes, including humans, suggests that endogenous regulatory mechanisms based on antisense mRNA also exist in higher organisms.  A recently identified naturally occurring antisense transcript corresponds to the murine hsp70.2 gene, a member of the heat shock protein (hsp) gene family (Murashov and Wolgemuth, 1995a, b).  Both sense and antisense transcripts of hsp70.2 gene are expressed in the mouse brain during embryonic and postnatal neural development.  In this application, we propose experiments to study the individual and combined roles of sense and antisense transcripts of the hsp70.2 gene during development of the central nervous system.  These experiments have two long-range goals: 1) to characterize interactions between the sense and antisense hsp70.2 transcripts in specific sites of the developing brain; and 2) to understand the functional role and significance of the hsp70.2 antisense transcript during neural development.  The specific goals for the four-year period are:

1.  Characterize the spatio-temporal patterns of expression of the hsp70.2 sense and antisense transcripts in the developing brain.   The expression will be monitored by in situ hybridization during embryonic and postnatal mouse development.  In parallel, the pattern of the expression of hsp70.2 sense transcript protein product during neural development will be assessed by immunocytochemistry.

2.   Determine the molecular structure of the hsp70.2 antisense mRNA and complete analysis of the corresponding genomic region.

3.  Investigate the possible function of the hsp70.2 antisense transcripts in the developing brain

BACKGROUND AND SIGNIFICANCE

            Many human diseases, including cancer and inherited physical and mental disorders, are caused by abnormal gene expression.  Most of such defects lack fully efficient treatments.  Advances in genetics and recombinant DNA technology have made gene therapy for treating human diseases increasingly feasible.  One of the approaches to suppressing the expression of a malfunctioning gene is the introduction of artificial antisense nucleic acids into cells.  The numerous successful applications of this technique have demonstrated that if complementary RNAs exist in the same cell simultaneously, the expression of one can be modulated by the other (Krystal, 1992).  However, the progress of gene therapy via antisense RNAs is limited by insufficient knowledge of the molecular mechanisms underlying antisense-regulated gene expression in eukaryotes.  Insight into the potential mechanisms underlying sense-antisense interactions can be elucidated from studies of naturally occurring antisense mRNAs.

            Naturally occurring antisense transcripts were first observed in prokaryotes and viruses (for review, see Eguchi et al.,  1991).  The studies in prokaryotes demonstrated that antisense RNAs can regulate mRNA transcription, processing, and translation as well as DNA replication.  A growing number of antisense transcripts has been detected in higher eukaryotes, suggesting that endogenous regulatory mechanisms based on expression of antisense mRNA exist in eukaryotic cells as well (Kimelman, 1992; Krystal, 1992).  Antisense transcripts have been reported in the rat gonadotropin-releasing hormone gene (Adelman et al., 1987; Jakubowski and Roberts, 1994), the mouse c-myc (Nepveu and Marcu, 1986; Spicer and Sonenshein, 1992) and N-myc (Krystal et al., 1990) proto-oncogenes, the rat r-erbAa (thyroid hormone receptor) gene (Lazar et al., 1989) and its human homologues ear-1 and ear-7 (Miyajima et al., 1989), the mouse (Rivkin et al., 1993) and the chicken insulin-like growth factor-II genes (Taylor et al., 1991), the mouse homeobox genes Hoxb-3 ( Swiatek and Gridley, 1993) and Hoxa-11 (Hsieh-Li et al., 1995), and recently, in the murine hsp70.2 gene, a member of the heat shock protein 70 (hsp70) gene family (Murashov and Wolgemuth, 1995a, b).  However, in none of these studies have the functions of the antisense transcripts been elucidated.

            The hsp genes are grouped on the basis of the molecular weight of their protein products and degree of structural  homology.  The three major classes include hsp90, hsp70 and hsp20.  In the murine hsp70 family, only hsp68 exhibits a true heat-inducible pattern of expression; the other genes are expressed constitutively, under nonstressed conditions (Burel et al., 1992; Terlecky et al., 1992), or as a part of normal developmental programs (Bensaude et al., 1983; Morange et al., 1984; Subjeck and Thung-Tai, 1986; Zakeri et al., 1990; Gruppi et al., 1991; Welch, 1992; Heikkila, 1993).

