While the majority of the genes identified in our analyses were specific to either the abdomen or the head/thorax samples, numerous changes were common to both tissue preparations. Twenty genes had decreased levels (including loj, see Additional file 1) in both mutant tissue samples, while 77 genes showed increased expression in loj mutants compared to controls (see Additional files 1 and 2). Seventeen of these 97 genes were found previously to be transcriptionally responsive to the ER stress inducing agent tunicamycin 43. Fourteen of our genes show a similar response to tunicamycin-induced changes, while 3 genes show the opposite pattern (see Additional files 1 and 2). Tequila, which is down regulated in loj mutants (see Additional file 1), is also repressed in tissue culture cells in which the UPR has been induced by dithiothreitol treatment; this decrease is dependent upon Ire-1 but not Xbp1 14.

Of the genes identified in both tissue data sets, 36 are known or likely Drosophila immune-regulated genes (DIRGs), the majority of which are targets of Toll or Imd signaling (see Additional file 2). GADD45β is an ER stress response gene that is regulated by NF-κB and exerts an anti-apoptotic effect by down regulating JNK signaling 44. Also up regulated in this common data set are genes encoding small immune peptides such as Cecropins and Attacins as well as the gram-positive bacteria peptidoglycan receptor PGRP-SA (see Additional file 2). The NF-κB gene Rel also has increased transcript levels in the loj mutant tissues.

To validate our microarray data, we used qPCR to test a subset of genes with altered expression profiles in both tissue samples. Each tissue was assayed independently and showed the expected directional change (see Methods and Additional files 1 and 2). For most of the genes tested, the gene expression levels are significantly different between loj mutants and controls in both tissues (p < 0.05). In only one instance (GstD5) was the difference non significant for either tissue.

We anticipated that increased expression of immune responsive and NF-κB target genes would be a general response to p24 gene mutations in multi-cellular animals. In Drosophila, mutations in two other p24 genes, eclair (eca) and baiser (bai), cause oviposition defects similar to those observed for loj mutants 2745. Both eca and bai are co-expressed with loj in ovarian follicle cells and are expressed in the CNS 28.

We generated viable eca mutant females and tested them for increased expression of four of the immune regulated genes that are up regulated in loj mutants (See Methods). Expression of CG6687, Tsp42Ed and Frost is significantly increased in eca mutants relative to controls, while Socs36E is not significantly increased in eca mutants (data not shown). These results indicate that activation of NF-κB and immune-regulated genes may be a general response to p24 loss-of-function mutations.

The egg tissue of the Drosophila abdomen expresses Loj throughout all stages of development, with particularly strong expression observed at stage 10 in the somatically-derived follicle cell layer 2728, whose primary function is secreting factors needed for eggshell formation. Low-level Loj expression is also observed in the nurse cells 2728 that produce mRNAs and other factors necessary for embryonic development. Our recent analysis indicates that the Loj protein is highly expressed in other adult abdominal tissues as well, including the fat body and the gut (K.A. Boltz, S. Grady, and G.E. Carney, unpublished results).

We identified a set of 366 genes with altered expression patterns in loj mutant abdominal tissue. Of these loci, 269 were noted only in the abdomen while the remaining 97 were also detected in the head/thorax samples (Table 1; see Additional files 3 and 4). Forty-seven of the 269 abdomen-specific genes were previously determined to be immune responsive. Most are transcriptional targets of Toll or Imd signaling while others are components of JNK signaling pathways (see Additional file 3). General immune-responsive genes are overrepresented in this data set (p = 4.915 × 10^-17, Fisher's Exact Test) as are components of JNK signaling (p = 9.192 × 10^-5, Fisher's Exact Test).

The abdominal tissue also differentially expressed 222 genes with no known function in immunity (see Additional file 4). Most of the published genome profiling experiments that identified immune response genes used an earlier version of the Drosophila genome array 2930313235 that lacked many of the transcriptional units on the Drosophila Genome 2.0 chips used in this study. Therefore, some of the genes in Additional file 4 likely have previously unrecognized roles in the immune response.

Genes in the up-regulated class have functions consistent with involvement in a UPR response (see Additional file 4). CG14207 encodes a small Hsp20-like chaperone. CG33486 encodes an asparagine synthetase, a protein previously implicated in the mammalian ER stress response 46. The predicted function of CG4415 is unfolded protein binding, suggesting it functions in the UPR as well. Many other up-regulated genes encode proteins with predicted functions such as proteolysis and peptidolysis, metabolism and oxidoreductase activity.

