Comparison of the Post-transcriptional Regulation of the mRNAs for the Surface Proteins PSA (GP46) and MSP (GP63) ofLeishmania chagasi *

MSP (GP63) and PSA (GP46) are abundant 63- and 46-kDa glycolipid-anchored proteins on the surface of the promastigote form of most Leishmania species. MSP is a zinc metalloprotease that confers resistance to host complement-mediated lysis. PSA contains internal repeats of 24 amino acids, and its function is unknown. The steady state levels of mRNAs for both glycoproteins are regulated post-transcriptionally, resulting in about a 30-fold increase as Leishmania chagasi promastigotes growin vitro from logarithmic phase to stationary phase. Previous studies showed the 3′-untranslated regions (3′-UTRs) of these mRNAs are essential for this post-transcriptional regulation. These two 3′-UTRs of 1.0 and 1.3 kilobases were cloned immediately downstream of a β-galactosidase reporter gene in a plasmid, and segments were systematically deleted to examine which portions of the 3′-UTRs contribute to the post-transcriptional regulation. The 92-nucleotide segment of greatest similarity between the two 3′-UTRs was deleted without loss of regulation, but the segments flanking this similarity region have positive regulatory elements essential for the regulation. We propose that similar, but non-identical, molecular mechanisms regulate the parallel expression of these two L. chagasimRNAs despite their lack of sequence identity. These post-transcriptional mechanisms resemble the mechanism recently suggested for the regulation of mRNAs encoding the dipeptide (EP) and pentapeptide (GPEET) repeat proteins in Trypanosoma brucei that involves interactions between positive and negative regulatory elements in the 3′-UTR.

Protozoan parasites of the genus Leishmania cause a diverse group of diseases collectively called leishmaniasis that range in severity from spontaneously healing cutaneous ulcers to potentially fatal visceral disease. During their life cycle the Leishmania sp. exist as two developmental stages, i.e. as extracellular promastigotes in the gut of the sandfly vector and as intracellular amastigotes in the phagolysosome of mammalian macrophages. Glycoproteins on the surface of the organism play important roles in its survival in both of these environ-ments. The two best characterized Leishmania surface glycoproteins are the major surface protease (MSP, 1 also called GP63 for 63-kDa glycoprotein) and the parasite surface antigen (PSA, also named GP46 for 46-kDa glycoprotein). Although these proteins have historically been called GP63 and GP46, the Nomenclature Working Group for Protozoan Parasites has recommended that protozoan proteins be assigned 3-6-letter names (1), so we will use the nomenclature of MSP and PSA here. Immunization with recombinant versions of either of these proteins or their genes via DNA vaccines provides experimental animals with partial protection against Leishmania challenge (2)(3)(4)(5)(6).
Leishmania MSP is a family of closely related zinc metalloproteases that have been found in different reports to (i) confer resistance of promastigotes to complement-mediated lysis (7), (ii) promote attachment to and internalization of promastigotes by host macrophages (8), and (iii) facilitate the intracellular survival of amastigotes in phagolysosomes of host macrophages (9). When virulent promastigotes develop during growth in culture from the less infectious logarithmic phase to the highly infectious stationary form, an 11-30-fold increase in MSP expression occurs (10 -12). In Leishmania chagasi, which causes visceral leishmaniasis in Latin America, MSP is encoded by more than 18 genes (MSPs) that fall into three classes on the basis of (i) the stage at which they are expressed in the life cycle and (ii) unique sequences in their 3Ј-untranslated regions (UTRs) and intergenic regions (IRs) (13). In virulent promastigotes, 3.0-kb MSPS RNAs are expressed in stationary (S) phase but not logarithmic phase of growth, whereas 2.7-kb MSPL RNAs are expressed during logarithmic (L) but not stationary phase. MSPC RNAs of 2.6 and 3.1 kb are constitutively (C) expressed at low levels in both logarithmic and stationary phase (14).
PSA is another family of closely related proteins that have been detected in all Leishmania species examined except for members of the Leishmania braziliensis complex (15)(16)(17)(18). All reported nascent PSA sequences contain hydrophobic aminoand carboxyl termini that are likely cleaved during translocation of the protein across the endoplasmic reticulum and its linkage to a glycolipid anchor. The function(s) of PSAs is not known, but they possess 3-13 internal leucine-rich repeats of 24 amino acids that have been shown in other proteins to participate in protein-protein interactions (19). The organization of the PSA genes (PSAs) has not been fully characterized in any Leishmania species, but in those that have been investigated the multiple non-identical PSAs occur in clusters (18 -20). In L. chagasi promastigotes, expression of the 2.8-kb PSA mRNA parallels that of 3.0-kb MSPS mRNA during growth in vitro. The steady state levels of both RNA species increase more than 30-fold as the promastigotes develop from the less infectious, logarithmic form to the highly infectious, stationary form (20). Stationary promastigotes have approximately equal amounts of the PSA and MSP RNAs, which together constitute 2-3% of the total mRNA in the cell.
