Green fluorescent protein (GFP) is a useful reporter to follow the in vivo behaviour of proteins, but the wild-type gfp gene does not function in many organisms, including many plants and filamentous fungi. We show that codon-modified forms of gfp, produced for use in plants, function effectively in Aspergillus nidulans both as gene expression reporters and as vital reporters for protein location. To demonstrate the use of these modified gfps as reporter genes we have used fluorescence to follow ethanol-induced GFP expression from the alcA promoter. Translational fusions with the modified gfp were used to follow protein location in living cells; plant ER-retention signals targeted GFP to the endoplasmic reticulum, whereas fusion to the GAL4 DNA-binding domain targeted it to the nucleus. Nuclear-targeted GFP allowed real-time observation of nuclear movement and division. These modified gfp genes should provide useful markers to follow gene expression, organelle behaviour and protein trafficking in real time.
The fungus Aspergillus nidulans is a useful model system for understanding the molecular basis of eukaryotic cellular morphogenesis as well as for asking more specific questions about fungal development. Genetic analysis of cell division has identified several important regulators, motors and structural components required for nuclear division (reviewed by Doonan, 1992). Genetic dissection of nuclear movement has similarly identified a number of novel genes and proven that cytoskeletal components, such as tubulin, are essential for correct nuclear positioning (reviewed by Beckwith et al., 1995). Development in Aspergillus involves the sequential production of a few well-defined cell types. An extensive collection of mutants, covering almost every aspect of the life of the fungus, has allowed a detailed genetic dissection of morphogenesis (reviewed by Timberlake, 1990). In addition, Aspergillus nidulans is closely related to species of economic or medical importance, making a good model for understanding the molecular and cellular basis of growth in an important group of organisms. A major limitation to exploiting this wealth of biological materials is the absence of convenient methods to monitor the behaviour of cellular components or to identify cell types within the living organism.
The green fluorescent protein (GFP), originally isolated from the jellyfish Aequorea victoria, is an attractive marker to follow gene expression and protein location in living cells (Chalfie et al., 1994). Other reporter genes have been developed and are widely used but most of these, for example glucuronidase (Jefferson et al., 1987) and luciferase (Gallie et al., 1989), require exogenous substrates and cell permeabilization. The GFP chromophore is formed as a result of protein folding and fluorescence does not require any extrinsic factors except blue or UV light. Thus, living cells can be observed with minimal perturbation and intercellular activities can be observed directly.
The native version of the gfp gene, however, is often poorly expressed in heterologous systems and requires modification. For instance, some plant species recognize a splice site within the coding sequence and so produce a partial gene product that does not fluoresce (Haseloff and Amos, 1995; Haseloff et al., 1997). GFP fluorescence in mammalian cells can be enhanced by culturing at low temperatures and expression can also be improved if specially adapted genes are used (Zolotukhin et al., 1996). The native gfp gene is poorly expressed in many fungi and Streptomycetes (Fernández-Ábalos, unpublished) but the reason(s) for this is not clear. Suboptimal codon usage in the native gfp gene is a possible reason for poor expression in fungi.
In this report, we show that several gfp genes with modified codon usage provide convenient reporters for gene expression in Aspergillus, whereas the native gene does not. The underlying reason for this is a failure of the native gene to produce significant amounts of protein in the fungus. The gfp genes that have been modified for use in plants do produce significant amounts of protein. We have used these modified gfp genes to make translational fusions that are targeted to either the endoplasmic reticulum (ER) or nucleus within living cells and have used the nuclear targeted GFP to follow the events of nuclear division for the first time in a filamentous fungus. Therefore, when adapted to its host, this versatile reporter gene is an efficient and effective means of monitoring gene expression and protein localization in filamentous fungi.
Results and discussion
gfp genes, optimized for expression in plants, function in Aspergillus
Initial experiments with the native gfp gene, cloned in the pAL5 expression vector, failed to produce detectable fluorescence when transformed into Aspergillus. We therefore explored the possibility that some of the modified gfp genes produced for use in other species could be used directly to overcome this problem. We cloned four different gfp genes into the alcohol-inducible expression vector, pAL5 (Fig. 1A; Doonan et al., 1991). We used this inducible expression system as expression of high levels of GFP could be toxic but neither the GFP nor the GFP fusions reported here had significant effects on fungal growth.