A subset of cells in the CNS also has high-level Loj expression, and CNS expression of loj in loj mutants rescues egg laying and fertility 28. We separated the loj CNS effects from those occurring in developing eggs and gut by dissecting the abdomen (containing the ovaries and gut but not CNS tissue) away from the remainder of the fly.

We identified a set of 275 genes differentially regulated in these samples compared to those from abdomens (Table 1; see Additional files 5 and 6). Many up-regulated genes function in processes such as lipid and protein metabolism. Again, approximately 20% of the genes have been implicated in immune signaling (Table 2; see Additional file 5). Immune responsive as well as JNK pathway components are also overrepresented in the head/thorax preparations (Immune gene p = 1.594 × 10^-19; JNK p = 3.110 × 10^-3; Fisher's Exact Test).

The Toll and Imd pathways activate different, but sometimes overlapping, immune-responsive genes depending upon the pathogen 3247. In adult flies, the NF-κB protein Dif is the Toll signaling effecter molecule in response to immune challenge, while the adult function of Dl is not clear. Rel functions downstream of Imd signaling to activate DIRGs [Reviewed in 48]. The fact that we observed increased message levels of signaling molecules that lie downstream of each pathway begs the question as to whether the three Drosophila NF-κB proteins are activated in the loj mutant. Therefore, we tested for genetic interactions between loj and the three Drosophila NF-κB genes to identify which may be involved in the stress response.

Since loj is expressed throughout development 27, we anticipated that activation of the ER stress response in loj mutants would be important for survival to adult eclosion from the pupal case. If NF-κB proteins are needed for stress response activation to survive pupation, we expected that reducing NF-κB levels in the loj mutant would affect adult survival. When we assayed eclosion rates in double mutant combinations of loj with Dif, dl or Rel, we found that fewer than the expected number of Dif; loj or loj, Rel mutant offspring emerged from the pupal case (Table 3). These results suggest that the Dif and Rel proteins are needed in loj mutants to ameliorate the stress response so that animals can survive development.

We discovered transcript-level changes in genes encoding factors involved in protein metabolism and folding that are consistent with induction of the UPR. However, we did not identify all of the genes found in a previous study on tunicamycin-induced ER stress in Drosophila 43, which identified approximately 600 transcriptional changes in untreated compared to tunicamycin-treated flies using a 1.5-fold change threshold. Tunicamycin activates the UPR as well as the EOR 13.

In yeast, loss of p24 function results in increased splicing of the Ire-1 target Xbp1 26. We did not find evidence for increased transcription of Xbp1 in our p24 mutant microarray experiment, which is consistent with the Girardot et al. report (2004) showing that the UPR transcriptional target Xbp1 is not up regulated in tunicamycin-treated flies. Furthermore, a second RT-PCR-based assay did not reveal increased splicing of Xbp1 in loj compared to control animals (data not shown) nor did we observe activation of Xbp1-EGFP 16 expression in loj mutants. The fact that increased Xbp1 transcription and splicing are not detected could be because these responses occur prior to the time point of our assays. Alternatively, some aspects of the tunicamycin and p24 mutant stress responses may be independent of Xbp1 signaling. Activated Atf6 increases expression of Xbp1 during ER stress 49, but Atf6 signaling has not been implicated in NF-κB activation during the EOR.

Our study identified 641 expression changes, 92 of which were demonstrated previously to be tunicamycin sensitive. Sixty-five transcripts had the same directionality as in Girardot et al. (2004) 43, while the remaining 27 show opposite patterns (see Additional files 1, 2, 3, 4, 5, 6). The tunicamycin experiment differed from ours in that the researchers used wild-type Drosophila adult male bodies and probed an earlier version of the Drosophila genome array with lower transcriptional coverage. Some of the differences in the two studies may be due to sex, genetic background and assayed transcripts. Additionally, the loj mutants should be chronically stressed due to loss of p24 function throughout development; the tunicamycin-treated animals experienced an acute stress response since they were assayed 24 hrs after being placed on sugar medium containing tunicamycin 43. Therefore, we expect that many differences in the two studies are due to differential effects from chronic compared to acute stress responses.

Our most interesting finding is the large number of immune-responsive genes with altered transcriptional profiles in the loj p24 mutant. Since the majority of currently published studies on Drosophila immune-responsive genes used the Affymetrix version 1 Drosophila Genome Arrays (based upon Berkeley Drosophila Genome Project v4.0), they could not identify all immune-regulated loci. Based upon their predicted functions, immune-regulated candidates from our study include Cht4 (see Additional file 1), ImpL2 and CG33093 (see Additional file 4).

