Molecular and Functional Analysis of PRKAR1A and its Locus (17q22–24) in Sporadic Adrenocortical Tumors: 17q Losses, Somatic Mutations, and Protein Kinase A Expression and Activity¹

Despite the relatively early discovery of the TP53 gene, as one of the primary tumor suppressor genes involved in ACT⁴ formation (1, 2), few other genetic defects that are associated with this process have been identified (3, 4). Factors that have been implicated in adrenal tumorigenesis include lack of 11p15 imprinting and overexpression of the insulin-like growth factor type 2 (IGF2) gene (5, 6), alterations of inhibin A (INHA), and/or its 2q locus (7, 8), yet unidentified genes on 2p16 (9) and 9q34 (10), and various substances or transcription factors, including urotensin II (11), novH (12), and the cAMP response-element binding protein and modulator (13, 14).

Various components of the cAMP-dependent PKA signaling pathway, including the ACTH-receptor (MC2R gene; Ref. 15) and the G_sα subunit (GNAS1; Ref. 16), have also been implicated in ACT pathogenesis (17, 18). Recently, the gene coding for the PKA regulatory-subunit type-I α (RIα), PRKAR1A, was found to be responsible for most cases of a relatively rare form of bilateral adrenocortical hyperplasia, PPNAD (19, 20). Germ-line PRKAR1A-inactivating mutations were found in both patients with isolated PPNAD and those with Carney complex (CNC), a multiple neoplasia syndrome with PPNAD as its main endocrine manifestation (21, 22). In PPNAD, as in other CNC tumors, LOH of the 17q PRKAR1A locus and an abnormal PKA activity have been demonstrated, suggesting that PRKAR1A acts, perhaps, as a tumor suppressor gene (22, 23), despite some in vitro evidence that RIα is overexpressed in cancer cell lines and may be a target for chemotherapy of certain tumors (24) or that PKA suppresses proliferation of leukemic T cells (25).

PPNAD may cause “classic” CS or an insidious, clinically atypical form of hypercortisolism, which may be associated with cyclical CS, simple disturbances of the normal circadian variation of cortisol secretion, and/or other atypical features (26). Most of these patients present with osteoporosis, and, occasionally, myopathy and cachexia, but tend to lack severe obesity, persistent hypertension, moon facies, and other manifestations of CS (26, 27). Patients with PPNAD respond to low and high doses of DXM with progressively increasing glucocorticoid excretion, a “paradoxical” response that may be used diagnostically to identify otherwise asymptomatic carriers or to distinguish PPNAD from other adrenocortical tumors (27, 28).

In this study, we examined whether somatic alterations of the PRKAR1A gene were associated with the development of sporadic ACTs in 44 patients, including 26 with ACTH-independent CS. Thus far, somatic PRKAR1A locus and gene changes have only been described in thyroid tumors (29). The limitations of LOH studies with polymorphic markers that are identified in the proximity of the gene of interest by sequence tag (STS)-mapping are well known. DNA from normal tissue is not always available, some markers are uninformative, and small deletions can be missed. Thus, in the present study, in addition to LOH, we used a BAC that contains the PRKAR1A gene to study allelic losses in several of the tumors. Our findings indicated that 17q22–24 allelic losses are frequent in ACTs. The PRKAR1A gene was sequenced in all samples; novel mutations were identified in three adenomas from unrelated patients, all with paradoxical responses to DXM and one with cyclical CS, suggesting a PRKAR1A genotype-clinical phenotype correlation, for the first time in sporadic ACTs. We also studied RIα expression and PKA activity in available samples and found that genetic alterations of PRKAR1A and/or its 17q22–24 locus correlated with decreased RIα expression and alterations in PKA activity.

Forty-four patients with ACTs were recruited for this study (Tables 1 and 2); 29 with single adenomas [6 males (M) and 23 females (F), age 47 ± 14.1 years (mean and standard)] and 15 patients with cancer (6 M and 9 F, age 42.5 ± 14.4 years). A total of 26 patients was diagnosed with ACTH-independent CS by standard diagnostic testing. The hormonal investigations were performed as described previously (30, 31). Briefly, the diagnosis of CS was based on elevated 24-h urinary cortisol excretion (>90 μg/day), with low ACTH plasma levels (<15 pg/ml, normal range: 20–60 pg/ml), and abnormal response to either the overnight 1 mg of DXM suppression test (plasma cortisol > 40 ng/ml) and/or the classic low DXM (2 mg/day for 2 days) suppression test (urinary cortisol > 10 μg/day).

