BRAF mutant gastrointestinal stromal tumor: first report of regression with BRAF inhibitor dabrafenib (GSK2118436) and whole exomic sequencing for analysis of acquired resistance

Activating oncogenic mutations of BRAF have been described in patients with gastrointestinal stromal tumor (GIST), but treatment of GIST with BRAF inhibitors and mechanisms of mediating the emergence of resistance in GIST have not been reported. Dabrafenib is a potent ATP-competitive inhibitor of BRAF kinase and is highly selective for mutant BRAF in kinase panel screening, cell lines, and xenografts. We report prolonged antitumor activity in the first patient with V600E BRAF-mutated GIST who was treated with a BRAF inhibitor. Whole exome sequencing performed in tumor tissue obtained at the time of progressive disease demonstrated a somatic gain-of-function PIK3CA mutation (H1047R) as well as a CDKN2A aberration, which may have contributed to eventual resistance to treatment.

Abstract P2-08-03: Phosphatidylinositol-3-kinase mutations are common in lobular neoplasia

The phosphatidylinositol-3-kinase pathway is one of the most commonly mutated in invasive breast carcinoma, with PIK3CA mutations present in ∼25% of invasive carcinomas, and several studies demonstrating an even higher prevalence of PIK3CA mutations in invasive lobular carcinomas. Lobular neoplasia (LN), including lobular carcinoma in situ (LCIS) and atypical lobular hyperplasia (ALH), are controversial lesions that may represent non-obligate precursor of invasive lobular carcinoma. However, lobular neoplasia has not yet been systematically studied for activating point mutations.

Twenty-six breast resection specimens containing LN and/or invasive lobular carcinoma (ILC) were identified from the files of Oregon Health & Science University. Adjacent lesions including columnar cell change (CCC), usual ductal hyperplasia (UDH), invasive or in-situ ductal carcinoma, and lymph node metastases were separately isolated where available. DNA was prepared from punches of formalin-fixed paraffin-embedded tissue blocks using standard methods. DNA extracts were screened for a panel of point mutations using a multiplex PCR panel with a mass-spectroscopy readout (Sequenom MassArray). The panel covers 643 point mutations in 53 genes including AKT1/2/3, ALK, BRAF, CDK4, CSF1R, CTNNB1, EGFR, ERBB2, ERCC6, FBX4, FBXW7, FES, FGFR1/2/3/4, FOXL2, GNA11, GNAQ, GNAS, HRAS, IDH1/2, IGF1R, KDR, KIT, KRAS, MAPK2K1/2/7, MET, MYC, NEK9, NRAS, NTRK1/2/3, PDGFRA, PIK3CA, PIK3R1/4/5, PKHD1, PRKCB1, RAF1, RET, SMO, SOS1, STAT1, TEC, and TP53; covering 41 substitutions in 23 codons of the PIK3CA gene.

PIK3CA mutations were identified in 8/22 LN (36%; PIK3CA exon 4 N345K-1; exon 9 E542K-1; E545K-3; exon 20 H1047R-3), and in 11/16 ILC (68%; PIK3CA exon 4 N345K-2; exon 9 E542K-1, E545K-2, Q546R-1; exon 20 H1047L-1, H1047R-4, one with concomitant HRAS G12D mutation). LN and coincident ILC were tested in 11 patients; 4 patients had the same point mutations in LN and ILC (concordant mutant); 4 patients were wildtype for all codons tested in LN and ILC (concordant wildtype); 3 patients had discordant mutation status (LCIS-E545K/ILC-H1047R; LCIS-E542K/ILC-H1047R; ALH-WT/ILC-H1047R & HRAS G12V). Four patients had LN and invasive ductal carcinoma (IDC); of these, 2 were discordant (LCIS-H1047R/IDC-wildtype; ALH-wildtype/IDC-E545K) and 2 concordant wildtype. Two additional patients had discrete IDC and ILC tumors; in both the IDC was wildtype, but the ILC harbored a PIK3CA mutation. Concurrent CCC and UDH were also screened, yielding 10/22 (45%) lesions with PIK3CA point mutations; in 3 instances the UDH mutation was concordant with LN/ILC, whereas CCC from the same specimen has discordant mutational status.