To examine the expression of GFP in hyphae, spores were patched on to ethanol induction medium and grown at 25°C. Plates containing five gfp transformants for each construct and a vector-only control were examined using an epifluorescence microscope. Transformants containing the gfp2 construct did not display any detectable green fluorescence when grown on ethanol media (Fig. 2A, GFP2) and were essentially indistinguishable from GR5 transformants containing an empty pAL5 vector (data not shown). The GFP2 protein contains the S65→T mutation, which has a single excitation peak centred at 490 (Heim et al., 1995) and the V163→A substitution, which enhances solubility, perhaps via improving protein folding (Siemering et al., 1996). However, transformants containing the gfp2-4, gfp2-5 and sgfp constructs were readily detectable, although fluorescence intensity varied depending on the construct. Germlings carrying each construct were photographed and printed under identical conditions, shown in Fig. 2A. gfp2-4 contains the S65→T substitution in addition to a region with optimal plant codon usage and produces a readily detectable level of fluorescence. GFP2-5 cells were brighter than GFP2-4, but GFP2-5 has three additional substitutions (V163→A, I167→T, and S175→G) that may contribute to enhanced fluorescence (Heim et al., 1995; Siemering et al., 1996). sgfp, containing only the S65→T mutation, produced the brightest fluorescence, despite the absence of the V163→A, I167→T, and S175→G mutations.
The absence of fluorescence from transformants carrying the gfp2 construct could be due to a number of reasons, including failure to produce mRNA or protein. Suboptimal codon usage has been reported for a number of other fungal species (Spellig et al., 1996; Cormack et al., 1997) and in plants the native gfp transcript contains inappropriate intron splice sites resulting in excision of part of the ORF (Haseloff et al., 1997). As a first step towards defining the reason why gfp2 does not produce any useful fluorescence in Aspergillus and why transfromants containing the other constructs vary in brightness we examined the amount of protein produced (Fig. 2B and C), as compared to the intensity of GFP fluorescence (D). This experiment clearly shows that the intensity of fluorescence in the protein extract from a given gfp transfromant is influenced primarily by the amount of GFP protein produced. First, gfp2 constructs did not produce any detectable GFP protein under either induced or non-induced conditions: gfp2 transformants (lane 2) were indistinguishable from pAL5 transformants (lane V) under both induced and non-induced conditions, as judged by both western blotting (Fig. 2C) and fluorimetry (Fig. 2D). gfp2-4, gfp2-5 and sgfp transfromants grown under inducing conditions produced protein (lanes 16, 17 and 32 in Fig. 2) detectable by western blotting (C) and observable as an extra band on SDS–PAGE gels (arrowed in B). gfp2-4 and gfp2-5 transformants produced approximately equal amounts of protein, as judged by western blotting, but the brightness of gfp2-5 transformants was consistently 1.5- to 2-fold greater than gfp2-4 transformants. The substitutions, V163→A, I167→T, and S175→G, found in GFP2-5 therefore seem to have a significant enhancing effect. In fluorimetry experiments, sgfp transformants tended to produce higher intensities than gfp2-4 or gfp2-5. However, western blot experiments show that sgfp transformants produce higher levels of protein, thereby contributing to the increased fluorescent signal. These differences are also reflected at the transcript level: the gfp 2 transcript cannot be detected on Northern blots, whereas sgfp produces stronger signals than gfp2-4 or gfp2-5. This suggests that the sgfp gene, having been modified throughout its entire length, may produce a more stable transcript than the partially modified gfp2-4 and gfp2-5, which, in turn, are more stable than the unmodified gfp gene. The sequence changes made in the modified genes tend to increase the overall GC content of the transcript and may lead to more stable transcripts. According to Haseloff et al. (1997) the changes made in the gfp sequence affect the central region of the gene (from NdeI to AccI sites, incorporated in our gfp2-4 and gfp2-5 versions) and affect not only the site of processing of the cryptic intron but also the usage of many codons in this region. The changes effectively prevent mis-splicing but no direct effect on translatability can be drawn from their data. Interestingly, sgfp was designed with modifications of the codon usage to adapt it to animal cells (Haas et al., 1996), but these modifications allowed improved expression in plant cells and removed the cryptic splicing points (Chiu et al., 1996). We (J. M. Fernández-Ábalos, unpublished) have found that only the sgfp gene functions in Streptomyces, a prokaryotic organism with highly biased codon usage (Wright and Bibb, 1992). Taken together, these observations suggest that the increase in GC content of the modified gfp genes increases the stability of the transcripts in these microorganisms.