            The hsp70.2 gene was originally identified and characterized by our lab, by virtue of its abundant expression in germ cells in the mouse testis (Zakeri et al., 1987; 1988).  Although hsp70.2 belongs to the hsp70 gene family due to its structural homology, it is not heat-inducible.  Rather, hsp70.2 exhibits a stage-specific pattern of expression during spermatogenesis, tissue restricted expression in other reproductive organs, and developmentally regulated expression in the central nervous system (CNS).  We have also detected the expression of antisense transcripts of the hsp70.2 gene in the brain (Murashov and Wolgemuth, 1995a).  Both the sense and antisense transcripts of hsp70.2 gene were shown to be expressed in cellular- and spatio-specific patterns in the adult central nervous system, which only partially overlap (Murashov and Wolgemuth, 1995b).  Preliminary in situ hybridization analysis in the developing brain, from d 17.5 post coitum (p.c.) to d 17 postnatal (p.n.) of mouse development, also suggests a neuron-specific expression of both sense and antisense hsp70.2 transcripts.  Particularly strong expression of both transcripts was observed in developing cerebellar and cerebral cortexes, hippocampus and olfactory nuclei.  The observed specific spatio-temporal patterns of expression in the developing brain suggests an important role for hsp70.2 sense and antisense transcripts in the coordination of neural development and maturation.  

            There are several reasons underlying our choice of the hsp70.2 gene as a model for studying of the natural mechanisms of interaction and function of antisense and sense RNAs.  Both the sense and antisense transcripts of hsp70.2 exhibit region-specific and developmentally regulated patterns of expression during neural development.  There are regions in the brain where only one type of transcript exists and conversely, regions where the expression of both the sense and antisense transcripts overlaps.  There are also antibodies against the hsp70.2 sense transcript protein product that will allow us to monitor its expression, including the sites where the hsp70.2 sense and antisense transcripts are co-expressed.  Finally, the expression of several members of the hsp70 gene family in the mammalian brain, particularly the rodent, has been well documented (for reviews see Brown, 1994; Nowak et al., 1994).  The expression of hsp68, hsp71, hsp72 , hsp74 genes and/or their protein products has been shown to occur in response to a number of stressful conditions, such as brain tissue injury (Brown et al., 1989; Gonzales et al., 1989), hyperthermia (Masing and Brown, 1989; Blake et al., 1990; Miller et al., 1991; David et al., 1994), epileptic attack (Vass et al., 1989), and ischemia (Vass et al., 1988; Ferriero et al., 1990; Nowak, 1990; Gonzales et al., 1991; Kawagoe et al., 1992).  Constitutive expression of certain hsp70 genes and proteins in the brain has also been reported (Sprang and Brown, 1987; Masing and Brown, 1989; Nowak, 1990; Olazabal et al., 1992; Kawagoe et al., 1992; Manzerra and Brown, 1992; Manzerra et al., 1993), notably in dentate granule cells, the CA1 and CA3 pyramidal neurons of the hippocampus (Sprang and Brown, 1987; Brown et al., 1989; Kawagoe et al., 1992), and the granular and Purkinje cells of the cerebellum (Masing and Brown, 1989; Manzerra and Brown, 1992; Manzerra et al., 1993).  To the best of our knowledge, recent observation of the expression of the hsp70.2 sense and antisense transcripts in the brain is the first report of the production of antisense orientation transcripts from the hsp gene family (Murashov and Wolgemuth, 1995a, b).

            The molecular and cellular mechanisms that regulate the induction and development of functional and structural compartments in the brain are poorly understood, but likely involve an orderly expression of genes in a cellular, temporal, and spatio-specific manner during neural development.  One approach to investigate the induction of diverse neuronal phenotypes in the developing brain is to analyze the regulation of genes that define functional and structural compartments in the central nervous system (Kurschner and Morgan, 1995).  Our observation of the sense and antisense hsp70.2 transcripts in the brain suggests that such regulatory mechanisms could involve the expression of antisense mRNAs.