Adrenal tumors were obtained during surgery and immediately dissected by the pathologist; sections were fixed and paraffin embedded for diagnostic studies, and others were frozen and stored in liquid nitrogen until further use for the genetic studies. All adrenal tumors were examined by the same pathologist in one institution and classified by the use of widely accepted pathological criteria and molecular markers as reported previously by Gicquel et al. (the COMETE network; Ref. 32). Informed consent was obtained for tumor collection as part of a protocol approved by the review boards of participating institutions.

DNA was extracted from peripheral blood leukocytes using the Wizard Genomic DNA Purification kit (Promega Corp., Madison, WI). Tumor DNA and RNA were purified by cesium chloride gradient ultracentrifugation as reported previously (32). cDNA was synthesized by Moloney murine leukemia virus-reverse transcriptase (Invitrogen, Groningen, the Netherlands) using 1 μg of total RNA in a final volume of 40 μl.

Touch preparations were made from fresh or frozen tumors; these specimens were not necessarily the same as those used for the LOH or protein studies (see below), because different amounts of tissue were needed for each study. The probe that was used for FISH was a BAC that maps to 17q22–24 and contains the entire PRKAR1A gene (CITB BAC 321-G-8; Refs. 19, 20, and 29). Other BAC probes from chromosomes 6 and 22 were used as controls (Fig. 1). The BACs were grown, and DNA was extracted as described elsewhere (33). Probe DNA was labeled with digoxigenin-11-dUTP (Roche Diagnostics Corp., Indianapolis, IN) by nick-translation and hybridized to touch preparations of the tumors as described previously (29, 34). After hybridization, cells were counterstained with 4′,6′-diamidino-2-phenylindol-dihydrochloride. Signals for hybridization were analyzed with the use of a Leica epifluorescence microscope, and fluorescence images were automatically captured on a Photometrics-cooled CCD camera (Photometrics, Ltd., Tucson, AZ) using IP Lab Image software (Scanalytics, Inc., Fairfax, VA). At least 100 nonoverlapping cells with strong hybridization signals were scored per case. Presence of >20% cells with only one PRKAR1A signal was interpreted as an allelic deletion. Control touch preparations from the corresponding normal tissues showed <8% of cells with one PRKAR1A signal, and control probes from chromosomes 6 and 22 showed <10% of cells with one signal with no significant variance between samples and probes. A chromosome 17-specific centromeric α-satellite probe (Vysis) was used to control for polyploidy.

Thirty-seven paired (blood-tumor) DNA samples were studied by PCR amplification of markers surrounding the PRKAR1A gene (as established by contig mapping, see Ref. 19; Tables 3 and 4). The sequences and genomic order of these primers are available in the genome database online and from the Internet sites of genome centers.⁵

The tissue fragments that were used for these DNA studies were dissected from the core of the tumor mass immediately after surgical exploration and from an experienced pathologist, avoiding contamination with normal tissue.

A two-round PCR reaction was performed. In the first round, we used 10 pmol of each primer, and 200 nmol of each nucleotide (GeneAmp PCR reagent kit; Applied Biosystems, Foster City, CA). In the second round, an aliquot of the first reaction was amplified with the addition of 1 μCi of α-³³P-dATP (Redivue deoxyadenosine 5′–[α-³³P]-triphosphate; Amersham Pharmacia Biotech, Piscataway, NJ). For each round, after initial denaturation (5 min at 94°C), and 35 cycles (94°C for 1 min, 57°C for 1 min, and 72°C for 3 min), samples were submitted to an extension time of 7 min at 72°C. PCR products were denaturated (94°C for 5 min), placed on ice, and loaded on a 6% nondenaturing polyacrylamide sequencing gel (Gel-Mix 6; Life Technologies, Inc., Rockville, MD) for ∼2.5 h at 30 mA, depending on the size of each marker. Normal and tumor samples were loaded subsequently and in parallel to a known sequence-labeled marker to identify the exact molecular size of the bands. After electrophoresis, the gel was transferred to a filter paper and dried for 1 h in a Gel Dryer Device (Bio-Rad, Hercules, CA). The bands, corresponding to the α-³³P-dATP-labeled markers, were detected by autoradiography (model BioMax–MR; Eastman Kodak Film, Rochester, NY).

