Review

Integration of Oxygen Signaling at the Consensus HRE

See allHide authors and affiliations

Science's STKE  18 Oct 2005:
Vol. 2005, Issue 306, pp. re12
DOI: 10.1126/stke.3062005re12

Abstract

The hypoxia-inducible factor 1 (HIF-1) was initially identified as a transcription factor that regulated erythropoietin gene expression in response to a decrease in oxygen availability in kidney tissue. Subsequently, a family of oxygen-dependent protein hydroxylases was found to regulate the abundance and activity of three oxygen-sensitive HIFα subunits, which, as part of the HIF heterodimer, regulated the transcription of at least 70 different effector genes. In addition to responding to a decrease in tissue oxygenation, HIF is proactively induced, even under normoxic conditions, in response to stimuli that lead to cell growth, ultimately leading to higher oxygen consumption. The growing cell thus profits from an anticipatory increase in HIF-dependent target gene expression. Growth stimuli–activated signaling pathways that influence the abundance and activity of HIFs include pathways in which kinases are activated and pathways in which reactive oxygen species are liberated. These pathways signal to the HIF protein hydroxylases, as well as to HIF itself, by means of covalent or redox modifications and protein-protein interactions. The final point of integration of all of these pathways is the hypoxia-response element (HRE) of effector genes. Here, we provide comprehensive compilations of the known growth stimuli that promote increases in HIF abundance, of protein-protein interactions involving HIF, and of the known HIF effector genes. The consensus HRE derived from a comparison of the HREs of these HIF effectors will be useful for identification of novel HIF target genes, design of oxygen-regulated gene therapy, and prediction of effects of future drugs targeting the HIF system.

Introduction

Oxygen availability regulates many physiological and pathophysiological processes, including embryonic development, adaptation to high altitudes, wound healing, and inflammation, as well as contributing to the pathophysiology of cancer and ischemic diseases such as infarction and stroke. The elucidation of the molecular mechanisms by which cells respond and adapt to insufficient oxygen supply (hypoxia) is central to our understanding of these processes. The last few years have brought a wealth of novel insights into these mechanisms. In particular, several oxygen-sensing protein hydroxylases have been discovered, which regulate the abundance and activity of hypoxia-inducible transcription factors and thereby the expression of effector genes involved in anticipatory metabolic changes, adaptive survival, or programmed death of the affected tissue (15).

Oxygen Sensing and Reactive HIF Regulation

The hypoxia-inducible factor (HIF) is a heterodimeric transcription factor consisting of one of three different oxygen-sensitive HIFα subunits (HIF-1α, HIF-2α, and HIF-3α) and a common constitutive HIFβ subunit. Whereas HIF-1 and HIF-2 αβ heterodimers function as transcriptional activators of oxygen-regulated target genes, the role of HIF-3α is less clear, and a short splice variant of HIF-3α, termed inhibitory PAS protein (IPAS), functions as a transcriptional repressor (13).

The partial pressure of cellular oxygen is sensed by a family of prolyl hydroxylases that covalently modify HIFα subunits (69). This enzyme family contains three members: prolyl-4-hydroxylase domain 1 (PHD1), PHD2, and PHD3, also known as HIF prolyl hydroxylase 3 (HPH3), HPH2, and HPH1, respectively (10, 11). A putative fourth family member, PH-4, affects HIF-1α and HIF-2α function only when overexpressed (12). Under normoxic conditions, HIFα is hydroxylated; hydroxylation promotes von Hippel–Lindau (VHL) tumor suppressor protein binding to HIFα, thereby targeting it for proteasomal destruction (13). Under hypoxic conditions, however, PHD activity (and thus HIFα hydroxylation) decreases. Thus, the high turnover rate of HIFα subunits enables the very rapid accumulation of HIFα under hypoxic conditions (14). Following a further decrease in oxygen availability, the asparagine hydroxylase function of another enzyme, the factor inhibiting HIF (FIH), also becomes impaired, resulting in a decrease in HIFα C-terminal hydroxylation. This decrease in C-terminal HIFα hydroxylation enables the increased recruitment of the p300 and CREB binding protein (p300/CBP) transcriptional coactivators, leading to the enhanced transcriptional activation of HIF target genes (15, 16).

The PHD and FIH oxygen sensors catalyze a complex reaction involving oxygen, ferrous iron, 2-oxoglutarate, the substrate protein, and probably ascorbate. Upon hydroxylation of target proteins, succinate and CO2 are released. Thus, protein hydroxylation is a nonreversible process, and protein hydroxylases are nonequilibrium enzymes. The multicomponent nature of the reaction per se provides the possibility of integrating input from several different signaling pathways into oxygen signaling. Indeed, it has been shown that ascorbate (17), transition metals (18, 19), reactive oxygen species (ROS) including NO (20, 21), and Krebs cycle intermediates (22, 23) all influence PHD activity. In addition, protein-protein interactions affect the abundance or activity, or both, of the PHDs and FIH, resulting in altered HIFα stability or activity, or both (Table 1).