Our study confirms the high prevalence of PIK3CA hotspot point mutations in ILC (68%). Importantly, we screened LCIS and ALH for a large panel of point mutations, and found only PIK3CA mutations (36% of lesions). Although a small cohort, the mutation status of concurrent LN and ILC was frequently concordant (8/11=72%), and in fact, quite similar to the degree of mutational concordance between paired DCIS and IDC in the literature and in our own experience (66–77%). This provides some support to the notion of LN as a precursor to ILC.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P2-08-03.

P2-06-04: Phosphatidylinositol-3-Kinase Pathway Mutations Are Common in Breast Columnar Cell Lesions

The phosphatidylinositol-3-kinase pathway is one of the most commonly mutated in invasive breast carcinoma, with PIK3CA mutations present in ∼25% of invasive carcinomas, and AKT1 mutations identified in up to 5%. Several studies have demonstrated the same complement of mutations in ductal carcinoma in-situ (DCIS), as well as benign papillomas. We sought to investigate whether PIK3CA mutations occur in breast columnar cell lesions (CCL).

Twenty-five breast resection specimens containing CCL (including columnar cell change, columnar cell hyperplasia, and flat epithelial atypia) were identified from the files of Stanford University Pathology; 15 of these had associated invasive carcinoma (IDC) or carcinoma in situ. DNA was prepared from punches of formalin-fixed paraffin-embedded tissue blocks using standard methods. DNA extracts were screened for a panel of point mutations using a multiplex PCR panel with a mass-spectroscopy readout (Sequenom MassARRAY). The panel covers 321 mutations in 30 genes, including ABL, AKT1/2/3, BRAF, CDK4, CTNNB1, EGFR1, ERBB2, FBX4, FBXW7, FGFR1/2/3, FLT3, GNAQ, HRAS, JAK2, KIT, KRAS, MAPK2K1/2, MET, NRAS, PDGFRA, PIK3CA, PTPN11, RET, SOS1, and TP53. The majority of mutations were confirmed by Sanger sequencing. PIK3CA mutations were identified in 12/25 CCL (48%); paired normal breast tissue was tested in 21 cases and was negative for mutations in all but one case. In associated DCIS, 4/8 (50%) harbored PIK3CA mutations, while 3/9 IDC had mutations (33%, 2 PIK3CA, 1 AKT1). The mutation status of CCL and carcinomas was frequently discordant. Of 15 cases, only 6 demonstrated the same genotype in matched samples of CCL and carcinoma (5 wildtype, 1 PIK3CA H1047R). Interestingly, 5 patients had mutations in CCL with wildtype DCIS or IDC; 2 patients had different point mutations in CCL and carcinoma, including one patient with discordant mutant DCIS and wildtype IDC. Only 3 cases had wildtype CCL and mutated carcinoma.

The nearly 50% PIK3CA mutation prevalence in CCL is greater than reported in most studies of invasive breast cancer. Further, CCL and carcinoma were frequently discordant for PIK3CA/AKT1 mutation status; most commonly the CCL harbored a PIK3CA mutation, while the associated carcinoma was wildtype. Although these findings need validation in a larger study, they raise interesting questions as to the role of PIK3CA/AKT pathway in breast carcinogenesis, and as to the biologic/precursor potential of CCL.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P2-06-04.

Evaluation of the presence of IGF1R overexpression in wild-type and kinase mutant GI stromal tumors