To determine the extent to which gfp could be used as a dynamic reporter of gene expression, we investigated the time course of alcohol induction and repression of GFP expression from the alcA promoter. Cells transformed with pMCB32 (sgfp) were grown overnight (about 12 h) in shaken rich glucose medium (YG) at 37°C and then switched to minimal ethanol medium at 30°C. Protein extracts were made from samples taken every 30 min for the appearance of green fluorescence. No fluorescence was detected until 1.5 h and thereafter accumulated over a period of several hours (Fig. 2E). When pMCB32 cells grown on ethanol were switched to minimal glucose medium, fluorescence (per μg of protein) was maintained at a high level for at least 4 h and then declined by about 40% every hour (data not shown), As the nuclear doubling time is about 90–120 min under these conditions, this decline represents the combined effect of dilution by cell growth and protein degradation. Considering that the alcA promoter responds quickly to changes in carbon source (Pateman et al., 1983), this suggests that GFP protein is turned over rather slowly in Aspergillus.
Location of GFP within the hyphae
One of the major attractions of GFP is the ability to study dynamic events associated with protein targeting and its regulation in vivo. The constructs pMCB16, 17 and 32 were all designed to produce soluble GFP. As judged by microscopic observation, soluble GFP was present throughout the cytoplasm and was able to enter the nucleus. The nuclei often appeared brighter than surrounding cytoplasm, suggesting that GFP accumulated within them. However, GFP was excluded from vacuoles and mitochondria, which were present along the length of the hyphae. Immunogold staining of hyphae using antibodies against GFP confirmed that GFP was absent from the mitochondria and vacuoles but abundant in the cytoplasm and nuclei (data not shown).
Subcellular localization of GFP using targeting sequences
The ability of gfp transcriptional fusions to monitor protein trafficking has been demonstrated in a number of systems (reviewed by Cubitt et al., 1995). To test if GFP could be targeted to a specific location, in this case the ER, we cloned mgfp5 (Siemering et al., 1996, J. Haseloff, personal communication) containing ER retention signals into pAL5. Two other versions of GFP, GFP2-5 and sGFP, were also engineered to place the plant chitinase export signal at the N-terminus of GFP along with the ER retention peptide signal on the C-terminus. These constructs directed GFP to a tubular network within the cell. GFP2-5 produced the brightest fluorescence. A z-series of optical sections through hyphae (Fig. 3) shows GFP2-5 directed to a branching tubular network containing variously shaped brighter nodes. This network extends throughout the cell to within 1–2 μm from the tip and surrounds the nucleus. Vacuoles, the cytoplasm and the nucleus all contained low levels of GFP. Equivalent constructs in plants have been demonstrated to target to the ER (Boevink et al., 1996), illuminating a network similar to that which we describe here.
GFP can be used to follow nuclear division in Aspergillus
GFP is widely used to follow organelle behaviour (for example Rizzuto et al., 1995). To follow nuclear division we used a GAL4 DNA-binding domain–GFP fusion, which we know locates to the nucleus of yeast cells (J. M. Fernández-Ábalos, unpublished). In Aspergillus the GAL4 BD–sGFP fusion is targeted to the nucleoplasm and the spindle plaque, producing an image very similar to the DNA-specific dye DAPI. The nucleolus contains very low levels of GFP. A useful aspect of the GAL4 BD–sGFP protein is its constant association with the nucleus throughout the cell cycle. This permits us to follow the dynamics of nuclear division. Time-lapse photography (Fig. 4) was used to follow a nucleus from the early stages of mitosis (panels 0 and 1,5), through mid-phase (panels 5 and 9), anaphase (panels 14.5 and 16.5) telophase (panels 18.5 and 21) and decondensation as the daughter nuclei re-enter interphase (panels 21, 24 and 29). Under these conditions (effective observation temperature, 19°C), mitosis takes between 15 and 22 min.