Fig. 1.

The consensus core HRE sequence RCGTG (where R is A or G) and flanking nucleotides. Relative occurrence of the nucleotide distributions within the 108 core HREs listed in Table 4 are indicated as percentage of total. Shadowed boxes (light blue) indicate nonrandomly distributed nucleotide compositions, which occur mainly at 5′ flanking bases where HIFα subunits contact the DNA. For position −1, only one T and two C’s were identified among the 108 core HREs shown in Table 4, raising concerns about their physiological relevance.

Fig. 2.

Oxygen-dependent and oxygen-independent signaling pathways involved in HIF-regulated gene expression. Oxygen-dependent processes include direct oxygen sensing by PHDs and FIH as well as indirect effects due to changes in mitochondrial function, signaled to PHDs and FIH by second messengers such as ROS or succinate, or by the redirection of cellular oxygen. Oxygen-independent processes include reduction-oxidation (redox)–active transition metals and other compounds (Table 3) that interfere with PHD and FIH function, as well as numerous growth factors (Table 3) that alter cellular metabolism, and hence oxygen consumption, and/or activate common kinase signaling pathways, leading to increased HIFα translation or phosphorylation. All of these mechanisms result in HIFα protein stabilization and increased HIF-dependent transcription of target genes (Table 4), ultimately leading to cellular adaptation to sensed or anticipated changes in oxygen availability. FIH, factor inhibiting HIF; HRE, hypoxia-response element; MARK, mito-gen-activated protein kinase; ODD, oxygen-dependent degradation domain; PAS, Per-ARNT-Sim; PHD, prolyl hydroxylase domain containing protein; PI3K, phosphatidylinositol 3-kinase; ROS, reactive oxygen species; TA, transactivation domain; TCA, tricarboxylic acid cycle.

Table 1. Proteins that interact with PHD or FIH to regulate HIFα protein stability or HIF transcriptional activity.

Once the HIFα protein is stabilized, it heterodimerizes with the constitutively expressed HIFβ subunit, ARNT (aryl hydrocarbon receptor nuclear translocator), to form the heterodimeric transcription factor HIF. HIF binds DNA at the hypoxia response element (HRE) and recruits transcriptional coactivators such as p300/CBP. Apart from the essential protein-protein interaction between HIF α and β subunits, a multitude of other proteins are known to interact with HIFα. Table 2 provides a comprehensive list of interaction partners that affect either HIFα stability or HIF transcriptional activity, and thereby allow multiple other signaling pathways to converge on the oxygen-regulated HIF pathway.

Table 2. Proteins that interact with HIFα. Many proteins bind HIFα either to regulate HIFα stability or HIF transcriptional activity, or because they are regulated themselves by HIFα. Regulation is achieved by covalent modifications such as hydroxylation, ubiquitination, acetylation, or phosphorylation, or by steric protection from modification/degradation. Another group of interactors consists of transcriptional cofactors interacting with HIF to enhance gene transcription, often in a target gene–specific manner.

PHDs—Far More than Simple Oxygen Sensors

If the only function of the HIFα prolyl-4-hydroxylases were to sense oxygen, an excess of sensory (PHD) over effector (HIFα) protein, with its activity solely regulated by oxygen availability, could have been expected. However, molecular oxygen sensing fulfills far more functions than simply measuring oxygen at a given threshold value.

First, different tissues are differently vascularized and unequally perfused, and have spatially and temporally variable rates of oxygen consumption. Thus, the mean oxygen partial pressure varies from tissue to tissue and, within a given tissue, oxygen gradients exist from the blood vessels to the most distant oxygen-consuming cells. Nevertheless, every tissue is capable of sensing reduced oxygenation and can adequately respond to such a reduction by inducing the expression of HIF-dependent genes. Therefore, hypoxia thresholds (the partial pressure of oxygen at which the PHDs cease to effectively hydroxylate HIFα) vary both spatially and temporally; the PHD oxygen sensors, however, evolved to meet these requirements and show also a spatially and temporally variable tissue-specific expression pattern (12, 2428). How the subcellular localization of PHDs is spatially regulated within the oxygen gradient from the plasma membrane to the mitochondrial oxygen sink, and whether there exists a cellular redistribution of oxygen toward the PHDs under conditions of mitochondrial inhibition (29), are open questions that are very difficult to answer experimentally.

