The Cyanophage Molecular Mixing Bowl of Photosynthesis Genes

Among the wealth of microbial organisms inhabiting marine environments, cyanobacteria (blue-green algae) are the most abundant photosynthetic cells. Prochlorococcus and Synechococcus, the two most common cyanobacteria, account for 30% of global carbon fixation (through the photosynthetic process in which sugars are manufactured from carbon dioxide and water). By drawing on natural resources, these microbes use photosystems (PS) I and II (the two reaction centers in photosynthesis) to harness energy. 
 
Intriguingly, some viruses that infect cyanobacteria (called cyanophage), carry genes that encode two PSII core reaction-center proteins: PsbA (the most rapidly turned over core protein in all oxygen-yielding photosynthetic organisms) and PsbD (which forms a complex with PsbA). By expressing their own copies of psbA and psbD during infection, these cyanophages have managed to co-opt host genes to suit their own purposes: enhancing photosynthesis. It seems likely that they do this in the interests of their own fitness, since cyanophage production is optimal when photosynthesis is maintained during infection. 
 
Until recently, only a small sample of cyanophages had been examined, leaving open the questions of how widespread PSII genes are in these organisms and where the genes came from. To answer these questions, Matthew Sullivan, Debbie Lindell, Sallie Chisholm, and colleagues examined a pool of 33 cyanophage isolates (cultured from samples collected from the Sargasso Sea and the Red Sea), along with data already available for nine other cyanophages, for the presence of psbA and psbD genes. They found psbA was present in 88% and psbD in 50% of the cyanophages studied. By analyzing the sequences of these genes along with those from Prochlorococcus and Synechococcus host genes, they reconstructed the evolutionary history of how the PSII genes entered the phage genomes. 
 
Cyanophages are divided morphologically into three main families (Podoviridae, Myoviridae, and Siphoviridae). Looking at the distributions of the PSII genes across the different families, Sullivan, Lindell, et al. saw that psbA was present in all myoviruses and all Prochlorococcus podoviruses, but not in Prochlorococcus siphoviruses or Synechococcus podoviruses. The high levels of sequence conservation between the different cyanophages suggest that this gene is probably functional and that it is likely to increase the reproductive fitness of the phage. The length of the latent period may impact the distribution pattern of psbA among these phage groups. However, more information about the physiological characteristics of cyanophages is needed to further investigate these possibilities. 
 
The second gene, psbD, was less prolific but was seen in four of the 20 Prochlorococcus myoviruses and 17 of the 20 Synechococcus myoviruses examined—all of which also encoded psbA. Myoviruses are known to infect a wider range of cyanobacteria than the other cyanophage families. Indeed, when investigated, the psbD-encoding myoviruses correlated with those known to have a broader host range. Perhaps the co-opting of both PSII genes ensures a functional PsbA–PsbD protein complex to enhance infection for these cyanophages that are able to infect a wider range of hosts. 
 
To determine when the PSII genes had been transferred into the phage and from where, Sullivan, Lindell, et al. investigated the nucleotide sequences of psbA and psbD from both Prochlorococcus and Synechococcus host and cyanophage. Using meticulous sequence analyses and standard statistical methods, they generated phylogenetic trees to explain the evolutionary history of these two PSII genes. 
 
By analyzing the clusters of sequence types within the resulting tree, the authors saw evidence that psbA was transferred from the cyanobacteria host genome into the phage genome on four independent occasions and two separate occasions for psbD. Exchange events were generally host-range specific, meaning that Prochlorococcus genes transferred to Prochlorococcus phages, and so on. However, a few intriguing exceptions, where genes did not cluster with their hosts, were observed; these might result from genetic exchange between members of two different phage families (one of broader host range) during co-infection of the same host. 
 
Sullivan, Lindell, et al. were also able to use their dataset to investigate a previous suggestion that alterations in the nucleotide distributions within individual PSII genes (creating a kind of patchwork gene) demonstrate that intragenic recombination has taken place. Indeed, they confirm that this occurs among Synechococcus myoviruses and Prochlorococcus podoviruses. In some cases involving Synechococcus, intragenic recombination appears to have happened in both host-to-phage and phage-to-host directions for both genes; and, for some Prochlorococcus genes, DNA from an unknown source also seems to have been inserted. Occasionally, intragenic exchanges are also seen between Synechococcus hosts. 
 