The microsatellite alterations seen in the tumors were classified as described previously (35). LOH was considered to be present when only one allele was evident in DNA extracted from patient’s tumor compared with two alleles in the DNA derived from peripheral blood lymphocytes or if the signal from one of the two alleles had a >50% reduction in intensity, compared with that of the corresponding band in peripheral blood. Microsatellite length instability was considered to be present when the amplified tumor DNA contained multiple bands or bands that differed from those seen in DNA from peripheral blood. Specimens that were not successfully amplified were excluded from the analysis. All samples were run from preset matched wells that were prepared with automated pipetting; in addition, the results from the other markers for each pair ensured that the samples with microsatellite length instability were not mismatched. The results were expressed as a percentage of the total number of informative loci in both tumor and blood samples.

The 12 exons and flanking intronic sequences of the PRKAR1A gene were separately PCR amplified using the primers; the conditions were described previously for exons 1A, 1B, 2, and 7 (19, 20, 36); and the oligonucleotides are listed in Table 5 for exons 3, 4A, 4B, 5, 6, 8, 9, and 10.

Amplified products were directly sequenced on an automated sequencer (ABI 3700; Perkin-Elmer Corp., Wellesley, MA) using the Big Dye Terminator method (Fig. 2). For samples with a mutated sequence, analysis was performed in both directions to confirm the mutation. Nucleotides were numbered in accordance with the reference sequence for PRKAR1A (GenBank accession no. NM_002734; Refs. 19 and 20).

Real-time PCR was performed using the Light Cycler apparatus (Roche Diagnostics GmbH, Mannheim, Germany). For this analysis, specimens were available from 4 tumors of the 13 nonfunctioning incidentally discovered tumors (“incidentalomas” or NSAA = nonsecreting adrenal adenomas), 9 of the 13 tumors associated with CS (that did not carry mutations of the PRKAR1A gene; Table 1), the AA21 and AA22 tumors with PRKAR1A-inactivating mutations (Table 1), and 10 specimens from the 15 ACs described in Table 2.

PCR reactions were performed in a 20-μl final volume, 0.5 μm oligonucleotide, 3.5 mm MgCl2, and the Fast Start DNA master SYBR Green (Roche) according to the manufacturer recommendations. Primers used for PRKAR1A and GAPDH amplification are shown in Table 5. Annealing was performed at 56°C for PRKAR1A amplification and 58°C for GAPDH. Product specificity was controlled by melting curve analysis and migration on a 1% agarose gel. Results are analyzed with the software of the Light Cycler apparatus. Standard curve is determined by the use of a plasmid containing the PCR product amplified from cDNA made from RNA extracted from the H293 cell line and cloned in pGEM-T easy, as reported previously (36).

To prepare total protein extracts, additional large segments of frozen tissue were homogenized in PBS; homogenates were resuspended in three volumes of M-PER reagent supplemented with complete protease inhibitor cocktail (Roche Biochemicals, Inc., Indianapolis, IN). Protein concentrations were determined by the bicinchoninic acid assay kit (Pierce, Rockford, IL), and 50 mg of protein were resolved in a 4–12% NuPage gel in 4-morpholineethanesulfonic acid buffer (Invitrogen, Carlsbad, CA) before transfer to polyvinylidene difluoride membranes. Western blots were performed using the Western Breeze kit for murine antibodies as directed by the manufacturer (Invitrogen). Monoclonal antibodies specific for R1α, R2α, R2β, and Cα were obtained from BD Transduction Laboratories (Lexington, KY) and used at dilutions specified by the manufacturer, as described elsewhere (20, 29).

To compare R1α protein levels in tumors with and without PRKAR1A mutations and/or 17q losses in the same experiment, protein was extracted anew from the three tumors with mutations (AA6, AA21, and AA22) and three without PRKAR1A/17q genetic changes (AA7 and two additional “control” adrenal tumors that were sequenced and checked for 17 q losses and had none). The lysates were processed, and the Western blots were performed as above. After exposure, the films were scanned, and absorbance was recorded by standard methods.

Sections from paraffin-embedded tissue from the two tumors with PRKAR1A mutations (AA21 and AA22), two tumors with LOH for 17q22–24 but without PRKAR1A mutations (AA 10 and 18), and three tumors without mutations or 17q losses (AA7 and the two additional tumors referred to above) were processed for IHC with monoclonal antibodies specific for RIα and the other main PKA subunits (RIIα, RIIβ, and Cα) under conditions specified by the manufacturer (BD Transduction Laboratories, San Diego, CA). At least two blinded readers graded the specimens for all stainings, as reported elsewhere (29). Briefly, the specimens were graded 0–2 with 0 being no different from the negative control (serial section of the same specimen processed for IHC without the specific antibody), 1 showing moderate expression and 2 reflecting strong immunoreactivity.