Recognizing that increased DIRG message levels could be explained by activation of one or more Drosophila NF-κB proteins, we tested for genetic interactions between loj and Dif, dl or Rel. We found that Dif²; loj and loj⁰⁰⁸⁹⁸, Rel^E20 mutant combinations dramatically reduced adult eclosion, while effects of the dl¹; loj⁰⁰⁸⁹⁸ double mutant combination did not differ from dl¹; loj⁰⁰⁸⁹⁸/+ or dl¹/+; loj⁰⁰⁸⁹⁸. These results indicate that NF-κB and loj mutations interact with one another. The results suggest that the loj mutant stress response is needed for loj mutant animals to survive development and that this response is partially mediated by Dif and Rel. It is possible that the two mutations independently make a "sick" fly worse off. However, Dif and Rel are not required for viability, and neither dl nor imd mutations enhance the loj mutant phenotype, although both genes affect viability in loj/+ animals.

In our study we did not identify expression changes in all of the known immune-regulated genes. If both Dif and Rel are activated in the loj mutant, one might expect that all potential NF-κB target genes should be affected. However, NF-κB target sites in immune-responsive promoters are variably responsive to NF-κB and can be modulated by other factors to provide specialized immune responses 50, possibly in a tissue-specific manner. Therefore, the loj mutant may not activate all possible NF-κB targets because other cellular factors are needed for a strong response.

In mammals two potential mechanisms for ER stress-induced NF-κB activation have been described. The first involves PERK-mediated translational inhibition of the NF-κB inhibitor protein IκB 1819. In Drosophila loss of IκB could only affect Dif signaling since Dif, but not Rel, is maintained in an inactive form through an association with the Drosophila IκB protein Cactus 48. Degradation of Cactus and release of its inhibitory effect on Dif allows downstream transcriptional changes. In contrast, the Drosophila NF-κB transcription factor Rel is a bi-partite protein containing inhibitory as well as activating domains. Similarly to the mammalian p105 and p100 NF-κB proteins, Rel activation involves a proteosome-mediated cleavage event that releases the inhibitory domain from the activating domain. Therefore, it is unlikely that PEK-mediated translational attenuation of an inhibitor molecule functions in Rel activation.

The second proposed mechanism for mammalian NF-κB activation due to ER stress is mediated by Ire-1 via an interaction with the TNF-receptor-associated factor 2 (Traf2) 20. Traf2 is an E3 ubiquitin ligase that activates NF-κB and JNK signaling [51, Reviewed in 52]. Drosophila encodes a Traf2-like molecule, and there is evidence that Drosophila Traf proteins are involved in Toll, Imd and JNK signaling 53545556. One possible scenario for NF-κB activation due to ER stress in Drosophila is that Ire-1 modulates the Rel-induced ER stress pathways while Pek translational attenuation regulates Dif signaling.

In yeast, loss of p24 protein function causes an ER stress response, one consequence of which is secretion of the heat shock protein and ER stress sensor BiP 26. Interestingly, there is evidence in mammals that BiP and related heat shock proteins bind to and activate Toll-like receptors [Reviewed in 575859]. Therefore, it may be possible to induce NF-κB independently of PEK, Atf6 or Ire-1 if BiP is secreted. In Drosophila, the Toll receptor regulates Dif and Dorsal signaling. Even if BiP activation of Toll receptors plays a role in stress-induced NF-κB activation, it cannot account for the observation that Rel, which is not regulated by Toll signaling, is induced and interacts genetically with loj. Since imd regulates Rel but does not interact genetically with loj, Rel signaling in this p24 mutant context could be mediated by the Ire-1/Traf2 pathway rather than via Imd. Similarly, the Ire-1/Traf2 pathway may modulate the JNK response genes that are up regulated in loj mutants (see Additional files 2, 3 and 5). Because Rel activation requires Rel cleavage rather than disruption of an inhibitory protein interaction, it is unlikely that Rel activation is immediately downstream of Pek.

Our analysis did not identify the same set of stress-responsive genes previously shown to require Ire-1 but not other components of the UPR/EOR response 14. Only one locus, Tequila, is down regulated in both experiments. However this difference in affected genes may be due to the stressors or the time periods examined.

There are other possibilities for the observed increases in NF-κB target genes in the loj mutant. One is that the observed stress response results from tissue damage occurring in loj mutants. Tissue damage could activate NF-κB and stress response genes in the fat body. Another possibility is that increased levels are due to increased message stability rather than activation of NF-κB in the adult. We cannot rule out the possibility that changes in stability of NF-κB target gene transcripts, rather than NF-κB activation, account for the observed expression differences between loj and wild-type females.