Leishmania and other trypanosomatids do not appear to have promoters for RNA polymerase II, which transcribes protein-encoding genes, even though transcription of these genes is sensitive to ␣-amanitin as it is in other eukaryotes (21). Instead, these genes are constitutively transcribed from large gene clusters, and the steady state levels of their mature mRNAs are regulated post-transcriptionally by mechanisms that often involve their 3Ј-UTR sequences (22)(23)(24)(25)(26)(27)(28). Because the abundance of the PSA and MSPS RNAs in L. chagasi promastigotes are regulated in parallel, we inspected their 3Ј-UTRs to see if their 1.0-kb (MSP) and 1.3-kb (PSA) 3Ј-UTRs contain sequences in common. The greatest similarity between these two 3Ј-UTRs is a 92-nucleotide segment with 66% identity. We found, using a ␤-galactosidase reporter gene, that neither this 92-nucleotide 3Ј-UTR segment nor the downstream IR between the tandem MSPs or PSAs contributes directly to the regulation of these two RNA species. Therefore, we generated systematic deletions of other segments of the two 3Ј-UTRs. We discovered that the regions immediately flanking this 92-nucleotide segment are involved in regulating the levels of both the MSPS and PSA mRNAs through similar, but non-identical, mechanisms. These mechanisms have features in common with a recently proposed model for the regulation of the Trypanosoma brucei genes for EP and GPEET, the most abundant proteins on the surface of the insect form of African trypanosomes (29 -32).

EXPERIMENTAL PROCEDURES
Parasites-A strain of L. chagasi (MHOM/BF/00/1669) was originally isolated from a Brazilian patient with visceral leishmaniasis. Virulent parasites were maintained in golden hamsters, and amastigotes were isolated from infected hamster spleens. Amastigotes convert into promastigotes when cultivated in vitro at 26°C in hemoflagellate-modified minimal essential medium supplemented with 10% heat-inactivated fetal bovine serum and 5.6 g hemin/ml at pH 7.4 (33). For most experiments, promastigote cultures were seeded at 1 ϫ 10 6 parasites/ ml, and stationary phase was achieved in 7-8 days. Logarithmic and stationary phase were defined according to concentration and morphologic changes as described (34). Promastigote cultures used for stable transfections were used within 3 weeks of isolation from a hamster. The medium used for selection and maintenance of stably transfected promastigotes contained 50 g of G418/ml (Invitrogen). All frozen stocks of promastigotes were prepared within two passages in liquid media after clonal isolation to avoid the effects of attenuation due to long term serial passaging in culture.
For subsequent analyses, aliquots of ϳ1.5 ϫ 10 8 promastigotes were removed either daily or during logarithmic (day 3) and stationary (day 7) phases of growth. Parasites were washed in sterile Dulbecco's phosphate-buffered saline (1ϫ) (Invitrogen) 3 times by centrifugation for 5 min at 4000 ϫ g, then resuspended in 1.4 ml of phosphate-buffered saline (1ϫ). The final cell pellet was frozen in a dry ice/ethanol bath and stored at Ϫ70°C. Cells for northern, Southern, and enzymatic analyses were harvested separately from the same tissue culture flask for each stage of growth. Cell densities were measured to confirm the stage of growth because the duration of the lag phase and the growth rates vary slightly from experiment to experiment. Data collected from each cell line represent multiple transfections with the same plasmid to minimize variability due to the particular condition of the parasites isolated from different hamsters.
Plasmid Constructions-The Leishmania expression vector, pX␤GAL2, was kindly provided by Stephen Beverley (36). In earlier studies, plasmids were constructed in which the corresponding 3Ј-UTRs and IR regions of the three MSP gene classes, i.e. MSPS, MSPL, and MSPC, were cloned at the XbaI site downstream of the ␤GAL-coding region in pX␤GAL2 (36,37). A fragment containing the 3Ј-UTR and IR of an L. chagasi PSA gene, called PSAA, was isolated by NotI digestion of a genomic DNA phage clone and ligated into a NotI site downstream of the ␤GAL gene in pX␤GAL2. Recombinant constructs containing mutations in these 3Ј-UTRs were initially constructed in pBluescript, after which the mutant 3Ј-UTRs were gel-purified and cloned downstream of ␤GAL in pX␤GAL2. Nucleotide sequences and orientations were confirmed by DNA sequencing.