Kinase activity was measured, as described previously (19, 29), in cpm using γ-³²P-dATP (deoxyadenosine 5′-[γ-³²P]-triphosphate; Amersham Pharmacia Biotech), in 2 mg of total protein cell extracts from tumors that had been snap frozen in liquid nitrogen at the time of their excision. All determinations of PKA activity were done twice for each tumor, corrected for protein content, and an average value was calculated for each experiment.

Total kinase activity represents enzymatic activity after stimulation with cAMP; total PKA-specific activity represents the difference between PKA activity before and after the addition of PKI, as described previously (19, 29). Free PKA activity was calculated also as described previously (29); a ratio was then calculated according to the following formula [both total and free PKA activity are expressed in units (U) per milligram of protein (units/mg)]:

PKA activity ratio = Free PKA activity (in units/mg protein content)/Total PKA activity (in units/mg protein content)

Data from all tumors were compared with STATISTICA software (StatSoft, Inc., Cary, NC) using the t test for individual comparisons between the two diagnostic groups [carcinomas (n = 8) and adenomas (n = 7)]. A P < 0.05 was considered to indicate significance; a P < 0.1 was interpreted as showing a tendency toward a statistically significant difference.

All data are shown with the mean ± SE, unless indicated otherwise. Statistical analysis for quantitative PCR data were performed by nonparametric test (Mann-Whitney) using the StatView 5.0 program (SAS Institute); significance was set at P < 0.05. The χ² test was used to compare rates of LOH or mutations between adenomas and cancer, responses to DXM 150% over baseline and higher, and tumor weights < 15 grams (with the Fischer correction, when needed); PKA activity comparisons between adenomas with and without LOH and cancers with and without LOH were done with the t test, as described above.

We have mapped previously the PRKAR1A gene to the chromosomal region 17q22–24 by sequence tag (STS)-mapping to a known contig of the area (19) and were able to use polymorphic markers from the region for LOH studies. We also mapped a BAC containing the PRKAR1A gene (BAC 321-G-8) by FISH to the 17q22–24 region and determined its position with 10 linked MMs in a 10 cM area flanked by D17S942 and D17S1295 on 17q22–24 (data not shown). Among them, 8 MMs surrounding (and one from within) the PRKAR1A gene, were chosen for the LOH studies. The 321-G-8 BAC was used for FISH analysis of frozen tumor preparations.

We studied a total of 44 ACTs from 29 patients with single adenomas and 15 patients with cancer; their clinical data and 17q22–24 molecular cytogenetic results are presented in Tables 1 and 2.

Allelic losses by FISH were seen in two of seven adenomas (Fig. 1) for which either DNA from normal tissue was not available or MM analysis was uninformative (Table 1). LOH by MM analysis was detected in 5 of 22 adenomas (23%) for which paired blood and tumor DNA samples were available and the studies were informative (Table 3).

Allelic losses by FISH were seen 6 of 8 cancers (compared with two of the seven adenomas; P = 0.1; Fig. 1). LOH by MM analysis was detected in 8 of 15 cancers (53%; Table 4), a difference with adenomas that was significant (P = 0.039).

In some carcinomas (from which adequate sample was available because of their large size; n = 8), we obtained both LOH and FISH studies (Table 6). In five tumors, the two studies produced identical results (63%); in three, the data were discrepant, and after reanalysis of the touch preparations and LOH studies, as it has been suggested elsewhere (37), we assigned their status for further analysis, based on several factors that are described below:

For AC6, all cells tested by FISH showed loss of one signal; PRKAR1A (CA)n, however, did not show LOH, and the next marker (D17S1882) was uninformative, whereas the very next (D17S784) showed LOH. We interpreted these data as consistent with 17q loss, most likely a deletion of one PRKAR1A allele that occurred after exon 1 of the gene.

For AC9, all tested markers showed LOH (with the exception of the two centromeric sequences D17S784 and D17S942, which are located at least 2 million bp away from the PRKAR1A gene and were uninformative; see Table 4). FISH in this specimen showed 29% of the cells (only 9% above the cutoff of 20%) with deletion, which was the lowest percentage among all our FISH-positive tumors (Table 6). This was a large carcinoma (see Table 2) with large segments of the specimen “contaminated” with normal cells, which may explain the lower percentage of cells with deletions in the FISH studies.

Finally, AC15 had LOH for all centromeric markers up to PRKAR1A(CA)n, which, however, was uninformative because of microsatellite instability (Table 4); the FISH studies were difficult to interpret because most cells had more than four signals, indicating polyploidy and, perhaps, explaining the instability seen at the microsatellite level (Table 6). Given the LOH studies, this tumor, too, was interpreted as having 17q losses.