By careful examination of optically sectioned nuclei in the early stages of mitosis, we can discriminate up to seven discrete strands in individual nuclei. The GFP image of early mitosis differs markedly form that observed using DNA-specific dyes on fixed material in that these strands are not normally observed. However, as the later stages of mitosis as observed by GAL4 BD–sGFP closely resemble that of fixed material and the organism has eight genetic linkage groups, we suggest that these strands may represent individual chromosomes. During the early stages of mitosis, the GFP-stained chromosomes appear to be attached at a single point, initially at their extremes but as mitosis proceeds the region of attachment becomes more diffuse and tends to move towards the middle of the strands. As the nuclei approach mid-phase, the chromosomes condense further so that discrete strands are impossible to resolve (panel 5). Anaphase A movement occurs very quickly, taking less than 1 min to resolve the chromatids into two groups, about 1–1.5 μm apart. After the initial separation, the daughter nuclei remain connected by a thin bridge of GFP-stained material and may hold this configuration for several minutes. Checks on DNA segregation may occur during this apparent pause as we have observed occasional bidirectional exchanges of material between the two chromatid masses: small amounts of GFP-stained material move from one nucleus into the other and then back (data not shown). When such exchanges do occur, this phase of mitosis seems to be prolonged, taking up to 15 min in one extreme case (as against the norm of 3–4 min). These observations would be compatible with the existence of a mid-anaphase checkpoint as observed in budding yeast (Yang et al., 1997) implicated as a final check on the correct segregation of the genetic material before the two daughters become separate physical entities. After anaphase, the separation between daughter nuclei increases to a maximum of 3–5 μm over a period of about two minutes. As the nuclei re-enter interphase (as judged by the reappearance of a nucleolus) they may recoil together again slightly before taking up their final positions. Nuclear movement during interphase is relatively slow (compared with mitotic movements). (See note added in proof.)
Unmodified gfp genes fail to produce useful green fluorescence when expressed in Aspergillus. We have taken advantage of the modified forms of gfp that have recently been produced for use in plants to demonstrate their use for following gene expression, protein localization and nuclear behaviour in Aspergillus. Expression of GFP in Arabidopsis is low because of the presence of a cryptic intron, 84 bp long (Haseloff and Amos, 1995; Haseloff et al., 1997). Haseloff's group modified gfp to specifically prevent intron splicing by changing the codons in this region (modifications that are present in our gfp2-4 version) and introduced enhancing mutations (present in our gfp2-5). Haas et al. (1996) have changed codon usage throughout the entire gene producing the sgfp gene, which should also eliminate the splice recognition sites and, because of improved codon usage, may enhance protein production (Chiu et al., 1996). Recently, the sgfp gene has been used as a vital marker in the phytopathogenic fungus Ustilago maydis (Spellig et al., 1996) and Cormack et al. (1997) have shown that codon adaptation is also necessary for expression of GFP in Candida. Here, we demonstrate that both the intron-modified gfp and the completely synthetic sgfp genes are both effective in Aspergillus. These data, combined with the observation that the sgfp gene is the only one expressed in other species with extreme codon bias such as Streptomyces (incidentally, the codon usage in sgfp closely approaches the codon preference in Streptomyces, J. M. Fernández-Ábalos, unpublished), suggest that it may be necessary to produce a specifically adapted gfp for each group of organisms.