10506

Background: Most adult GI stromal tumors have gain-of-function mutations in KIT (80%) or PDGFRA (5–7%). These pathogenetic mutations are the target for kinase inhibitor therapy. The pathogenesis of the 10–15% of GISTs lacking kinase mutations (WT GIST) is unknown; this includes most pediatric GISTs. Recently, Tarn et al. (PNAS 2008) identified IGF1R over-expression in all WT GIST in their series of cases, including one pediatric case. IGF1R may represent a novel therapeutic target for WT GISTs. Methods: We developed a quantitative RQ-PCR assay for IGF1R and GAPDH transcripts that is linear over 5 logs (R2 values of 0.99 and 0.98, respectively). We prepared cDNA from RNA isolated from 79 archival, formalin-fixed, paraffin-embedded GIST specimens and quantified IGF1R expression as the % ratio of IGF1R/GAPDH transcripts. We had previously genotyped the tumors to identify any kinase mutations. Results: Adult WT GIST had a 10–15 fold higher expression of IGF1R transcript than kinase mutant GISTs (WT GIST 13.7 ±18, n=35; KIT exon 11 mutant GIST 1.0 ± 1.1, n=26; PDGFRA mutant GIST 0.9 ± 1.0, n=8; KIT 9 or 13 mutation 0.8 ± 0.6, n=5). Interestingly, the WT GIST group could be further divided into tumors with low vs. high IGF1R expression. The low IGF1R expression group (mean 0.6, range 0.1–1.9, n=14), had similar IGF1R expression to kinase mutant GISTs. In contrast, the high expression group had more than 30-fold higher levels of IGF1R transcript (mean 22.5, range 7.3–76.5, n=21). We also examined IGF1R expression in 5 WT pediatric GISTs. The mean IGF1R expression in pediatric GISTs resembled that of the adult high expression group, with 4 of 5 cases showing IGF1R over-expression. Conclusions: WT GISTs are heterogeneous with regard to IGF1R expression and can be divided into two subgroups. One third of the WT GISTs overlap with kinase-mutant GISTs, while the remainder express IGF1R at more than 30-fold higher levels. These results have implications for evolving models of GIST pathogenesis and the clinical testing of IGF1R monoclonal antibodies for treatment of metastatic GIST.

[Table: see text]

Cytokine response gene 8 (CR8) regulates the cell cycle G1-S phase transition and promotes cellular survival

Cellular proliferation and survival are modulated by the expression of specific genes. Cytokine response gene 8 (CR8), which was originally cloned as an IL-2-induced gene in human T lymphocytes, encodes a basic helix--loop--helix (bHLH) transcription factor. The CR8 gene product is highly conserved among human, mouse and rat, and contains sequence motifs that distinguish it from other bHLH families. The CR8 gene is ubiquitously expressed, and CR8 gene expression is induced by both growth-promoting as well as growth-inhibitory stimuli. As bHLH proteins have been found to regulate both the G1-S phase cell cycle transition, as well as cellular survival, the effects of CR8 on these processes were investigated. Ectopic CR8 expression in asynchronous U2OS cell cultures reduces the percentage of cells in the cell cycle S phase, and also slows the entry of G1-synchronized cells into S phase. The prolonged G1 interval correlates with impaired elevation of cyclin E protein and prolonged p21 protein expression in G1. CR8 expression also protects U2OS cells from serum-withdrawal induced apoptosis. These results indicate that CR8 is an important modulator of both the G1-S phase cell cycle transition, and cellular survival.

Cytokine response gene 6 induces p21 and regulates both cell growth and arrest

Cytokine response gene #6 (CR6), cloned from interleukin 2-stimulated T lymphocytes, is homologous to GADD45 and MyD118, genes which promote cell cycle arrest and apoptosis. To determine how this gene family could possibly mediate both cell survival/proliferation and cell cycle arrest/death, transfectants were generated so that the genes could be expressed ectopically, independently from their normal inducing agents. In cycling retinoblastoma protein-negative (pRb-) cells, ectopic CR6 expression blocked G2/M transition, but did not prevent G1/S transition so that endoreduplication resulted. By comparison, when CR6, GADD45, and MyD118 genes were expressed ectopically in proliferating pRb+ cells, either G1/S or G2/M transition was effectively blocked, so that there was no endoreduplication. Consistent with these findings, in proliferating pRb-cells, ectopic expression of CR6 promoted the expression of both G1 and G2/M cyclins. By comparison, in pRb+ cells, the expression of G1 cyclins was increased, while expression of the mitotic cyclins was decreased. However, in pRb+ cells, cyclin-dependent kinase activities associated with both G1 and G2/M cyclins were decreased. Moreover, ectopic expression of all three genes resulted in the expression of the CKI, p21, both in pRb- and pRb+ cells. The physiologic induction of CR6 expression by IL2 in quiescent normal human T cells occurs transiently in the first half of G1, coordinately with the expression of p21. Therefore, this gene family regulates G1 and G2, and promotes either cell growth or arrest by a common mechanism.