The authors compare their cultured results to those from wild phage sequences from the Pacific Ocean and see that much of the natural diversity is similar to the sequences from the cyanophage isolates, despite their origination from different ocean basins. Overall, therefore, a considerable amount of genetic shuffling takes place within these two PSII genes in cyanophages, and this creates a reservoir of photosynthetic diversity from which both host and phage are likely to benefit. This study offers a compelling example of global-scale microbial and phage co-evolution that likely influences the biological success of these prolific marine organisms.

The development of simultaneous resistance to structurally unrelated drugs is a major obstacle to chemical treatment of numerous cancers. In vitro studies performed in tumour cell lines resistant to antitumour drugs showed that several interrelated mechanisms could be involved in the resistance process (Pastan & Gottesman, 1987;Moscow & Cowan, 1988;Endicott & Ling, 1989;D'Arpa & Liu, 1989;Beck, 1989;Giovanella et al., 1989;Tew, 1989). For example, a P-170 kDa transmembrane glycoprotein coded by a multidrug resistance (MDRI) gene is present at high levels in cell lines that are cross-resistant to anthracyclines and Vinca alkaloids (Pastan & Gottesman, 1987;Moscow & Cowan, 1988;Endicott & Ling, 1989). P-glycoprotein acts as a drug efflux pump regulating intracellular accumulation of these drugs. DNA topoisomerases I and II are nuclear enzymes important for solving topological problems arising during DNA replication and transcription (D'Arpa & Liu, 1989). These enzymes have been shown to represent important targets for a variety of anticancer drugs and to play a part in resistance process of cancer cells (Beck, 1989;Giovanella et al., 1989). Another mechanism of action involving glutathione S transferase was described in cancer cells resistant to drugs as different as cis-platinum, alkylating agents and anthracyclines (Tew, 1989). The anionic isoenzyme glutathione S transferase (GSTx) which belongs to a complex group of drug-detoxifying enzymes, was present at high levels in the MCF7 breast cancer cell line resistant to Doxorubicine (DXR) together with high levels of MDRI transcripts (Batist et al., 1986;. Although the precise role of GST7 in the development of resistance is not known, its expression may be one among the numerous phenotypic and biochemical changes which accompany drug resistance in some cancers. Patients with invasive carcinoma of the uterine cervix are treated by surgery, radiation and/or chemotherapy (Friedlander et al., 1983;McGuire et al., 1989). Until now, results obtained with chemical agents suggest that cervical cancers may exhibit a resistance phenotype. Therefore, the aim of this study was to analyse the expression of the GSTi gene in invasive cervical cancers of different clinical stages and of different histological types in comparison with normal cervical epitheliums. Such studies may contribute to a better comprehension of the failure of drug therapy in cervical cancers.

Materials and methods
Cervical cancers and normal cervical epitheliums One hundred and six specimens of invasive cervical cancer were obtained by biopsy or rarely by surgical excision from untreated patients (93 primary tumours and 5 node metastases) and from treated patients (six recurrent tumours and two liver metastases). These treated patients have received external beam radiotherapy (RX) and/or drug regimen (DR) composed of cis-platinum, Methotrexate, Chlorambucil and Vincristine (Table I). Three specimens of carcinoma in situ (CIS) were also obtained at surgery. Specimens of normal cervical epithelium were obtained from 18 patients treated by surgery for fibroma of the uterine corpus and from six patients with adjacent primary cervical cancers. Tumour and cell samples were immediately stored in liquid nitrogen.
DNA and RNA preparations DNAs and total RNAs were prepared from the same tissue samples (about 100 mg) by the guanidinium isothiocyanate-CsCl gradient method (Maniatis et al., 1982;Sheng et al., 1990). Briefly, tissues were ground in liquid nitrogen, then lysed in the guanidinium-isothiocyanate buffer. Lysate was layered onto a 5.7 M CsCl cushion and submitted to a centrifugation at 37 000 r.p.m. for 17 h at 20°C (SW55 Rotor Beckman Ultracentrifuge model L5). DNA was collected from the supernatant, dialysed and treated with proteinase K. After deproteinisation by phenol-CHCI3, DNA was precipitated by absolute ethanol. DNA preparations in solution in appropriate buffers, were incubated with HindIII restriction endonuclease and the digest products analysed by Southern blot hybridisation under stringent conditions using human DNA probes.
Total RNA was collected at the bottom of the centrifuge tube, solubilised in Tris EDTA, 0.1% SDS and precipitated by absolute alcohol. RNA in solution in suitable buffer was then incubated with 2 fig ml-' of DNAase RNAase-free (Sigma) for 60 min, at 37°C. Denatured RNA samples (10 ,.g per well) were fractionated on a formaldehyde 1.2% agarose gel and transferred to a Hybond C extra filter and analysed by Northern blot hybridisation. The quality of the RNA was verified by the integrity of the 28S and 18S ribosomal bands coloured by ethidium bromide.  Figure 3. bMetastases, five pelvic lymph node metastases and two liver metastases. The two liver metastases were obtained after patients were treated for primary tumour by vincristine (VCR) or external beam + radiotherapy (RX) and chemotherapy (DR). cLocal recurrences (6 cases) after therapeutic failure (DX + DR). dTwo carcinoma specimens were obtained from the same patient in four cases. eAnaplastic cell carcinoma was of stage III and sarcoma of stage I.
(Schleicher & Schuell). Hybridisation were performed in stringent conditions with the appropriate denatured human probes 32P-labelled by nick-translation (about iO0 c.p.m.). Filters were exposed for various periods of times to Kodak XAR5 films. The cervical cancer cell line CaSki (American Type Culture Collection) was used to quantitate the levels of GSTir mRNAs. The GSTi transcript level in this cell line was stable and arbitrarily considered as the basic level (one unit). GSTIE mRNA levels were determined by densitometer scanning of the autoradiographic bands (Chromoscan 3, Joyce Loebl). In order to provide a control for the amount of RNA on the filters, the GSTn gene signal was removed and the same filters were rehybridised with an actin probe.  (Batist et al., 1986).