            The function of the antisense RNAs in eukaryotes in general, and in particular in the developing nervous system, is far from clear.  Thus, the study of the functions of the hsp70.2 sense and antisense transcripts during neural development is a useful model for investigation of the molecular mechanisms of sense-antisense mRNA interactions.  This study will contribute to a better understanding of both the natural mechanisms underlying antisense-regulated gene expression in eukaryotes as well as the regulation of gene expression during neural development.

RESULTS

(1).  Molecular structure of the hsp70.2 gen

            Hsp70.2, was initially identified in our lab by virtue of its abundant and developmentally regulated pattern of expression during spermatogenesis (Zakeri et al., 1988).  The genomic sequence of hsp70.2 contains a single unspliced open reading frame capable of encoding a 634 amino acid protein with a predicted molecular weight of ~69 kDa.  Hsp70.2 has extensive similarity to heat shock-inducible members of hsp70 gene family within the coding region but diverges in both its 3’ and 5’ untranslated regions.  There is a TATA box located upstream from the translation start site (Fig.1).  The hsp70.2 promoter region contains a heat shock-like element (CTGAGAGTTTCCAG), which lacks an exact match to the heat shock element (HSE) consensus sequence CNNGAANNTTCNNG.

Fig. 1.  Schematic representation of the hsp70.2 genomic clone and the fragment used for generating sense and antisense riboprobes.  The black bar represents the sense coding region; the thick line below represents the 230 bp fragment between SmaI and TaqI used for generating riboprobes.  Nucleotide sequences with presumed regulatory function are indicated as follows: HSE*, an incomplete heat shock element; TATAAG, TATA box; ATG, presumptive start of translation

            To generate sense and antisense RNA probes, we subcloned a 230-bp SmaI to TaqI fragment (430-660 nt) of genomic sequence of hsp70.2 into pBluescript SK+ (Stratagene), generating pHSP70.2-230.   This plasmid contains 200 bp of 5’ untranslated region and 30 bp of the putative coding region (Fig. 1).  The probe generated from this part of genomic sequence of hsp70.2 has been shown in our previous studies to be highly specific for hsp70.2 transcripts and does not cross-hybridize with any other members of the mouse hsp70 gene family (Zakeri et al., 1988).

(2).  Detection of antisense-orientation mRNA corresponding to the 5’ untranslated region of the hsp70.2 gene

            We have recently detected the expression of sense (2.7 kb) and antisense (2.8 kb) transcripts from the genomic region of the hsp70.2 gene in several adult mouse tissues, including the adult brain (Murashov and Wolgemuth, 1995a,b).  Given the complex structural organization of the central nervous system (CNS), we investigated whether the sense and antisense hsp70.2 transcripts were expressed throughout the entire CNS or were spatially restricted (Murashov and Wolgemuth, 1995b).  To this end, total RNA isolated from different dissected regions of the CNS was examined by Northern blot hybridization for the presence of hsp70.2 sense and antisense mRNAs.  As expected, abundant 2.7 kb sense transcripts were observed in RNA from adult mouse testis (positive control).  Exposure for a longer period revealed low levels of expression of the 2.7 kb transcripts in most areas of the brain, with distinctly higher signal in the hippocampus (Fig.2).

Fig. 2.  Expression of hsp70.2 sense and antisense transcripts in various areas of the mouse CNS.  Samples contained 20 mg of total RNA from total brain, various brain structures, spinal cord and adult testis (positive control), as indicated.   Upper panel: hybridization with ³²P- labeled hsp70.2-230 antisense riboprobe.  Exposure time was 7 days.  Middle panel: hybridization with ³²P- labeled hsp70.2-230 sense riboprobe.  Exposure time was 4 days.  Lower panel: hybridization with ³²P- labeled antisense riboprobe specific for 18S ribosomal RNA. Exposure time was 6 hours.  The open arrows indicate the position of the 28S and 18S ribosomal RNA bands.

           To determine if the 2.8 kb antisense transcripts of hsp70.2 displayed a similar distribution pattern, the blot was re-hybridized with a sense-oriented ³²P-labeled riboprobe.  The 2.8 kb antisense transcripts were clearly detected in RNA from several of the dissected brain structures.  However, the pattern of intensity and distribution of this 2.8 kb (antisense) transcripts were strikingly different from that exhibited by the 2.7 kb (sense) transcripts.  The highest levels of the 2.8 kb transcripts were observed in the brainstem, cerebellum, cortex, hippocampus, thalamus and olfactory bulbs.  Low levels of the 2.8 kb transcripts were observed in other areas of CNS including the spinal cord (Fig.2).