The plasmids used in the 3Ј-UTR and IR swapping experiments ( Fig.  1) and the 3Ј-UTR deletion experiments (Figs. 2, 4, 6, and 8) were derived from the above-described plasmids containing the full PSAA and MSPL 3Ј-UTR and IRs using spliced overlap extension (SOE) PCR (38) followed by appropriate restriction enzyme digestion and ligation. The SOE PCR strategy facilitated the specific deletion of nucleotides from large plasmids in which removal by restriction enzyme digestion was limited. The primers used for the deletion constructs are shown in Figs. 3 and 7. Some primers were designed to create a unique AvrII site (5Ј-CCTAGG) at the site of deletion to facilitate the insertion of nonleishmanial DNA sequences.
Replacement constructs were prepared in which non-leishmanial linker DNA sequence and/or wild type sequence was inserted at the location of the deletion in the deletion constructs. The inserted nonleishmanial DNA sequence was selected to conserve length and GC nucleotide content. The template for these PCR reactions was the pBluescript polylinker region. PCR products were digested with AvrII and ligated into this unique restriction site in the pX␤GAL2 construct at the deletion. Constructs A⌬3A and A⌬3B (Fig. 4) were generated from A⌬3 by constructing a fusion between part of the wild type region 3 sequence and the non-leishmanial sequence (pBluescript). The resulting recombinant fragment was then ligated into A⌬3 at the unique AvrII site at the deletion. Deletions in the 3Ј-UTR of MSPS (Figs. 7 and 8) were made by the same SOE PCR procedures as those described above for the deletions in the 3Ј-UTR of PSAA.
Generation and Activity Assays of Stable Tranfectants-L. chagasi promastigotes were stably transfected and cloned according to the published protocol (33), a procedure that took an average of 2-3 months to obtain each stably transfected clone. Transfected cells were harvested, and cell pellets were resuspended in 100 ml of lysis buffer (100 mM KH 2 PO 4 , pH 7.8, 0.33% Triton X-100), lysed by three freeze-thaw cycles, sonicated for 5 min, and centrifuged for 5 min at 10,000 ϫ g. The supernatant was assayed for protein concentration (bicinchoninic acid reagent and assay, Pierce) and ␤GAL activity (using Galacton-Star chemiluminescent substrate, CLONTECH, Palo Alto, CA). Fluorescence was measured in triplicate in a Monolight 2010 luminometer™ (Analytical Luminescence Laboratory, San Diego, CA).
Data from each experimental clone were normalized to a control transfectant containing the parent plasmid, pX␤GAL2, to eliminate variability due to differences in isolates and growth conditions. Multiple clones from each transfection condition were used in assays to establish consistency among different clones.
Calculations and Statistics-␤GAL activities for each transfectant were recorded as the average of three readings at a 1:10 dilution of cell lysate as prepared above. Total protein concentrations were recorded in g/l. Relative fluorescence units (RFU) were calculated as (␤GAL activity/l of cell lysate assayed)/1000. RFU/g of total protein was then calculated as (RFU/l of supernatant)/(protein concentration in g/l).
RFU/g of protein for each transfectant was normalized to the RFU for the control transfectant containing the parent plasmid. The ratios of normalized stationary phase activity to normalized logarithmic phase activity were calculated for each transfectant. Figures contain the means Ϯ S.D. of these normalized RFUs and stationary/log (S/L) ratios. The latter are referred to as the mean normalized S/L ␤GAL values. Statistical comparisons were done using Student's t test with Sigma-Stat ® software (version 2.03, SPSS Inc.).

RESULTS
The 3Ј-UTR of PSAA (3Ј-UTR PSAA ), but not Its Downstream IR, Influences Expression of a ␤GAL Reporter Gene-We previously cloned the 3Ј-UTRs and downstream IRs of the three L. chagasi MSP gene classes, MSPS, MSPL, and MSPC, and of a specific L. chagasi PSA gene, PSAA, downstream of ␤GAL in the Leishmania expression plasmid, pX␤GAL2 (36), and exam-ined their effects on ␤GAL expression (20,37). Constructs of the different plasmids were transfected into virulent L. chagasi promastigotes, and cloned transfectants were grown from logarithmic to stationary phase in vitro. Samples were removed at different times of growth for determination of ␤GAL enzyme activities and ␤GAL RNA levels. These experiments showed that expression of ␤GAL mRNA and activity closely paralleled expression of the MSPS, MSPL, or PSA whose 3Ј-UTR ϩ IR was cloned downstream of ␤GAL. For example, when the 3Ј-UTR ϩ IR of MSPS or PSAA was after ␤GAL, the ␤GAL activities and RNA levels were low in logarithmic phase and steadily increased to 20 -30-fold higher as the recombinant promastigotes grew to stationary phase (20). In contrast, when the 3Ј-UTR ϩ IR of MSPL was inserted after ␤GAL, ␤GAL expression remained at a low basal level throughout promastigote growth even though the wild type MSPL RNA is expressed more highly in logarithmic than in stationary growth (20). Southern blots and nuclear run-on assays were used to show that in these cloned stable transfectants the ␤GAL gene copy number does not change during promastigote growth in culture under constant drug selection and that the pX vector sequences are constitutively transcribed (shown in Refs. 20 and 39).