Sequencing of the 12 exons of PRKAR1A was performed in tumoral DNA extracted from all 44 ACTs. Three different genetic changes were found in three adrenal adenomas (AA): (a) a nonsense mutation (751 A>T in exon 6) was found in AA6; (b) a splice site mutation (9IVS-1 G>A) was found in AA21; and (c) a point mutation (1050 T>C) followed by a 22-bp deletion in exon 9 of the gene was observed in AA22 (Fig. 2). These mutations were not observed in the leukocytes of these patients, a finding consistent with somatic alterations. All of these changes lead to predicted shorter R1α protein products; however, the shorter protein was not detected in Western blotting with an R1α-specific antibody (Fig. 3 A), indicating that they, most likely, lead to nonsense mRNA decay, as we have shown to be the case in other PRKAR1A-inactivating mutations (19, 20, 39). The same antibody detects shorter R1α protein products made of the mutant gene after in vitro translation (19, 20).

It is noteworthy that these three PRKAR1A mutations are novel; they are not among those that we (19, 20, 29, 36, 39) and others (40) have described in CNC patients. These patients were screened extensively and did not meet the criteria for CNC (28), and their mutations were somatic (Fig. 2). Interestingly, the three adenomas that harbored these mutations were responsible each for CS and appeared pigmented on pathological examination, although the cortex had no other features of PPNAD (data not shown). Cyclic hypercortisolism, a frequent feature of PPNAD in the context of CNC (26, 27), was also observed in patient AA21; this patient had alternating periods of high and low cortisol excretion associated with ACTH-independent CS and adrenocortical insufficiency, respectively, during the 1 year that she was followed preoperatively.

An additional clinical feature that the 3 patients with the adenomas that harbored PRKAR1A mutations all shared was a paradoxical increase of cortisol secretion during DXM administration (Fig. 4,A). This response is considered diagnostic for PPNAD in the presence of bilateral adrenocortical hyperplasia; in a recent study, only 15% of sporadic adrenocortical adenomas had low level glucocorticoid rise in response to DXM (27). Indeed, in the present study, such paradoxical responses were only rarely present in the adenomas without PRKAR1A mutations; of a total of 18 studied, 15 had cortisol responses <150% from the baseline, whereas two of the three adenomas with mutations had a cortisol rise over 160%, and the third had a rise of 130% (Fig. 4 A; P = 0.02). Some of the tumors with 17q losses had some increase of cortisol excretion over the baseline in response to DXM, which, however, was neither significant nor more frequent than that seen in sporadic adrenal adenomas or cancer (27).

Another feature shared between known PPNAD characteristics (21, 22, 27) and those of the adenomas bearing PRKAR1A mutations was the relatively small size of the latter. The average weight of the three adenomas with mutations was 11.2 ± 0.8 grams; only two of the other adenomas were <15 grams (P = 0.007), and their average weight was 23.4 ± 12.05 grams (Table 1; Fig. 4 B).

It is noteworthy that none of the 15 cancers had mutations of the PRKAR1A gene (data not shown), a difference from adenomas that was, however, not significant (P = 0.4292).

Quantitative, real-time PCR demonstrated variable levels of PRKAR1A among ACTs. All experiments were done in duplicate, however, and the results were almost identical for each specimen, despite the interspecimen variability (data not shown). For all of the tumors that were studied, the data were expressed as the number of copies of PRKAR1A mRNA versus the number of copies of the housekeeping GADPH mRNA, as explained in “Materials and Methods.” These data are shown in Fig. 5. Enough material to study PRKAR1A expression by real-time PCR was available from two of the adenomas with mutations (AA21 and AA22). Among the nonmutant tumors, nonsecreting adenomas had the highest level of PRKAR1A mRNA (0.232 ± 0.098 copies of PRKAR1A corrected for GAPDH), when compared with adrenal adenomas responsible for CS (0.034 ± 0.009) and cancers (0.004 ± 0.002; P < 0.001; Fig. 5). The two adenomas with PRKAR1A mutations exhibited a very low level of PRKAR1A mRNA (0.001 ± 0.0000245) that was comparable with that of adrenal tissue from patients with germ-line PRKAR1A-inactivating mutations (data not shown).

Western blotting with a R1α-specific antibody showed differences consistent with the mRNA studies; R1α protein was decreased in the three tumors (AA6, AA21, and AA22) with PRKAR1A mutations when compared with tumors that had no mutations or 17q genetic changes (Fig. 3, A and B).