Since loj mutants have increased expression of many genes implicated in ER stress responses and loj mutations interact genetically with mutations in two NF-κB genes, it is likely that intracellular trafficking is reduced in loj mutants. Genes that encode components of the eggshell (dec-1, Vm32E and Cp15) are down regulated in abdominal tissue (see Additional file 4). It is possible that trafficking of these proteins or is slowed in loj mutants and provides feedback inhibition of transcription. Alternatively, these products may be decreased in the loj mutant because egg production is slowed and fewer eggs of the stages that express these genes are present in loj females.

Another interesting finding is that transcript levels of the octopamine receptor Oamb are also decreased in the loj mutant (see Additional file 4). Octopamine signaling is required for egg laying in flies and other insects 606162, and reduced Oamb protein expression impairs female ovulation 63. Loss of octopamine signaling via decreased trafficking of the receptor or molecules involved in octopamine or Oamb production could account for the egg-laying defect in loj mutants.

In both loj mutant tissue samples the p24-2 gene (CG33105) is up regulated (see Additional file 1). p24-2 is a member of the Drosophila p24-alpha subfamily, while loj is a member of the p24-gamma subfamily 2228. It is possible that p24-2 can substitute for some loj functions during development or in the adult. We plan to test these hypotheses during future investigations into p24 protein function in flies.

Drosophila melanogaster stocks and cross progeny were maintained at 25°C on a 12 h light-dark cycle using a standard cornmeal, sugar, agar and yeast culture medium. The loj⁰⁰⁸⁹⁸ and loj⁰⁴⁰²⁶ alleles were described previously 27. The eca¹ allele was provided by S. Bartoszewski 45. dl¹ and Df(3R)GB104, red/TM3, Sb Ser (which removes eca) were obtained from the Bloomington Drosophila Stock Center (stock numbers 3236 and 1937, respectively). The Dif² 41 and Rel^E20 42 alleles used in this study were provided by B. J. Taylor. A description of each allele is available in Flybase (see Availability and requirements section for URL).

Females of the genotypes loj⁰⁰⁸⁹⁸/loj⁰⁴⁰²⁶, loj⁰⁰⁸⁹⁸/+, and loj⁰⁴⁰²⁶/+ were collected as virgins within 1–2 h of the beginning of the 12 h lighted portion of the light-dark cycle, kept in glass vials in groups of 20 or fewer flies, and aspirated singly into new vials on day 3 after collection. On day 4, a single wild-type male was aspirated into a vial containing a single female and observed for mating. Only females that mated within 30 min for 18–30 min were collected for subsequent RNA extractions. Female abdomens were dissected away from the head/thorax 3 hours after the end of the mating period and each of the two tissue types (abdomen or head with thorax) was quick-frozen in Trizol reagent (Invitrogen). Total RNA from abdomen or head/thorax tissue was extracted in Trizol using the manufacturer's protocol. Twelve female tissue samples were collected for each of 6 independent RNA extractions for loj⁰⁰⁸⁹⁸/loj⁰⁴⁰²⁶. Our control samples consisted of 6 loj⁰⁰⁸⁹⁸/+ and 6 loj⁰⁴⁰²⁶/+ samples combined together for each of 6 independent RNA extractions. Three experimental (loj⁰⁰⁸⁹⁸/loj⁰⁴⁰²⁶) and three control RNA samples for each set of tissue were used for array hybridization. Sample labeling and hybridization to Drosophila Genome 2.0 Arrays (Affymetrix) was performed at the University of Kentucky Microarray Core Facility using standard Affymetrix protocols for a total of 12 arrays (3 sets of experimental and control samples for each of the two tissues).

Cyber-T was used to perform Bayesian statistical analyses on expression values derived from dChipPM-MM and dChipPM 64 and GCOS (Affymetrix) essentially as described previously 65. We set stringent parameters and required that the final set of genes with altered expression levels have a significance value of p < 0.001 for all three expression values (dChipPM-MM, dChipPM and GCOS) and show at least a 1.5-fold difference from the control samples. This analysis left us with a data set of 641 genes with expression changes in one or both tissues.

To validate the microarray results, the remaining twelve RNA samples (3 experimental and 3 control preparations for each of the two tissues) were used to prepare independent cDNA samples with the Superscript 1^st Strand Synthesis Kit (Invitrogen). cDNA preparations were diluted 1:15, and 1.5 ul was used in each reaction for qPCR using the SYBR Green PCR Mastermix (Applied Biosystems). Reactions were performed in the ABI7700 or ABI7500 Fast Real-time PCR system using the default run parameters. For each plate, a melting curve analysis was performed at the end of each run to test for primer specificity. Additionally, selected reactions were electrophoresed on agarose gels to view PCR products as a second test for presence of the correct product.