To determine whether either the 3Ј-UTR or IR or both are the sequences responsible for the growth-associated regulation of PSAA, we first "swapped" the 3Ј-UTRs and IRs of PSAA and MSPL, as shown in Fig. 1. Plasmid constructs were made in which the ␤GAL-coding region in plasmid pX␤GAL2 was followed either by 3Ј-UTR PSAA ϩ IR MSPL or by 3Ј-UTR MSPL ϩ IR PSAA (see "Experimental Procedures"). These linearized constructs were stably transfected into virulent L. chagasi promastigotes. The ␤GAL activities and RNA levels in these cloned cells in logarithmic (3 days of growth) versus stationary phase (7 days of growth) were compared with the corresponding constructs containing ␤GAL followed by the complete 3Ј-UTR ϩ IR of either PSAA or MSPL (Fig. 1).
The ratio of ␤GAL activity in stationary versus logarithmic cells (S/L in Fig. 1) was about 30 when the 3Ј-UTR ϩ IR of PSAA was after ␤GAL. The ratio was unchanged when ␤GAL was followed by 3Ј-UTR PSAA ϩ IR MSPL . In each case, the 30fold increase in ␤GAL activity in stationary cells compared with logarithmic cells was accompanied by a corresponding increase in the ␤GAL RNA steady state level, as measured by instant imager analyses of Northern blots (summarized in the right-hand column of Fig. 1) and as demonstrated previously for PSAA (20). In contrast, when ␤GAL was followed either by 3Ј-UTR MSPL ϩ IR MSPL or by 3Ј-UTR MSPL ϩ IR PSAA , the S/L ratio of both ␤GAL activity and ␤GAL RNA was less than one. Thus, the 3Ј-UTR sequence appears to account in large part for the increased expression of PSAA and the lack of change in MSPL RNA during the logarithmic-to-stationary transition. In contrast, the downstream IR sequences do not appear to play a role. A further conclusion from both these results and earlier results (20,37) is that ␤GAL activity reflects the steady state level of ␤GAL RNA in these constructs containing downstream PSAA and MSPS sequences Deletion of a 92-Nucleotide Segment of Sequence Similarity in the 3Ј-UTRs of PSAA and MSPL Does Not Affect Gene Regulation-Previously we showed that the steady state levels of both PSAA and MSPS RNAs increase in parallel as virulent promastigotes grow from logarithmic to stationary phase (20). We also showed that, similar to the 3Ј-UTR PSAA results above, the increase in the MSPS RNA abundance is regulated primarily by elements in the 3Ј-UTR MSPS (37). We therefore inspected the 1.3-kb 3Ј-UTR PSAA and the 1.0-kb 3Ј-UTR MSPL for shared sequence elements ( Fig. 2A). The region of greatest similarity is a 92-nucleotide segment of 66% identity called the overlap or olp region. Two other small (Ͻ10 bp) regions of limited similarity are located upstream of olp. To see if this similar 92nucleotide olp segment in each of the two 3Ј-UTRs contributes to the parallel regulation of their RNAs, the 92 nucleotides were deleted from each 3Ј-UTR by the SOE PCR technique (see "Experimental Procedures"), and the resultant ⌬olp 3Ј-UTR ϩ IR was cloned immediately downstream of ␤GAL in pX␤GAL2 for subsequent stable transfections and ␤GAL activity measurements. Fig. 2B shows that deletion of this olp region from these two 3Ј-UTRs had little if any effect on the regulatory role of either 3Ј-UTR. The S/L ␤GAL activity ratio was about 30 Ϯ 3.5 when the 3Ј-UTR PSAA was present and about 46 Ϯ 16 when the 3Ј-UTR PSAA⌬olp was present, yielding a statistically insignificant p value of 0.155. Likewise, the S/L ␤GAL ratio was about 63 Ϯ 9 and 59 Ϯ 15 in the presence of the 3Ј-UTR MSPS and 3Ј-UTR MSPS⌬olp , respectively. Thus, by this targeted deletion analysis it appears the 3Ј-UTRs do not need the olp segment to up-regulate their RNA levels during growth to stationary phase. This conclusion prompted us to conduct a more systematic deletion analysis of the two 3Ј-UTRs.