Second, every successful adaptation to hypoxia eventually results in reoxygenation of the affected tissue. Because there are much larger amounts of HIFα in the cell following hypoxia, the oxygen-dependent hydroxylation and degradation machinery must increase its capacity to cope with degradation of the increased protein mass. Indeed, the hypoxic increase in HIFα protein abundance is a transient process, even with overexpression of very large amounts of exogenous HIF-1α (30). PHD2 and PHD3, but not PHD1, are transcriptionally induced under hypoxic conditions, at least in part by HIF-dependent mechanisms (3137).

Third, even if the HIF system evolved exclusively for the adaptation of organisms to hypoxia, it has become evident over the last few years that both HIFα protein abundance and HIF transactivation activity can also be induced under normoxic conditions in response to various nonhypoxic stimuli (see below). How can HIFα protein stabilization occur in the presence of active PHDs and sufficient oxygen supply? The answer lies in the fact that PHDs are not present within cells in a large excess over HIFα protein but rather that there is a fine balance between HIFα production and PHD-dependent degradation. If either one is up-regulated, it overcomes the other. Thus, just as an increase in PHD synthesis leads to increased HIFα degradation, an increase in HIFα synthesis can lead to stabilization of HIFα protein and the subsequent increase in abundance, even under normoxic conditions.

Cell Growth and Proactive HIF Regulation

The finely tuned balance between oxygen-dependent HIFα protein degradation and oxygen-independent HIFα protein synthesis enables an additional level of HIF regulation. Indeed, normoxic induction of HIFα protein abundance or HIF transactivation activity by various factors, including numerous stimuli that promote cell growth, has been reported (Table 3). Because increased cell growth is associated with augmented oxygen consumption, these stimuli enable the cell to anticipate an increase in its oxygen requirements. These stimuli can be roughly grouped by their ability to induce common cellular signaling pathways, mostly pathways involving kinases, ROS, or both. The p38 mitogen-activated protein kinase (p38), p42/p44 extracellular signal–regulated kinase 1 and 2, and phosphatidylinositol 3-kinase (PI3K) pathways are often involved in HIF regulation. Intriguingly, stimulation of the PI3K pathway by heregulin or insulin-like growth factor 1 (IGF-1) results in a generalized increase in protein translation. Because up-regulation of HIFα translation can overcome its oxygen-dependent degradation, an apparently specific increase in HIFα abundance can be observed (38, 39). These prototypical experiments likely explain how HIFα content can be stabilized, even under normoxic conditions, by saturation of the degradation machinery.

Table 3. Stimuli that induce HIFα protein abundance, HIF transactivation activity, or both independently of hypoxia. Many proteins, peptides, and small molecules are known to induce HIFα under normoxic conditions. Growth factors usually increase cell growth, and small-molecule ligands, viruses, and various environmental stimuli often elicit specific cell responses, all of which are associated with increased metabolism and oxygen consumption, finally leading to hypoxia. On the other hand, some stimuli, such as NO, ROS, and certain cations, might directly interact with the PHD oxygen sensors, leading to HIFα stabilization.

In addition to signaling pathways involving kinases, numerous reports implicating ROS in HIFα protein stabilization in response to either hypoxia or stimuli that promote cell growth have appeared. Thus far, the source(s) of the ROS, and whether hypoxia leads to an increase or decrease in ROS levels, are uncertain (40). However, ROS do have the potential to interfere with the complex process of protein hydroxylation. Indeed, an increase in H2O2 observed in junD−/− cells leads to a decrease in PHD activity and hence to HIF-1α accumulation (20). Thus, this study provides an elegant explanation for the molecular mechanism whereby nontoxic ROS levels could serve as specific signaling molecules regulating HIF activity.