Regulators of G protein signaling exhibit distinct patterns of gene expression and target G protein specificity in human lymphocytes

The newly recognized regulators of G protein signaling (RGS) attenuate heterotrimeric G protein signaling pathways. We have cloned an IL-2-induced gene from human T cells, cytokine-responsive gene 1, which encodes a member of the RGS family, RGS16. The RGS16 protein binds Gialpha and Gqalpha proteins present in T cells, and inhibits Gi- and Gq-mediated signaling pathways. By comparison, the mitogen-induced RGS2 inhibits Gq but not Gi signaling. Moreover, the two RGS genes exhibit marked differences in expression patterns. The IL-2-induced expression of the RGS16 gene in T cells is suppressed by elevated cAMP, whereas the RGS2 gene shows a reciprocal pattern of regulation by these stimuli. Because the mitogen and cytokine receptors that trigger expression of RGS2 and RGS16 in T cells do not activate heterotrimeric G proteins, these RGS proteins and the G proteins that they regulate may play a heretofore unrecognized role in T cell functional responses to Ag and cytokine activation.

Elf-1 and Stat5 bind to a critical element in a new enhancer of the human interleukin-2 receptor alpha gene

The interleukin 2 receptor alpha-chain (IL-2R alpha) gene is a key regulator of lymphocyte proliferation. IL-2R alpha is rapidly and potently induced in T cells in response to mitogenic stimuli. Interleukin 2 (IL-2) stimulates IL-2R alpha. transcription, thereby amplifying expression of its own high-affinity receptor. IL-2R alpha transcription is at least in part controlled by two positive regulatory regions, PRRI and PRRII. PRRI is an inducible proximal enhancer, located between nucleotides -276 and -244, which contains NF-kappaB and SRE/CArG motifs. PRRII is a T-cell-specific enhancer, located between nucleotides -137 and -64, which binds the T-cell-specific Ets protein Elf-1 and HMG-I(Y) proteins. However, none of these proximal regions account for the induction of IL-2R alpha transcription by IL-2. To find new regulatory regions of the IL-2R alpha gene, 8.5 kb of the 5' end noncoding sequence of the IL-2R alpha gene have been sequenced. We identified an 86-nucleotide fragment that is 90% identical to the recently characterized murine IL-2-responsive element (mIL-2rE). This putative human IL-2rE, designated PRRIII, confers IL-2 responsiveness on a heterologous promoter. PRRIII contains a Stat protein binding site that overlaps with an EBS motif (GASd/EBSd). These are essential for IL-2 inducibility of PRRIII/CAT reporter constructs. IL-2 induced the binding of Stat5a and b proteins to the human GASd element. To confirm the physiological relevance of these findings, we carried out in vivo footprinting experiments which showed that stimulation of IL-2R alpha expression correlated with occupancy of the GASd element. Our data demonstrate a major role of the GASd/EBSd element in IL-2R alpha regulation and suggest that the T-cell-specific Elf-1 factor can serve as a transcriptional repressor.

Interleukin-2 activation of STAT5 requires the convergent action of tyrosine kinases and a serine/threonine kinase pathway distinct from the Raf1/ERK2 MAP kinase pathway

Interleukin-2 (IL-2) induces DNA binding of STAT5, a member of the family of cytokine-regulated transcription factors termed 'signal transducers and activators of transcription'. IL-2-stimulated STAT5-DNA complexes include two tyrosine phosphoproteins which exhibit distinct mobilities in SDS-PAGE gels. Our studies have shown that IL-2 rapidly induces both tyrosine phosphorylation and serine phosphorylation of STAT5 and that the two STAT5 tyrosine phosphoproteins detected in IL-2-activated cells differ in their levels of phosphorylation on serine residues. The two different phosphoforms of STAT5 have identical in vitro DNA binding specificity and reactivity with tyrosine phosphopeptides, but differ in their cellular localization. As well, the present data indicate that the transcriptional activity of STAT5 is regulated by serine kinases in T lymphocytes. Two previously characterized serine kinases activated by IL-2, MAP kinase/ERK2 and p70 S6 kinase, do not appear to be involved in STAT5 regulation by this cytokine. Accordingly, STAT5 activation in T cells requires the convergent action of tyrosine kinases and a distinct serine/threonine kinase which has not previously been implicated in IL-2 signalling.