Specific Segments of the 3Ј-UTR PSAA Are Involved in Gene Regulation-Because deletion of the olp segment did not abrogate logarithmic-stationary gene regulation, we used RNA secondary structure prediction programs to examine the 3Ј-UTRs of PSAA and MSPS for potential secondary structures that might provide clues about their involvement in regulation. Both 3Ј-UTRs are about 90% GϩCϩU, so unfortunately, when both G-C and G-U base pairing are allowed, the number, sizes, and complexities of possible hairpin loops in these 3Ј-UTRs of 1.0 and 1.3 kb are immense. Thus, these secondary structure analyses were not informative, even when smaller regions of the 3Ј-UTRs were examined (not shown). Therefore, the two 3Ј-UTRs were tested further for regulatory sequence elements by deleting ϳ200-bp segments across the entire 3Ј-UTR using the SOE PCR technique. In the case of the 3Ј-UTR PSAA , five adjacent segments were individually deleted (Fig. 3) and the constructs, called A⌬1-A⌬5, were cloned into pX␤GAL2 for stable transfection into virulent promastigotes and subsequent analyses (Fig. 4). The right-hand boundary of the deletion in construct A⌬5 was designed to occur 13 nucleotides upstream of the polyadenylation site to preserve this site. Point mutations were introduced into SOE PCR primers so that a unique restriction site, AvrII, would be present in recombinant constructs at the site of the deletion. To determine whether changes in ␤GAL expression were merely due to changes in spacing generated by deletions, the AvrII sites were used to insert non-leishmanial DNA (from pBluescript) of the same length and GC nucleotide content as the 200-bp deletion. This generated another series of recombinant constructs called A⌬1link-A⌬5link. The plasmid deletion constructs and their corresponding linker constructs were stably transfected into promastigotes, and the ␤GAL activities in logarithmic and stationary phase cells were analyzed. In every case the ␤GAL activities derived from a given 200-bp deletion construct and its corresponding "link(er)" construct were found to be equivalent within experimental error (not shown). Thus, Fig. 4 shows the data for only the deletion constructs.
Deletion A⌬1 caused a 2.7-fold increase in the S/L ␤GAL ratio compared with the wild type (wt) 3Ј-UTR PSAA , suggesting there may be a negative control element in this segment. Other deletions had no effect (A⌬2) or caused a decrease (A⌬3, A⌬4, A⌬5) in the S/L ratio. Deletions A⌬3 and A⌬4 exhibited the largest effects, i.e. decreases of 11-and 15-fold in the S/L ratio, respectively. Interestingly, these deleted segments flank olp, whose deletion had no effect (compare Figs. 2 and 4). Thus, sequences deleted in the A⌬3 and A⌬4 constructs possess a positive control element(s) that increases ␤GAL activity in stationary cells.
To further map putative positive regulatory elements, plasmid constructs A⌬3A and A⌬3B were generated to examine the individual effects of each half of segment 3. In each of these constructs, a non-leishmanial sequence was introduced to preserve the position of the wild type sequence within the 3Ј-UTR. To our surprise, the presence of either half of segment 3 did not restore wild type ␤GAL activity in stationary phase (Fig. 4). A similar half-deletion of segment 4 was not constructed.
The locations of the two nucleotide replacements in construct A⌬3 that were used to create the AvrII site (5Ј CCTAGG) at the site of the A⌬3 deletion are shown in Fig. 5. Insertion of a 200-bp linker sequence into this site did not significantly change the ␤GAL S/L ratio from that shown for A⌬3 in Fig. 4  (not shown). Surprisingly, however, re-insertion of the wild type segment 3 sequence into the AvrII site also did not restore the wild type S/L ␤GAL ratio (Fig. 6, compare A⌬3 and A⌬3wt). Because the only difference between the A⌬3wt and PSAA 3Ј-UTR constructs is the AvrII site, we tested whether the AvrII site itself might be responsible for this unexpected result. Three additional constructs were prepared (Fig. 6). To prepare construct A⌬3wt5Ј, the SOE PCR primers were designed to eliminate the AvrII site on the 5Ј side of segment 3 and retain it on the 3Ј side. In construct A⌬3wt3Ј, the AvrII site was eliminated on the 3Ј side of segment 3 and retained on the 5Ј side. Finally, in construct A⌬3Avr(Ϫ), segment 3 was deleted from the 3Ј-UTR PSAA by SOE PCR without generation of an additional AvrII site. In the resulting stable transfectants, the wild type S/L ␤GAL ratio was restored in construct A⌬3wt3Ј but not in construct A⌬3wt5Ј (Fig. 6). Thus, replacement of two nucleotides to generate an AvrII site at the segment 3-olp boundary of A⌬3wt5Ј (Fig. 5) is sufficient to cause loss of logarithmic-stationary regulation. However, when segment 3 was deleted without the concomitant insertion of an AvrII site in construct A⌬3Avr(Ϫ),the logarithmic-stationary regulation of ␤GAL was abrogated even more completely than the A⌬3  construct (20.6-fold decrease versus 11.3-fold decrease). Thus, the presence of the non-mutated sequence at the 3Ј-end of segment 3 alone is not sufficient to confer wild type regulation.