Statistical analysis
The Student test was performed for comparison of mean values, and the Chi-square test for other correlations.

Results
Analysis of total RNA for GSTrc transcripts Using the GSTi probe, a 0.7 kb transcript band was observed in the 51 specimens of invasive carcinoma as well as in the 12 specimens of normal cervical epitheliums which were analysed by Northern blot hybridisation (Figure 1). The analysis of total RNAs from the breast cancer cell lines MCF7 sensitive (MCF7/p) and resistant to doxorubicin (MCF7/DXR) provided negative and positive controls for GSTir mRNA as previously described (Batist et al., 1986) ( Figure 1). The transcript levels given in arbitrary units were quantitated by slot blot hybridisation relatively to the GSTir mRNA level found in the uterine cervix carcinoma cell line, CaSki, as described in Materials and methods. A representative slot blot is shown in Figure 2. The values of the GSTr mRNA levels found in individual cervical specimens are scored in Figure 3. High GSTt mRNA levels were found in invasive squamous cell carcinomas (mean: 2.6 ± 1.5) while low levels were detected in normal cervical epitheliums (mean: 0.7±0.1) (Student test P<10-a). Levels were also low in anaplastic cell carcinoma and sarcoma (0.5 unit) ( Table I). Intermediary GSTi mRNA levels were found in adenocarcinomas (1.7 ± 1.3) and carcinomas in situ (1.3 ± 0.8) ( Table I). An overexpression (level superior to 1 unit) of the GSTn gene was observed in 89/106 (84%) invasive cancers while observed only in 1/24 (4%) normal cervical epithelium (P< 10-4). As shown in Figure 3, the frequency of invasive cancers exhibiting an overexpressed GSTi gene was significantly higher in squamous cell carcinomas (P< 0.01) than in cancers of other histological types (adenocarcinomas, anaplastic cell carcinoma and sarcoma). However data showed that overexpression was not found to be different in early stages (I and II) and in advanced stages (III and GSTr -0.7 kb Figure 1 Northern blot analysis of total RNA (10 gg in each well) for analysis of GSTin transcripts. CaSki, MCF7/p, MCF7/ DXR cell lines, normal cervical epitheliums (NC), invasive squamous cell carcinomas (T, primary tumour). No GSTi transcript was detected in MCF7/p and a high level was found in MCF7/DXR as previously described (Batist et al., 1986). Upper panel represents the transferred blots after the agarose gel was coloured with ethidium bromide (EB). Exposure time to Kodak XAR5 films was days. Three pairs of specimens obtained from the same patients were shown; NC-T (0.5 and 1.2 units respectively); T-M (3.5 and 8.4 units respectively); T-LN (2.9 and 2.5 units respectively). Filter was dehybridised and rehybridised using actin probe. Exposure time to Kodak XAR5 film was two days for both signals.
IV) ( Table I). The GSTIC mRNA levels were not found to be significantly higher in recurrent tumours and lymph node metastases than in primary tumours. However, in the case of one liver metastasis treated with VCR, the GSTi mRNA level was found higher (8.5 units) than that observed in the ed (data not shown). No evidence for amplification of the GSTc gene was found in the 49 DNA preparations analysed including those from tumours where the gene was overexpressed. A previous immunohistochemical study (Shiratori et al., 1987) had suggested that the human papillomavirus (HPV), a virus most likely involved in the carcinogenesis of cervical cancers (Zur Hausen, 1989), could favour the production of GSTc in precancerous cervical lesions. Therefore we have analysed the GSTc mRNA level in relation to the presence of HPV DNA sequences in tumours and normal cervical epitheliums. Of the 96 invasive cervical cancers for which both HPV detection (Riou et al., 1990a) and GSTi analysis could be done, no difference in GSTi mRNA levels was found between HPV-positive (85% of tumours) and HPVnegative tumours. For epidermoid carcinomas, the mean levels of GSTi mRNA were 2.6 U for both HPV-positive and HPV-negative tumours. For adenocarcinomas these mean levels were 1.6 and 2.2 for HPV-positive and HPVnegative tumours respectively. Moreover the GSTic mRNA level was not found to be significantly higher in the four HPV-positive than in the 11 HPV-negative normal cervices. The normal cervical epithelium which displays more than 2 units (Figure 3) was HPV-negative.