Interleukin-7 can induce the activation of Jak 1, Jak 3 and STAT 5 proteins in murine T cells

The activation of Janus protein tyrosine kinases (Jak) and STAT (signal transducer and activator of transcription) proteins has recently been linked to the signal transduction mechanism of several cytokines. IL-7 was observed to induce a rapid and dose-dependent tyrosine phosphorylation of Jak 1 and Jak 3 and concomitantly, the tyrosine phosphorylation and DNA binding activity of multiple STAT proteins. The STAT proteins utilized by IL-7 were identical to those induced by IL-2 and could be identified as various STAT 5 isoforms. Moreover, the induction of both Jak 1 and 3, and STAT 5 activity strongly correlated with the growth-promoting effects of IL-7, suggesting that this signal transduction mechanism may play a key role in IL-7-induced proliferation.

Activation of JAK kinases and STAT proteins by interleukin-2 and interferon alpha, but not the T cell antigen receptor, in human T lymphocytes

The activation of Janus protein tyrosine kinases (JAKs) and signal transducer and activator of transcription (STAT) proteins by interleukin (IL)-2, the T cell antigen receptor (TCR) and interferon (IFN) alpha was explored in human peripheral blood-derived T cells and the leukemic T cell line Kit225. An IL-2-induced increase in JAK1 and JAK3, but not JAK2 or Tyk2, tyrosine phosphorylation was observed. In contrast, no induction of tyrosine phosphorylation of JAKs was detected upon stimulation of the TCR. IFN alpha induced the tyrosine phosphorylation of JAK1 and Tyk2, but not JAK2 or JAK3. IFN alpha activated STAT1, STAT2 and STAT3 in T cells, but no detectable activation of these STATs was induced by IL-2. However, IL-2 regulates the DNA binding and tyrosine phosphorylation of two STAT-like protein complexes which do not include STAT1, STAT2 or STAT3. STAT4 is not activated by IL-2. The activation of STAT5 cannot be excluded, so the IL-2-activated complexes most probably include at least one novel STAT. No STAT activity was detected in TCR-stimulated lymphocytes, indicating that the JAK/STAT pathway defined in this study constitutes an IL-2R-mediated signaling event which is not shared by the TCR. Finally, in other cell types the correlation between JAK1 activation and the induction of STAT1 has suggested that JAK1 may activate STAT1. The observation that IL-2 and IFN alpha activate JAK1 to a comparable degree, but only IFN alpha activates STAT1, indicates that JAK1 activation is not the only determining factor for STAT1 activation. Moreover, the data show that JAK1 stimulation is also not sufficient for STAT3 activation.

In search of cytokine-response genes

The development of efficient cloning methods for the isolation of cytokine-response genes is vital to our understanding of how cytokines elicit distinct cellular responses. Here, Carol Beadling and Kendall Smith describe a straightforward and rapid method that has been used to clone interleukin 2 (IL-2)-induced genes.

Tactics for the isolation of interleukin-2-induced immediate-early genes

Interleukin-2 (IL-2), a 15 kDa protein secreted by antigen-activated T lymphocytes, plays a central role in the generation of an immune response. Initially characterized as a T cell growth factor, IL-2 has subsequently been found to act on all lymphoid cell types, inducing both their proliferation and functional differentiation. The effects of IL-2 are mediated by the activation of a specific cell-surface receptor (IL-2R), which is comprised of at least three distinct proteins. Early biochemical signaling events triggered upon IL-2R activation have recently begun to be elucidated, and a number of genes have been identified which are specifically induced by IL-2. The task now at hand is to link the biochemical signaling pathways activated by IL-2 with the regulation of gene expression, and to delineate the roles of these gene products in IL-2-mediated proliferation and differentiation.

Isolation of interleukin 2-induced immediate-early genes

Clonal expansion of antigen-reactive T lymphocytes is driven by the lymphokine interleukin 2 (IL-2). To further elucidate the mechanisms of IL-2 action, we have utilized a differential hybridization procedure to clone IL-2-induced immediate-early genes from an IL-2-stimulated human T-cell cDNA library. To increase the frequency of IL-2-induced transcripts represented in the library, the protein synthesis inhibitor cycloheximide was included during the 2-hr IL-2 stimulation to superinduce gene expression, and the uridine analogue 4-thiouridine was utilized to enable selective purification of newly synthesized transcripts. From the enriched library, we have isolated eight IL-2-induced genes, six of which represent previously unrecognized human sequences. Northern blot analysis revealed that the induction of seven of the genes is specific to the IL-2-mediated G1 "progression" phase of the cell cycle, in that only one gene is also induced during the T-cell receptor-triggered G0-G1 "competence" phase. These results indicate that the effects of IL-2 are mediated by the specific induction of a number of immediate-early genes and provide a means with which to further delineate the mechanisms whereby IL-2 stimulates T-lymphocyte proliferation and differentiation. The methods described in this report should also be of general utility in the dissection of the signaling pathways activated by diverse cytokine receptors.