Because an AvrII site had also been engineered at the deletion site in the A⌬4 construct (Figs. 4 and 6), the segment 4 wild type sequence was also reinserted into this AvrII site to generate construct A⌬4wt. In this case and in contrast to construct A⌬3wt, the wild type S/L ␤GAL ratio was restored in A⌬4wt (Fig. 6). Thus, the presence of the AvrII site does not appear to alter regulation of ␤GAL expression in the A⌬4 constructs as it does with the A⌬3 construct. Similarly, in experiments not shown, transfectants containing plasmid constructs in which only each end of segment 4 is mutated to an AvrII site have wild type S/L ␤GAL ratios.
Specific Segments of the 3Ј-UTR MSPS Are Also Involved in Gene Regulation-A deletion analysis of the 1.0-kb 3Ј-UTR MSPS was also undertaken similar to the 3Ј-UTR PSAA analysis (Figs. 7 and 8). Four deletion constructs with deleted segments replaced by a single AvrII site, called S⌬1-S⌬4, were cloned into pX␤GAL2 and stably transfected into virulent promastigotes. Also similar to the PSAA analysis, in each case non-leishmanial DNA (from pBluescript) of the same size and GC content as the deleted segment was cloned into the AvrII site, producing a corresponding set of constructs called S⌬1link-S⌬4link. As was found with the 3Ј-UTR PSAA , the S/L ␤GAL ratio of a given deletion construct and its corresponding linker construct were found to be the same within experimental error, so only the deletion data are shown (Fig. 8).
Deletion S⌬1 had no effect on the ␤GAL S/L ratio compared with wild type, and S⌬2 resulted in only about a 2-fold change. In contrast, S⌬3 and S⌬4 caused 7.5-and 3.5-fold decreases in the S/L ratio, respectively. Thus, similar to the 3Ј-UTR PSAA data, deletions of the segments flanking the olp region exerted the largest effects, and in both cases, the deletions caused a decrease of the ␤GAL S/L ratio. Constructs S⌬3A and S⌬3B were also generated to examine the effects of each half of segment 3 (Fig. 8). Similar to the findings with A⌬3A and A⌬3B (Fig. 4), the presence of either half of segment 3 of the 3Ј-UTR MSPS did not restore the wild type S/L ratio. Thus, the positive regulatory element(s) extends across both halves of segment 3 of the 3Ј-UTR MSPS as it does in segment 3 of the 3Ј-UTR PSAA .
Because of the effects of the engineered AvrII site in some of the 3Ј-UTR PSAA constructs, the wild type segment 3 sequence was reinserted at the AvrII site of construct S⌬3 to generate construct S⌬3wt. Fig. 5 shows the three-nucleotide replacements that were generated during SOE PCR to create this AvrII site. In contrast to A⌬3wt (Figs. 4 and 6), insertion of wild type segment 3 sequence into the AvrII site of S⌬3 restored the wild type S/L ␤GAL ratio (S⌬3wt in Fig. 8). A similar restoration of wild type activity was obtained when the wild type segment 4 sequence was replaced into the AvrII site of S⌬4 (not shown). Thus, the nucleotide replacements used to create the AvrII site in these S⌬ constructs do not affect expression of the upstream ␤GAL gene.