(3).  In situ hybridization analysis of the cellular localization of hsp70.2 sense and antisense transcripts

            To determine the cellular localization of hsp70.2 sense and antisense transcripts and their spatial pattern of distribution within the adult brain in greater detail, in situ hybridization analysis using antisense and sense ³⁵S-labeled riboprobes was conducted (Murashov and Wolgemuth, 1995b).  Hybridization of coronal sections with an hsp70.2-230 riboprobe that detects the 2.7 kb sense transcript revealed expression in regions of the hippocampus (Fig.3). 

Fig. 3.      In situ hybridization analysis using ³⁵S- labeled hsp70.2-230 antisense or sense riboprobes.  Coronal sections of mouse brain were photographed using dark-field illumination.  Left panel: hybridization with antisense hsp70.2- 230 riboprobe.   Right panel: hybridization with sense hsp70.2-230 riboprobe.  CA, pyramidal cell layers; DG, dentate gyrus; Cx, cortex; ZI, Zona incerta.  Exposure time was 14 days.

            Hybridization of adjacent sections with riboprobes to detect the 2.8 kb antisense transcript showed that both transcripts were localized to the same regions (dentate gyrus and pyramidal layers of the hippocampus).  However, strikingly higher levels of the 2.8 kb antisense transcript were detected in the cortex and thalamus (Zona incerta) (Fig.3). 

            Given the importance of determining whether similar cell types can express both sense and antisense mRNAs, in terms of understanding their functions, we extended our analysis using biotinylated riboprobes and colorimetric detection approaches, which permitted localization of transcripts to specific cell types (Murashov and Wolgemuth, 1995b; see Appendix Fig. 4).  These experiments revealed that the expression of both sense and antisense transcripts was associated with neuronal phenotypes.  Localization of the hsp70.2 sense and antisense transcripts was observed in the dentate granule cells, and the pyramidal neurons of the hippocampus.  In addition, other sites of overlapping expression of  hsp70.2 sense and antisense transcripts were detected, including the locus ceruleus, the granular layer of cerebellar cortex and the superior olives (Fig.4).

Fig. 4.  Localization of hsp70.2 transcripts recognized by antisense and sense biotinylated riboprobes in the mouse brain.  Sections in panels A, C, E, were hybridized with hsp70.2-230 antisense riboprobe.  Sections in panels B, D, F,  were hybridized with hsp70.2-230 sense riboprobe.  A and B: adjacent coronal sections of the locus ceruleus.  C and D: adjacent coronal sections of the cebebellar cortex.  The black arrows indicate the Purkinje cell layer, the white arrows indicate the granular layer.  E and F: adjacent coronal sections of the brainstem.  The short black arrows indicate the olives, the long black arrows indicate the reticular formation. 

            In contrast, the patterns of expression of the hsp70.2 sense and antisense transcripts were strikingly different in other regions of the brain.  We observed the highest levels of hsp70.2 antisense transcripts in the layer of Purkinje cells of the cerebellum, while hsp70.2 sense transcripts were not detectable (Fig.4).   Abundant hsp70.2 antisense transcripts were also observed in the frontal cortex, especially in layer II, in the anterior and posterior colliculi, the subthalamic nucleus and zona incerta, and the stratum granulosum of the area dentate, however, sense transcripts were weak or barely detectable in the same structures (Fig.5).

Fig. 5.     Detection of sense and antisense hsp70.2 transcripts in parasagittal sections of the mouse brain using biotinylated antisense and sense riboprobes.  Sections in panels A, C, E were hybridized with hsp70.2-230 antisense riboprobe.  Sections in panels B, D, F were hybridized with hsp70.2-230 sense riboprobe.  A and B: adjacent sections of frontal cortex. The white arrows indicate neurons of layer II.  C and D: adjacent sections of the mesencephalic tectum.  AC, anterior colliculus; PC, posterior colliculus.  E and F: adjacent sections of the thalamus.  Sg, the stratum granulosum area dentate; the open black arrows indicate the subthalamic nucleus; the white arrows indicate the zona incerta.  