The findings on IHC (with the same antibodies that were used above) correlated well with those of Western blotting. Consistent with what is shown in Fig. 3,A, there was no detectable expression of the R1α protein in AA21 and AA22 (Fig. 6,B) compared with that of adenoma AA7 or other nodular adrenal tissue that did not harbor any 17q genetic changes or PRKAR1A mutations (Fig. 6,G). Interestingly, too, adenomas with LOH for 17q22–24 showed lower R1α expression in adenomatous tissue compared with that in surrounding normal adrenal cortex; a representative staining is shown in Fig. 6,E for AA18, which had LOH from D17S1882 to the PRKAR1A locus (Table 3), but did not have PRKAR1A mutations.

When all of the PKA subunits were studied by Western blotting, among tumors without R1α mutations, in these samples with LOH and/or other allelic 17q losses, there was higher R2α (P = 0.027) but not R2β (P = 0.48), and a significantly lower Cα (P < 0.001; Fig. 3,C); the R1α levels per se were not significantly different (P = 0.67). Overall, although tumors without 17q losses had a R1α:R2α ratio of 0.53 ± 0.007, tumors with 17q losses had a ratio of 1.21 ± 0.08 (P < 0.001). Accordingly, the former had a total R subunit (R1α + R2α + R2β) to Cα ratio of 0.72 ± 0.0001, whereas in the latter, this ratio was 6.39 ± 1.05, a significant difference (P = 0.023). IHC supported these data; adenomas with PRKAR1A mutations or 17q losses tended to have strong immunoreactivity for type-II subunits (Fig. 6 C).

At baseline, all groups of tumors that were studied [a total of eight cancers with (group 1, n = 5) and without 17q allelic losses (group 2, n = 3) and 7 adenomas with (group 3, n = 4) and without 17q allelic losses (group 4, n = 3)] had similar kinase activity (Fig. 7 a): group 1 (17,565 ± 2,763 cpm), group 2 (13,288 ± 1,660 cpm), group 3 (17,985 ± 2,368 cpm), and group 4 (12,050 ± 2,097 cpm; for all comparisons, P > 0.05).

After exposure to cAMP, all tumors responded with an increase in kinase activity (P < 0.001, for all comparisons between each group’s baseline and after cAMP-peak value; Fig. 7 a). However, adenomas overall had a much higher stimulation of their kinase activity (198,806 ± 27,915 cpm) than cancers (56,836 ± 4,474 cpm; P < 0.001). The opposite was true for inhibition of the cAMP-stimulated activity by PKI, a PKA-specific inhibitor: cancers were inhibited less (12,515 ± 1,544) than adenomas (5,645 ± 1,364; P = 0.006).

Both cancers and adenomas with 17q allelic losses had higher stimulation of kinase activity in response to cAMP versus that without 17q losses: group 1 (63,386 ± 5,163 cpm) versus group 2 (45,920 ± 1,646 cpm; P = 0.046) and group 3 (250,597 ± 20,370 cpm) versus group 4 (129,752 ± 21,277 cpm; P = 0.009). There were no statistically significant differences between each subgroup after inhibition with PKI: group 1 (13,501 ± 2,294 cpm) versus group 2 (10,871 ± 1,600 cpm; P = 0.45) and group 3 (6,014 ± 1,788 cpm) versus group 4 (5,154 ± 2,535 cpm; P = 0.78).

Adenomas had higher total (14,486 ± 2,064 units/mg) and free PKA activity than cancers (9,433 ± 729 units/mg; Fig. 7 b). Both total and free PKA activity were higher in adenomas with 17q losses (18,343 ± 1,520 and 8,977 ± 1,363 units/mg, respectively) than those without (9,344 ± 1,437 and 5,172 ± 613 units/mg, respectively), although the free PKA comparisons did not reach significance levels (P = 0.009 and 0.074, respectively). Between cancers with and without 17q losses, there was a tendency for higher total PKA activity in the former (10,457 ± 866 units/mg) than latter (7,726 ± 367 units/mg; P = 0.06); free PKA activity was not different between the two groups (2,666 ± 465 and 2,056 ± 312 units/mg, respectively; P = 0.39).

PKA activity ratio was higher in adenomas than carcinomas (0.529 ± 0.057 versus 0.259 ± 0.024, respectively; P < 0.001; Fig. 7 c). This ratio was not significantly lower in adenomas with 17q losses (0.486 ± 0.054) versus those without losses (0.586 ± 0.118; P = 0.43); it was also not different between cancers with (0.255 ± 0.036) and without 17q losses (0.266 ± 0.037; P = 0.84).