A monoclonal mouse anti-human IL-2 receptor antibody (anti-Tac) will recognize molecules on the surface of Xenopus laevis immunocytes which specifically bind rIL-2 and are only slightly larger than the human Tac protein

We have made visible the binding of a mouse monoclonal anti-human interleukin 2 (IL-2) receptor (anti-Tac) antibody on the surface of phytohemagglutinin (PHA)-stimulated Xenopus thymocytes using a colloidal gold-conjugated goat anti-mouse antibody and transmission electron microscopy. No binding was found when a different mouse monoclonal antibody (mAb) of the same isotype and subclass was tested, or when the anti-Tac antibody was omitted from the procedure. After metabolic radiolabeling of the IL-2 receptors with [35S]methionine using PHA-stimulated thymocytes of Xenopus laevis, the South African clawed toad, we show that a concentrated preparation of the mouse anti-human Tac antibody will immunoprecipitate a radiolabeled molecule just slightly larger than 55 kDa. Phorbol dibutyrate (PDB), an effective T cell mitogen, and cyclosporin A, an inhibitor of T cell mitogenesis in this species, are both capable of regulating the expression of this IL-2-binding molecule on Xenopus immunocytes. Here, we use the calcium ionophore A23187 to show that the relationship between IL-2 receptor expression and mitogenesis, which was previously established in X. laevis, is associated with a calcium ion flux. Flow cytometry is used for assaying alterations in epitope expression after binding the lectin-stimulated cells under test with a fluorescence (Fl*) conjugate of the anti-Tac antibody or a control mAb, which is either anti-DNP or anti-keyhole limpet hemocyanin (KLH) in specificity, but of the same mouse isotype and subclass as the anti-IL-2 receptor antibody.

Differential induction of H-2K versus H-2D class I major histocompatibility antigens by recombinant gamma interferon. Lack of Kk augmentation in a leukemia virus-induced tumor is due to a cis-dominant effect

T-T tumor hybrids were constructed between the AKR SL3 thymoma and an H-2-distinguishable thymoma cell line. Hybrids were stimulated with IFN-gamma to determine whether the differential augmentation of H-2D vs. H-2K class I antigen expression by AKR SL3 in response to IFN-gamma was due to effects cis or trans to the noninducible Kk gene. For each of a large number of hybrids tested, the expression of H-2Db, Kb, and Dk, but not Kk, was substantially enhanced by murine rIFN-gamma. These results suggested that the lack of induction of the Kk gene was due to an alteration cis to Kk rather than to the presence or absence of K region-specific, trans-acting negative or positive factors, respectively.

Toad splenocytes bind human IL-2 and anti-human IL-2 receptor antibody specifically

Human r-DNA IL-2 and fluorescent (Fl) mouse anti-human IL-2 receptor antibody have been tested separately and in competition with each other for their capacities to bind to the splenocytes of Xenopus laevis, the South African clawed toad. Binding by Fl*-mouse anti-DNP antibody of the same subclass (IgG1, kappa) was used as a control. The results of visual tests using rIL-2 coated fluorescent Covaspheres demonstrate that the human mediator will bind cells of the toad spleen. Moreover, the mediator inhibits binding of the antibody against the human IL-2 receptor, as detected by cytofluorimetry. Some of the IL-2 receptors on the toad cells appear to be constitutive, since they are expressed on freshly biopsied lymphocytes. Activation of these cells in vitro will increase the percentage of those cells able to bind both the anti-receptor antibody and rIL-2. Since the human mediator is only able to modulate in vivo immune activity in antigen-activated toads, it appears that in spite of having some constitutive IL-2 receptors, a quantitative increase in receptor expression is required before immunological behavior can be effected. More stringent controls of receptor expression may have provided an additional regulatory level as mammalian mechanisms evolved.