Insertion of Region 3 from the 3Ј-UTR PSAA Did Not Restore Regulation in Construct S⌬3-Because deletions of the segments flanking the olp region in both 3Ј-UTR MSPS and 3Ј-UTR PSAA had similar effects, we tested whether these segments in one 3Ј-UTR could be replaced with the corresponding segments of the other 3Ј-UTR and still retain the regulation. Therefore, PSAA segment 3 was inserted into the AvrII site of construct S⌬3 to generate construct S⌬3-Awt3 (bottom of Fig.  8). In contrast to S⌬3wt, the wild type S/L ␤GAL ratio of the 3Ј-UTR MSPS was not restored in S⌬3-Awt3. Instead, the S/L ratio of the original S⌬3 dropped even further, i.e. from 7.5-to 22.2-fold. A similar result was obtained when PSAA segment 4 was inserted into construct S⌬4, i.e. the wild type S/L ␤GAL ratio dropped still further instead of being restored (data not shown). Therefore, in these two examples the regulatory region of one 3Ј-UTR could not replace the correspondingly positioned regulatory region of the other 3Ј-UTR, suggesting that these regulatory regions function only within the context of their own 3Ј-UTRs. DISCUSSION The purpose of the current work was to map sequences in the 3Ј-UTRs of two tandemly repeated gene classes of L. chagasi whose mRNAs are expressed at similar times in the growth cycle of the parasite. We hypothesized that similar molecular features would account for their similar patterns of expression. Our data revealed that the mechanisms regulating levels of MSPS and GP46A RNAs are likely to be complex, involving at least several regions of their 3Ј-UTRs. Furthermore, our data suggest that different features of each of these 3Ј-UTR sequences regulate gene expression.
Many differentially expressed trypanosomatid genes are regulated post-transcriptionally by molecular mechanisms involving their 3Ј-UTRs (21-28, 40 -42). The most extensively studied group of post-transcriptionally regulated genes in trypanosomatids is the T. brucei gene family encoding the related acidic repetitive proteins, EP and GPEET (previously called procyclic acidic repetitive protein or PARP (43)), found exclusively on the surface of the procyclic (insect) form of T. brucei (29,30). The coding regions of the EPs and GPEETs in the T. brucei genome are similar, but their short 3Ј-UTRs (300 bp) share only a conserved 26-mer sequence. The 100-fold higher steady state level of the EP and GPEET mRNAs in the procyclic form than in the bloodstream form is controlled mainly by elements in their 3Ј-UTRs (30,44,45). Deletion analysis of the EP1 3Ј-UTR (30,32) and characterization of its secondary structure by RNase digestion and lead hydrolysis (31) indicate that this 3Ј-UTR consists of three domains, I, II, and III. The 5Ј and 3Ј domains I and III, respectively, form independent stem-loop structures in the RNA, whereas the central domain II contains the conserved 26-mer as a singlestrand. Domains I and III both have positive regulatory elements, and it has been proposed that in procyclic trypanosomes one or more factors bind to these positive elements in the flanking stem-loops, shielding domain II from endonuclease degradation. In bloodstream trypanosomes, which presumably lack these positive regulators, the single-stranded domain II is exposed to endonuclease activity and quickly degraded (29,31).
Similar to T. brucei EP and GPEET, Leishmania PSA and MSP are the major surface glycoproteins on the insect form of parasite, and these Leishmania and T. brucei genes are posttranscriptionally regulated in a parallel manner via their dissimilar 3Ј-UTR sequences. The extent to which the patterns of cis-acting regulatory elements in the 3Ј-UTRs of their mRNAs resemble each other is intriguing. Similar to the EPs and GPEETs, PSAA and MSPS are regulated in a parallel fashion by their 3Ј-UTRs, yet these 3Ј-UTRs contain little sequence similarity. Likewise, similar to domains I and III of the EP1 3Ј-UTR, segments 3 and 4 of the PSAA and MSPS3Ј-UTRs contain positive regulatory elements that flank a conserved region (domain II in the EPs and olp in PSAA and MSPS). In the 3Ј-UTR PSAA , deletions of segments 3 and 4 result in an 11and 15-fold loss, respectively, in PSAA up-expression in stationary phase (Fig. 4). In the 3Ј-UTR MSPS , deletions of segments 3 and 4 cause a 7.5-and 3.5-fold drop, respectively, in MSPS up-expression in stationary phase (Fig. 8). In both of these Leishmania genes the positive regulatory element(s) in segment 3 could not be further localized by deleting just the 5Ј or the 3Ј half of segment 3. Similarly, the positive regulatory element(s) in domains I and III of the EP1 3Ј-UTR extend across the stem-loop in most of the domain (30 -32). The PSAA and MSPS data also suggest that the regulatory elements in segments 3 and 4 are both necessary, but neither is sufficient to confer full up-regulation of gene expression, again similar to domains I and III of EP1. It is not known whether possible hairpin loops in segments 3 and 4 of PSAA and MSPS play the same roles as the hairpin loops in domains I and III of the EP13Ј-UTR (31). The 200-nucleotide sequences of segments 3 and 4 have potential hairpin loops, as detected by RNA secondary structure prediction programs (not shown), but these sequences are too long for the predictions to be reliable. RNase digestion and lead hydrolysis experiments, similar to those conducted on the EP1 3Ј-UTR (31), will be necessary to clarify this question.