Readily detectable expression of the antisense transcripts of hsp70.2 was also observed throughout the whole brainstem, including the reticular formation. 

           An interesting aspect of the localization of the sense and antisense transcripts was the observed cytoplasmic localization of sense transcripts and the nuclear localization of antisense transcripts, which was obvious at higher magnification.   In the reticular formation of the brainstem, the antisense RNA was detected in dendrites, axons, perikarya and nuclei of neurons (Fig.6).

Fig. 6.     Nuclear localization of the antisense transcript of hsp70.2.  Higher magnification (x250) of the coronal section through the reticular formation of the brainstem hybridized with hsp70.2-230 biotinylated sense riboprobe.  The black arrow indicates a neuron of the reticular formation with staining in dendrites, perikarya and nucleus. 

(4).  Developmentally regulated and spatio-specific expression of hsp70.2 antisense and sense transcripts in the brain

            To begin to determine whether the regional and cellular specificity of the expression of the hsp70.2 sense and antisense transcripts changes during embryonic and postnatal neural development, we examined brains of mice at d 17.5 p.c., d 1, 3, 5- 7 p.n., and d 17 p.n. by in situ hybridization .  Hsp70.2 antisense and sense transcripts were detected in mouse brains from each of the stages examined to date. 

Particularly strong expression was detected in the developing cerebellar and cerebral cortexes, anterior olfactory nucleus, and hippocampus at all examined ages (Fig. 7).  On day 6-7 p.n. the expression of the hsp70.2 sense and antisense transcripts were no longer detectable in the thalamus and weakened in the hippocampus , but were again observed on day 17 p.c.   Determining when during this period expression reappears and how this correlates with events in neural development is one of the goals of our proposed studies (Specific Aim 1).

(5).  Expression of hsp70.2 protein in the developing and adult mouse brain

            To begin to determine the expression of the hsp70.2 protein in the developing brain, immunocytochemistry was performed on paraformaldehyde-fixed free, floating sections of mouse brain at d 17.5 p.c., d 1, 3, 5-7 p.n., and d 17 p.n., using the A2 anti-hsp70.2 antibody (Rosario et al., 1993; generously provided by M. Eddy) and the ABC horseradish peroxidase kit from Vector Laboratories.  The hsp70.2 protein appears to be expressed in a pattern similar to that of the hsp70.2 sense transcript in those stages of embryonic and postnatal neural development examined to date.  Expression was observed in the cerebellar and cerebral cortexes, hippocampus and olfactory nucleus.   In the thalamus, we noticed that hsp70.2 expression was no longer detectable (at the protein level) on day 6-7 p.n., but reappeared on day 17 p.c., again similar to that observed for the hsp70.2 sense transcript.  In adult animals the hsp70.2 protein was localized primarily to the frontal cortex, hippocampus, and cerebellar granular layer (Fig. 8).  The observed specific spatio-temporal pattern of hsp70.2 expression in developing brain suggests that hsp70.2 may have a role in neural development.

Fig. 7.     Developmentally regulated expression of the hsp70.2 antisense transcripts. 

A.  d 1 p.n.  B.  d 7 p.n.  C.  d 17 p.n. Parasagittal sections of mouse brain:  Cb, the cerebellum;  Cx, the cortex; Hp, the hippocampus;  Ol, the anterior olfactory nucleus;  Th, the thalamus.

Fig. 8.  Localization of the Hsp70.2 protein in parasagittal sections of adult mouse brain. 

A. The cortex.  Black arrow indicates neurons in the frontal cortex.  B. The hippocampus. Black arrows indicate pyramidal cell layers (CA) and dentate gyrus (DG).  C.  The cerebellum.   Black arrow indicates the granular layer of the cerebellum.

(6).  Preliminary immunoblot analysis.

            Preliminary data from immunoblot analysis showed that the molecular weight of the protein expressed in the brain was higher than the protein expressed in the testis (Fig.9). 