Several genetic abnormalities have been described in adrenal tumors (9, 10). In general, adrenal adenomas tend to have fewer genetic changes than cancers (1, 4, 32, 40); the extent of their genetic alteration correlates positively with size, which in ACTs is a reliable indicator of the benign or malignant nature of the tumor (41, 42). In the present study, we investigated the hypothesis that sporadic ACTs had somatic genetic changes of the 17q22–24 locus and/or the PRKAR1A gene. A recent study by comparative genomic hybridization, a technique with significantly less sensitivity than FISH and LOH (43), found 17q alterations in ACTs (44). Zhao et al. found gains of the most telomeric portion of 17q, an area that is distant from the PRKAR1A locus; other data indicate that more centromeric regions are likely to be involved in deletions (42). Our FISH and LOH data indeed confirmed frequent losses of the 17q22–24 region in ACTs; again, carcinomas had more frequent losses than adenomas (Tables 3 and 4).

This is the only study, to our knowledge, that has examined by both FISH and LOH several adrenal tumors. The two investigations produced similar results in 63% of the specimens, a rate that is generally similar to that in other settings, e.g., bladder cancer (38) or endocrine pancreatic tumors (45). In our reanalysis, factors such as the amount of tissue available (whether the touch preparation was done from the center of the mass or its edges), the cells that were found monosomic and/or polyploid by FISH, the benign or malignant nature of the tumor, as well as whether LOH was present in other, proximal MMs were taken into account (Table 6).

Several lines of evidence suggest that these changes are specifically related to PRKAR1A: (a) in addition to 17q BAC losses and LOH of several MMs, we also studied an intragenic PRKAR1A marker (19, 20); (b) unlike the widely present correlation between the size of ACTs and number of their genetic defects (41, 42), we found no correlation between the size of our tumors and frequency or extent of 17q losses; and (c) perhaps most importantly, sequencing of all ACTs in this study showed that three adenomas from unrelated patients with CS but without CNC or any other inherited condition harbored somatic PRKAR1A-inactivating mutations. These tumors were the smallest in size and associated with paradoxical cortisol responses to DXM (Fig. 4), all features of PPNAD (26, 27, 28).

It should be noted, however, that in two of the three tumors with PRKAR1A mutations (patients AA6 and AA22), we were not able to show losses of the PRKAR1A locus with the caveats that, at least, in patient AA6, few of the studied MMs were informative (Table 3) and that we did not use microdissected specimens for our FISH and DNA studies, which makes “contamination” with normal cells from these benign tumors not unlikely. PRKAR1A’s complex molecular regulation and possible methylation of its promoter⁶ may also account for the lack of LOH in these specimens, as well as for the absence of PRKAR1A coding sequence mutations in the DNA of the other tumors with 17q allelic losses. Indeed, down-regulation of the R1α protein was suggested by immunostaining in two specimens with 17q losses and lack of mutations of the coding sequence of the PRKAR1A gene (Fig. 6 E). In at least one other example of a tumor suppressor gene involved in endocrine tumorigenesis, methylation of the PTEN promoter was the apparent explanation behind the discrepancy between high rates of 10q LOH and absence of detectable sequence alterations in thyroid tumors (46). Furthermore, more recent data from our laboratory (39) also support the notion that PRKAR1A haploinsufficiency may be sufficient for tumorigenesis in at least some cases. According to these data, PRKAR1A may cause tumors even after simple haploinsufficiency is caused by either mutations of epigenetic silencing from one allele, whereas loss of the normal allele may further exacerbate its tumorigenic activity by annihilating any PRKAR1A expression in certain tissues or cells.

Could the lack of PRKAR1A mutations in tumors with 17q allelic losses indicate presence of another tumor suppressor gene in its proximity? This was the case with at least one other gene near PTEN at 10q; the identification of MINPP1 showed that at least some thyroid tumors with 10q LOH (47) had mutations of this gene (and not of PTEN; Ref. 48). However, LOH around PTEN involved genomic regions significantly larger (47) than the PRKAR1A locus, and examination of the available 17q22–24 genomic maps between D17S942 and D17S1295 did not show any other expressed sequences likely to be functioning as tumor suppressor genes (19, 20).