Unexpectedly, when the sequence of segment 3 was reintroduced into the engineered AvrII site in the A⌬3 deletion construct, wild type activity was not restored (Fig. 6). However, when the sequence at the 3Ј boundary but not at the 5Ј boundary of segment 3 was restored to wild type (i.e. without the AvrII site), activity was restored to wild type level. Thus, the wild type context at the 3Ј end of segment 3 is necessary for regulation of PSAA stationary phase expression. However, it is not a sufficient regulatory factor, since wild type regulation was not achieved by the presence of the wild type 3Ј end of segment 3 alone, as demonstrated by experiments with transfectants containing the A⌬Avr(Ϫ) deletion construct (Fig. 6). This A⌬Avr(Ϫ) construct shows that the two altered bases at the 3Ј end of segment 3 do not substitute for the entire segment 3. Furthermore, the positive effect of segment 3 and its wild type 3Ј sequence (Figs. 4 -6) was active only when the olp sequence was present. When the olp sequence was deleted, A⌬olp, replacement of the two bases to generate the AvrII site at the 3Ј end of segment 3 (Fig. 5) did not have the negative effect on regulation that is demonstrated by A⌬3wt and A⌬3wt5Ј (Fig. 6). These data are consistent with a model in which positive regulation by segment 3 results from shielding a negative element in the olp segment, such as a degradation signal. When the degradation signal is absent, as in the A⌬olp deletion construct, the protective effect of segment 3 is unnecessary, and wild type regulation is achieved. This is analogous to the involvement of domains I and II in the regulation of EP/GPEET expression.
When segment 4 of 3Ј-UTR PSAA is deleted, there is a 15-fold loss of regulation that can be restored by reintroduction of wild type sequence into the engineered AvrII site (Fig. 6), again similar in this case to the involvement of domains II and III in the regulation of EP/GPEET expression. Thus, segments 3 and 4 are necessary but not sufficient alone for regulation of PSAA gene expression. To determine whether these segments interact, a construct in which both segments are deleted will be necessary.
In summary, the regulatory effects of the 3Ј-UTR PSSA and 3Ј-UTR MSPS are clearly complex and multifaceted. Our data are consistent with a model in which segment 1 of the 3Ј-UTR PSAA contains a modest negative regulatory element, resulting in a 2.7-fold negative regulatory effect. Segments 3 and 4 each contain positive regulatory elements that appear to shield the olp region. Similarly, the 3Ј-UTR MSPS contains positive regulatory elements in segments 3 and 4 that flank the olp region, although there does not appear to be a weak negative regulator in segment 1 as there is in the 3Ј-UTR PSAA . Nucleotides at the boundary between segment 3 and olp appear to be critical for PSAA regulation, but we have no evidence for their involvement in MSPS regulation (Figs. 5, 6, and 8). The AvrII site in the 3Ј-UTR PSAA constructs was generated by two replacement mutations separated by a single base pair, both of which are in segment 3. The AvrII site in the 3Ј-UTR MSPS was generated by three replacements, one of which is in segment 3 and two of which are adjacent to each other three base pairs downstream in the olp sequence. Five additional adjacent nucleotides in segment 3 are shared between the 3Ј-UTR PSAA and 3Ј-UTR MSPS , as indicated by the diagonal lines in Fig. 5. All three replacements used to generate the AvrII site in the 3Ј-UTR MSPS disrupt identical nucleotides in the two 3Ј-UTRs, yet there was only an effect on the regulation of PSAA3Ј-UTR constructs. It will be worthwhile to determine whether these five nucleotides are important in the regulation conferred by the 3Ј-UTR PSAA but not 3Ј-UTR MSPS .
The similar features in the regulation of these Leishmania genes and those of the T. brucei EP/GPEETs support the possibility that there are common themes to the molecular mechanisms determining post-transcriptional regulation of trypanosomatid gene expression through sequences in their 3Ј-UTRs. Further elucidation of these regulatory mechanisms will require an even more detailed dissection of the 3Ј-UTR sequences and the potential proteins with which they interact than has been undertaken to date.  Fig. 4. The mean normalized ␤GAL S/L ratio was calculated as described in the legend for Fig. 4 and under "Experimental Procedures." For example, there is a 7.5-fold (or 62.5/8.37) decrease in the mean normalized S/L ratio in the S⌬3 transfectants compared with wild type.