Fig. 9.     Immunoblot analysis of Hsp70.2 protein.  Immunodetection was performed using A2 antibodies, dilution 1: 2000.  100 mg of proteins were loaded per lane.  T. Sample from adult mouse testis.  B. Sample from adult mouse brain.

This suggests that the hsp70.2 protein in the brain may undergo posttranslational modifications, including glycosylation and/or phosphorylation, which are widespread in the brain.

(7).  Preliminary structural analysis of 2.8 kb antisense transcript

            To determine if the 230 bp RNA probe recognizes more than one genomic region, we tested this probe on genomic Southern blots.  Hybridization to blots containing mouse DNA digested with either EcoRI, BamHI or HindIII respectively, revealed a single band in all three lanes.   This indicates that the 230 bp RNA probe does not cross-hybridize with any other DNA genomic sequences and likely corresponds to a single gene.

            To begin to elucidate the relationship between the 2.8 kb and 2.7 kb transcripts, we performed Northern blot hybridizations using RNA probes generated in the sense orientation from four distinct regions of the hsp70.2 genomic sequence: nucleotides 1-430, 757-1307, 1571-2027, and 1-2027 (Murashov and Wolgemuth, 1995b).  Each of the probes clearly detected the 2.8 kb antisense transcript in the brain but did not recognize any transcripts in the testis (negative control) (Fig.10). 

Fig. 10.   Hybridization to detect antisense transcripts using ³²P-labeled RNA probes corresponding to different regions of the hsp70.2 genomic sequence (1-2027 nt).  The left lane of each set of blots contained 20 mg of total RNA from brain; the right lane contained 20 mg of total RNA from testis.   Exposure time was 5 days.

These results suggested that the majority of the 2.8 kb transcript corresponds to sequences found in the 2.7 kb transcript, but in an antisense orientation.  This indicates that the sense 2.7 kb and the antisense 2.8 kb transcripts share, at least in part, an ~2 kb region (nucleotides 1- 2027) of the 5’ part of the hsp70.2 genomic sequence.  Although definitive proof awaits isolation of cDNAs corresponding to the antisense transcripts, they most likely arise from the opposite strand of hsp70.2.

            Examination of the antisense coding strand of the hsp70.2 genomic sequence did not reveal the presence of any long ORF, rather, only short ORFs capable of encoding peptides of 8 to 30 aa.  Moreover, the DNA sequence of the antisense strand of hsp70.2 did not show any significant homology to nucleic acid or protein sequence data bases in the GenBank.  However, there are three SP1 elements (at positions 1970, 2381 and 2414), a reversed TATA-box (position 2669), and a region CAGGAGCTTCTGG (position 1942) with high but incomplete homology to the heat shock element consensus sequence CNNGAANNTTCNNG (Fig.11).

Fig.11.    Schematic representation of the antisense strand of the hsp70.2 genomic sequence and the fragment used for generating sense and antisense riboprobes.  The thick line below represents the 230 bp SmaI- TaqI fragment used for generating riboprobes.  Nucleotide sequences with presumed regulatory function are indicated as follows: HSE*, an incomplete heat shock element that lacks an exact match to the heat shock element consensus sequence; SP1, regulatory elements, TTTATA, reversed TATA box.

(8).  Initial observation on the isolation of hsp70.2 genomic clones

            We have begun the isolation and mapping of genomic clones corresponding to the hsp70.2 gene.   This will enable us to characterize 5’ and 3’ flanking sequences of the hsp70.2 gene and subsequently to identify regulatory elements governing the expression of the hsp70.2 sense and antisense transcripts.  A genomic library from mouse embryonic DNA of the 129/Sv strain in Lambda FIXII was screened with a 230 bp DNA probe generated from the same insert we used for Northern blot analysis.  We isolated two clones designated 12/2 and 21/6.  Restriction mapping of the two clones showed that clone 12/2 had a complete genomic sequence of hsp70.2 plus about ~11 kb of 5’ flanking region and  ~5 kb of 3’ region.  Clone 21/6 lacks the 3’ end but has about ~15.6 kb of 5’ flanking region (Fig.12).  Both clones will be useful for future structural analysis of 5’ and 3’ flanking regions of the hsp70.2 genomic sequence.  The clones identified in this screen will also permit us to create recombinant constructs for targeted mutagenesis.