In favor of PRKAR1A’s specific involvement in the tumors included in this report are three additional studies that we performed: (a) quantitative PCR showed down-regulation of the PRKAR1A mRNA in tumors with inactivating mutations and adenomas with 17q allelic losses only (Fig. 5); (b) Western blotting showed that tumors with 17q losses had more type II regulatory subunits than tumors without any 17q changes (Fig. 3), just as in PPNAD (49) and cultured fibroblasts (40) carrying PRKAR1A inactivating mutations. A switch to type II PKA has been postulated as the underlying anomaly in cAMP signaling in CNC tumor cells (19, 20, 40, 50). Increased stability of mRNA or protein, and in some cases transcriptional enhancement, may be responsible for the increased presence of type-II regulatory subunits in these tumors (51, 52, 53, 54, 55); and (c) we found significant PKA activity alterations between tumors with and without 17q allelic losses and between cancers and adenomas. In general, loss of 17q22–24 alleles was associated with higher response of total kinase activity in response to cAMP and relatively decreased PKA activity ratio (Fig. 7). We have reported both these features in tumors associated with CNC and carrying PPNAD-inactivating mutations and 17q22–24 allelic losses (19, 22), as well as in sporadic thyroid carcinomas with somatic PRKAR1A down-regulation (29). Although regulation of the PKA tetramer differs significantly between tissues and cells at different developmental stages (56), these data are also consistent with the lower kinase activity found in RIIβ−/− mouse cells that have a compensatory up-regulation of RIα (57, 58).

A significant difference between the results of this study and our previous one on sporadic thyroid tumors (29) is the greater cAMP-induced PKA activity in adrenocortical adenomas; thyroid adenomas, to the contrary, had lower kinase activity than undifferentiated (anaplastic) carcinomas (29). The present study also showed that adrenocortical adenomas, even those with 17q allelic losses, had a greater PKA activity ratio, which is an indirect measure of RIα activity (59). These data agreed well with decreased PRKAR1A mRNA presence in cancers versus adenomas (Fig. 5). Despite their lower total kinase activity, adrenocortical carcinomas retained higher cAMP-induced kinase activity in the presence of PKI, a PKA-specific inhibitor. These data may indicate not only that RIα activity is decreased in adrenocortical cancer independently of the status of the 17q locus (as suggested by the PKA activity ratio) but also that cAMP in these tumors induces non-PKA-dependent alterations in total kinase activity through pathways that may be differentially expressed in cancer versus adenomas (60).

The lack of PRKAR1A mutations in AC and their presence in adrenocortical adenomas, which is the reverse of the situation in thyroid tumors (29), suggest that diverse tumorigenic processes with regards to the role of PKA exist in endocrine cells. These observations parallel our clinical experience; in >350 patients with CNC and/or PPNAD known to us, there has never been a case of adrenocortical cancer, in contrast to several cases of thyroid carcinomas (61). We speculate that PRKAR1A down-regulation through mutations, allelic haploinsuffiency, or epigenetic silencing may participate in the evolution of thyroid cancer toward more undifferentiated forms, as it has been described in thyroid oncogenesis (62) in a linear model and as reported in other tissues (63). In the adrenal cortex, however (and perhaps in the earlier stages of thyroid tumorigenesis), two alternative and not mutually exclusive hypotheses may be proposed: (a) PRKAR1A’s down-regulation could be a very early step in carcinogenesis and, therefore, quickly bypassed by other pathways more critical for the survival of neoplastic cells; some of these signaling cascades may be cAMP-linked but not necessarily PKA-dependent, as in other tissues (60, 64); and (b) when it occurs, especially in the form of PRKAR1A-inactivating mutations accompanied by LOH, it may lead to a well-differentiated adenoma with PPNAD-like functional properties that is very unlikely to ever proceed to AC, just like Lkb1 −/− cells appear to be resistant to at least one form of malignant transformation (64). The reduction of the expression of the PKA-regulated transcription factor cAMP-responsive element binding protein in adrenocortical cancer (13, 14, 65) is further supportive of the relatively decreased role that cAMP-dependent PKA signaling may play in at least the late stages of malignant adrenocortical transformation.

In conclusion, the present study supports a role for PRKAR1A in the formation of sporadic adrenocortical adenomas; in some of these tumors, PRKAR1A inactivation is associated with CS that shares some of the characteristics of cortisol hypersecretion observed in Carney complex and/or PPNAD.

We thank Constantine Farmakidis who participated in some of the earliest experiments measuring PKA activity in adrenal tumors. We also thank Dr. Cho-Chung (National Cancer Institute) for useful discussions on PKA and its regulation, Dr. A. Louvel (Laboratory of Pathology, Hospital Cochin, Paris) for pathological examination, and F. René-Corail (Hospital Cochin, Paris) for excellent technical assistance. Finally, we thank Dr. William Farrell (Keele University, Stoke-on-Tent, United Kingdom) who reviewed some of the experiments and the final version of this manuscript.