Each plate also contained reactions to test for amplification specificity in the presence or absence of template. rp49 primers were used for control amplification reactions 27. cDNAs from abdomen or head/thorax preparations were analyzed separately. We selected 9 up-regulated (AttC, CG6188, CG6687, CG15829, Frost, GstD5, IM23, Socs36E, Tsp42Ed) and 5 down-regulated candidates (CG8768, CG13793, ninaD, Tequila, UGP) for this analysis and report the average fold changes from the tissues in Additional files 1 and 2. In every case, samples showed a transcriptional change in the direction expected from microarray analyses and in most instances the differences between loj mutants and controls were significantly different (p < 0.05) in both sample preparations. GstD5 transcripts were increased in loj head/thorax and abdomen samples but the results were not significantly different from controls. Down regulation of loj transcripts in loj mutant tissue was reported previously 27.

To test eclair (eca) mutants for increased expression of immune response genes, we performed a similar protocol on cDNA prepared from eca/Df(3R)GB104 and control (equal numbers of eca/+ and Df(3R)GB104/+) females using primers that specifically amplified gene products that are downstream of various stress or immune-activated pathways: CG6687 (Toll/Imd), Tsp42Ed (Imd/JNK), Frost (Toll) and Socs36E (JAK/STAT). We were not able to collect bai mutant females for analysis due to lethality of bai mutations.

We determined the number of animals of each of the possible genotypes that eclosed from stock vials for the following strains: (1) w; Dif², cn bw/Cyo, ftz-lacZ; loj⁰⁰⁸⁹⁸/TM3, Sb ftz-lacZ (701 total animals scored); (2) loj⁰⁰⁸⁹⁸, Rel^E20/TM3, Sb (321 total animals scored); (3) dl¹ cn¹ sca¹/Cyo; loj⁰⁰⁸⁹⁸/TM2 (128 total animals scored); (4) w; imd EY08573/Cyo; loj⁰⁰⁸⁹⁸/TM6b, Tb (205 total animals scored). We also crossed loj⁰⁰⁸⁹⁸, Rel^E20/TM3, Sb females to either loj⁰⁰⁸⁹⁸/TM3, Sb (206 animals scored) or Rel^E20/TM3, Sb (77 animals scored) males and determined the progeny genotypes. w; Dif², cn bw/Cyo, ftz-lacZ; loj⁰⁰⁸⁹⁸/TM3, Sb ftz-lacZ was crossed to w; Dif², cn bw/Cyo, ftz-lacZ; loj⁰⁴⁰²⁶/TM3, Sb ftz-lacZ (210 animals scored) to assay interactions between Dif²and loj transheterozygous mutants. A similar cross with dl¹ in the loj⁰⁰⁸⁹⁸or loj⁰⁴⁰²⁶ background did not produce any dl¹homozygous progeny (data not shown). Stocks were maintained in non-crowded conditions and the number of animals of each genotype that eclosed was determined on multiple days over a multi-week period. Vials were cleared the day before adult offspring were counted.

For strains containing Dif or dl alleles, each of the 4 possible genotypes from the stock cross is predicted to equally represented due to Mendelian segregation. Since loj and Rel are both on the 3^rd chromosome, there are only two possible genotypes from the stock cross from loj⁰⁰⁸⁹⁸, Rel^E20/TM3, Sb. The percentage of expected animals eclosing was calculated by dividing the total number of adult animals of a given genotype by the total number of animals of the stock genotype.

From the crosses of loj⁰⁰⁸⁹⁸, Rel^E20/TM3, Sb to either loj⁰⁰⁸⁹⁸/TM3, Sb or Rel^E20/TM3, Sb there are three classes of expected progeny (e.g., (1) loj⁰⁰⁸⁹⁸, Rel^E20/TM3, Sb (2) loj⁰⁰⁸⁹⁸/TM3, Sb and (3) loj⁰⁰⁸⁹⁸, Rel^E20/loj⁰⁰⁸⁹⁸) that should each account for one-third of the total number of offspring. We are not able to distinguish the two classes of progeny that contain the TM3, Sb chromosome. Therefore, we calculate the percentage of expected loj⁰⁰⁸⁹⁸, Rel^E20/loj⁰⁰⁸⁹⁸ progeny by dividing the total number of loj⁰⁰⁸⁹⁸, Rel^E20/loj⁰⁰⁸⁹⁸ progeny by one-half of the total number of TM3, Sb progeny.