Analysis of tumoral DNA for GSTit gene amplification
Forty-nine preparations of genomic DNA from cervical cancers (40 samples) and normal cervices (nine samples) were analysed by Southern blot hybridisation using GSTic probe. As expected, two DNA bands of 6.2 and 5.0 kb were observ-Cervical cancers are treated by surgery, radiation and cytotoxic drugs (Friedlander et al., 1983;Haie et al., 1988;McGuire et al., 1989). Treatment depends on prognosis which is determined by clinicopathological parameters of which the most important are clinical stage at diagnosis and nodal status (Pejovic et al., 1981). However in most cases, cervical cancers respond poorly to chemotherapy. Therefore, the present study was designed to test whether the GSTt gene, suspected of involvement in drug resistance, was overexpressed in cervical cancer cells.
Using Northern blot hybridisation techniques we detected in the cervical tissues the expected 0.7 kb GSTc transcript band (Figure 1). Transcripts were easily detectable in all the specimens analysed. When normalised to actin mRNA, GSTi transcript levels in squamous cell carcinomas were found to undergo variations (range 0.5 to 8 units), but no significant difference was observed between cancers of different clinical stages suggesting that GSTi is not associated with the progression of cervical cancers. In one liver metastasis the GSTic mRNA level was found to be higher than in the primary tumour. This could however be due to the presence of normal tissue in the primary tumour. A significant difference of the frequency of tumours with GSTic overexpression (P<0.01) was found between squamous cell carcinomas and cancers of other histological types (adenocarcinomas, anaplastic cell carcinoma and sarcoma).
Our data confirmed previous immunohistochemical studies (Shiratori et al., 1987) showing a GSTn expression in about 90% of invasive cervical cancers. However they differ from those of Shiratori et al. (1987) since the presence of GSTx transcripts was detected in all invasive cervical cancer as well as in all normal cervical epitheliums. Moreover we show that the presence of HPV DNA sequences in invasive cervical cancers and in normal cervical epitheliums does not influence the GSTn mRNA level.
The overexpression of the GSTi gene found in most invasive cervical cancers indicates that this gene is associated with the process of carcinogenesis. Such gene activation is more prevalent in squamous cell carcinomas than in adenocarcinomas. This is in accordance with the fact that squamous cell carcinomas are usually less sensitive to cytotoxic drugs than cancers of other histological types. In previous studies we have shown the presence of MDRI transcripts at low, but significant levels in 43% of invasive cervical cancers and in 68% of normal cervical epitheliums (Riou et al., 1990b) suggesting that expression of this gene may also be involved in the drug resistance phenotype of certain cervical cancers.
In conclusion, it is most likely that several mechanisms are involved in the drug resistance of invasive cervical cancers. The presence of high GSTic mRNA levels may be a consequence of multiple biochemical alterations which accompany carcinogenesis and indirectly lead to a drug resistant phenotype.
We would like to thank Dr K. Cowan (Medicine Branch, NCI, Bethesda, USA) for his generous gifts of GSTr cDNA probe and of MCF7/DXR cell line and Dr G. Orth (Laboratoire des Papillomavirus, Institut Pasteur, Paris) for data on HPV.
Supported by Association pour la Recherche sur le Cancer (ARC, Villejuif, France), Institut Gustave Roussy (Contrat de Recherche Clinique 90D1O) and F6d6ration Nationale des Groupements des Entreprises Fran9aises dans la Lutte contre le Cancer.