Our results also indicate that subcellular localisation signals from other species, such as the ER retention signals from plants and the nuclear localization signals present in GAL4 DNA-binding domain from yeast, are functionally conserved. Thus, GFP fusion proteins containing such signals are effective for marking organelles and subcellular compartments in filamentous fungi without major alterations and should be useful for following protein trafficking. However, we did not find that it was necessary to restrict GFP to the ER to obtain acceptable levels of fluorescence as is the case in Arabidopsis (Haseloff et al., 1997). Finally, the GAL4 BD-GFP fusion marks nuclei and chromatin, allowing us to describe the dynamics of nuclear division in Aspergillus for the first time.
Construction of GFP expression plasmids
gfp expression constructs, summarized in Fig. 1, were produced using standard protocols (Sambrook et al., 1986). A gfp variant with mutations S56→T and V163→A (called gfp2 throughout this work) and cloned in pBluescript KS(+) (plasmid pKSGFP2) was used as the starting point for most of the constructions. The NdeI/BstBI in gfp2 was replaced with the corresponding NdeI/BstBI fragment from mgfp4 or mgfp5 that carry modifications of the DNA sequence to adapt the codon usage in this region to that in higher plants and to incorporate several point mutations (Siemering et al., 1996) giving gfp2-4 and gfp2-5 versions. All three variants were inserted in the Aspergillus vector pAL5 (Doonan et al., 1991) as Asp718I/BamHI fragments to produce pAL5GFP2, pMCB16 and pMCB17 respectively (Fig. 1A).
In a different set of constructions we used a synthetic sgfp gene (sGFPS65T, Chiu et al., 1996, Haas et al., 1996, referred to as sgfp in this work) in which the whole sequence of the gfp gene has been changed to adapt its codon usage for mammalian and plant cells and carrying mutation S65→T (Chiu et al., 1996; Haas et al., 1996,). As the starting point for our constructions we used plasmid pB1 (that contains sgfp before the ‘nos’ transcriptional terminator, J. Sheen personal communication). The ‘nos’ terminator was looped out by digestion with PstI and religation, giving plasmid pMCB30 that contains appropriate Asp718I and BamHI sites before and after sgfp, and that were used to transfer it to pAL5 to give plasmid pMCB32 (Fig. 1A).
Construction of GFP-translational fusions
For ER-targeted GFPs, we used mgfp5 (Siemering et al., 1996) that contains a plant chitinase signal peptide at the N-terminus and an ER retention signal (His–Asp–Glu–Leu) at the C-terminus, either directly or with replacement of the internal region of the gene with appropriate fragments from gfp2-5 or sgfp. All three versions were inserted in pAL5 at the Asp718I/SmaI sites, giving plasmids pMCB40 (with mgfp5), pMCB41 (with sgfp) and pMCB45 (with gfp2–5 ) (Fig. 1B). A translational fusion was made to target GFP to the nucleus using the GAL4 DNA-binding domain from budding yeast (Fields and Song, 1989). The NcoI/BamHI fragment from pMCB30 that carries sgfp was inserted in NcoI/BamHI sites of the yeast vector pAS2. This generates an in-frame fusion of sgfp with the DNA-binding domain of GAL4 (GAL4BD) and the HA tag present in the vector, and rendered plasmid pMCB31. Such a fusion is targeted to the nucleus in yeast (data not shown). Digestion of pMCB31 with HindIII (before the coding region of the GAL4BD) and PstI (after sgfp) provides a DNA fragment that was cloned in pBluescript SK(+) to give plasmid pMCB38, when appropriate Asp718I and BamHI sites allowed for cloning of the fusion in pAL5. The resulting plasmid, pMCB42, contains the fusion GAL4BD:HA tag:sGFP (Fig. 1C).
Culture conditions, strains and transformation
The Aspergillus strain GR5 (pyrG89, wA3), used for all manipulations, was grown on YG media (2% w/v yeast extract, 1% glucose, 1% agar) containing 10 mM uridine and uracil and transformed as previously described (Osmani et al., 1988). Briefly, protoplasts were prepared from freshly germinated conidiospores and transformed with plasmid DNA using PEG. Transformants were selected on YG media containing 0.6 M KCl and lacking uridine/uracil and allowed to sporulate. Conidiospores were streaked to single colony three times on selective media to obtain pure transformants. Expression of GFP was monitored microscopically by plating spores on minimal agar media containing 1% ethanol to induce gene expression or on 100 mM glucose to repress expression (Waring et al., 1989); a coverslip was placed on top of the germlings and observations made as described below.
To quantify the GFP production, spores (107 ml−1) were germinated in shaken liquid YG medium for 12 h, harvested by filtration through Miracloth and washed with 200 ml of sterile water, resuspended in either induction (2% lactose/1% ethanol) or repression (3% glucose/1% sucrose) media and grown for 4 h at 30°C. Cells were harvested as before and disrupted in lysis buffer (100 mM Tris-HCl pH 7.5, 1 mM EDTA, 5 mM DTT, 1 mM freshly added PMSF, 5 μg ml−1 aprotinin and 5 μg ml−1 pepstatin-A, Calera et al., 1997) in 2 ml tubes containing 0.5 g of 0.4–0.6 mm glass beads, using 3 × 20 s pulses at 6.0 speed in a Hybaid Ribolyser. Lysates were cooled on ice, spun at 14 000 r.p.m. for 5 min at room temperature (RT) and the cleared supernatant (representing total soluble proteins) was kept frozen at −20°C. Protein concentration was determined using a modified Bradford assay (BioRad) and samples were prepared for SDS–PAGE by adding 90 μl of cleared lysate to 30 μl of preheated (95°C) 4× sample buffer. A 10 μg sample of protein was loaded to each lane of a 15% polyacrylamide gel. After separation, the gel was blotted to Immobilon P membrane (Millipore) and, after blocking in 5% dried milk in Tris-buffered saline (TBS, 10 mM Tris-HCl, 150 mM NaCl, pH 8), was probed with anti-GFP polyclonal antibodies (Clontech) at 1:20000 dilution in TBS containing 1% dried milk. The blot was washed for 15 min each in (I) TBS + 0.05% Trition X-100 (TBST); (ii) TBST + 0.5 M NaCl; and (iii) TBST, before incubating with HRP-conjugated swine anti-rabbit IgG 1:10000 in TBS with 1% dried milk. After washing as described above, the blot was developed using the SuperSignal ULTRA chemioluminscence detection system (Pierce) and recorded using Kodak X-Omat film. Bands were scanned into Adobe Photoshop and the relative amounts of GFP protein estimated using the same package.
Cell lysates obtained as described above were used for fluorimetry. Duplicate samples were measured in a Titertek Fluoroskan II fluorimeter using a 480/538 nm excitation/emission filter block, with 100 mM Tris-HCl pH 7.5 buffer as a blank. Fluorescence was recorded as absolute fluorescence units, normalized against the protein concentration in the samples and expressed as arbitrary fluorescence units per μg protein. Extracts from a transformant containing an empty pAL5 expression plasmid were used to determine the background fluorescence.
Conidiospores from at least five independent transformants were patched on ethanol induction media (Waring et al., 1989) and allowed to germinate overnight at 25°C. To reduce background the media did not contain riboflavin. Just before observation a coverslip was placed on top of the germinating spores and they were observed using Neofluar ×100 lens, NA 1.3 on either a Zeiss epifluorescence microscope or on a BioRad MRC-600 confocal microscope using a 488 nm Krypton laser and standard Zeiss FITC filter block. Images were collected using NIH Image and transferred to Adobe Photoshop version 4.0.
Note added in proof
While this work was under review the following paper described the use of SGFP fusions with stuA to follow nuclear migration in Aspergillus: Suelmann, R., Sievers, N., and Fischer, R. (1997) Nuclear traffic in fungal hyphae: in vivo study of nuclear migration and positioning in Aspergillus nidulans. Mol Microbiol25: 757–769.
We are grateful to Jim Haseloff for modified mgfp4 and mgfp5 clones and helpful advice, to Roger Tsein, Jen Sheen and Tilman Spellig for the synthetic sgfp, to Fernando Leal for advice on protein extracts and to Carol Wymer and Peter Shaw for assistance with confocal microscopy and image analysis. J.M.F-A was supported by a short-term grant from the European Union PTP Programme and HF and CP by Studentships from the BBSRC.