A reactive oxygen compound of particular interest is NO. It has been reported that NO stabilizes HIF-1α under normoxic conditions (4144). However, contradictory reports indicate that treatment with NO donors (as well as with CO), or ectopic expression of inducible NO synthase (itself a HIF-1 target), both abrogated HIF-1 activity (4548). This discrepancy could be explained by the finding that NO effects appear to be transient: Initial NO treatment increased, whereas prolonged NO treatment decreased, HIF-1α protein abundance and the resulting increase in expression of erythropoietin under hypoxic conditions (49). Indeed, NO inhibits PHD activity (21). Thus, an initial increase in HIFα abundance might be counteracted by a subsequent HIF-dependent increase in PHD expression. Moreover, NO inhibits mitochondrial respiration and might thus lead to the redistribution of O2 from mitochondria to the PHD oxygen sensors (29). (It is important to note that, for technical reasons, intracellular oxygen levels have never been measured, and thus an actual redistribution of O2 has not be demonstrated experimentally.) However, a recent report showed that, upon inhibition of the mitochondrial electron transport chain under mild hypoxic conditions (3% O2), HIF-1α stabilization is impaired in normal (oxygen-impermeable) cell culture dishes but not in oxygen-permeable dishes. HIF-1α stabilization also remained intact upon inhibition of the mitochondrial electron transport chain under conditions of severe hypoxia (0.1% O2), regardless of the type of cell culture dish (50). Similar data were obtained in cells deficient in the p38 pathway, in which mitochondria-dependent HIF-1α induction was impaired in hypoxia but not in anoxia (51). These data suggest that enough oxygen could be redirected toward the PHD oxygen sensors in mild hypoxia or in gas-impermeable dishes to allow HIFα hydroxylation. But when oxygen is too rapidly equilibrated through the gas-permeable dishes or when hypoxia is too severe, insufficient amounts of oxygen are saved by mitochondrial dysfunction to allow HIFα hydroxylation and subsequent degradation.

In conclusion, numerous nonhypoxic stimuli lead to the tissue-specific activation of a few common signaling pathways that normally result in either the translational induction of HIFα or the inhibition of HIFα hydroxylation. In any case, the normoxic induction of the HIF system is usually only transient, because HIF-dependent PHD up-regulation forms a negative-feedback loop, down-regulating HIFα under conditions of excess oxygen supply.

The Consensus HRE Sequence

Once stabilized and activated, HIF binds to the consensus HRE, which is present in the oxygen-regulated elements of 70 known HIF target genes. Microarray experiments indicate that far more than 200 HIF target genes might exist; however, not all of these genes are likely to be directly regulated by an HRE in their regulatory regions. Rather, other HIF-dependent or -independent oxygen-regulated transcription factors might be responsible for their hypoxic induction. In this Review, the actual HIF DNA binding site within the HRE is referred to as the core HRE, whereas we use HRE to refer to the minimal cis-regulatory element required for hypoxic induction of gene transcription. A single core HRE is necessary but not sufficient for efficient gene activation in response to hypoxia. Although the core HRE is the minimal DNA domain required for interaction with HIF, a fully functional HRE usually contains neighboring DNA binding sites for additional transcription factors. These transcription factors are not necessarily hypoxia inducible, but they might amplify the hypoxic response or confer tissue-restricted activity to an HRE. For instance, HIF-1 cooperates with ATF-1 and CREB-1 to transcriptionally activate the lactate dehydrogenase A gene (52, 53) or with AP-1 binding factors to activate the gene encoding vascular endothelial growth factor (VEGF) (54). Tandemly arrayed core HREs can also form a functional HRE. Two or three adjacent core HREs were found, for example, in the genes encoding transferrin, several glycolytic enzymes, and glucose transporter–1.

Table 4 provides a comprehensive overview of the 70 known HIF-dependent target genes. Figure 1 summarizes the frequency with which particular nucleotides occur in a given position of the core HRE on the basis of the 107 core HREs listed in Table 4. The resulting mandatory consensus core HRE sequence is CGTG, with particular nucleotides occurring with nonrandom frequency in the surrounding positions, especially in the 5′ flanking bases. To determine the expected nucleotide frequencies for each position, mean values of the apparently randomly distributed nucleotides in the 3′ flanking positions +6 to +12 (Fig. 1) were calculated. In addition to the core HRE sequence CGTG, several other positions show a nonrandom distribution of nucleotides when compared to these mean values. In particular, A occurs at the −1 position 4.5 times as often as expected. Moreover, the −3 position shows a 4.2-fold underrepresentation of T. Positions −5 and −2 show a 2.0- and 1.8-fold underrepresentation of T and A, respectively, and position +5 reveals a 1.7-fold overrepresentation of C (Fig. 1).

Table 4. Compilation of the known HIF target genes. Only those genes were included in which binding of HIF to the target DNA sequence in a DNA binding (DB) assay or functional transactivation (TA) of reporter gene expression, or both, have been shown. FS, flanking sequence; IVS, intervening sequence; UT, untranslated region. For abbreviations of gene names, refer to the references.

A systematic mutational analysis of the −2 position in the context of the ACGTG core HRE revealed that it can confer different transcriptional activation by HIF under 0.5% O2 conditions with the preference T >> G > C (55). Although the functional relevance is unknown, HIFβ (also known as ARNT) can form homodimers that bind the palindromic sequence CACGTG (56). This might explain at least partially why this particular core HRE is less able to mediate transcriptional responses to hypoxia than are core HREs consisting of TACGTG or GACGTG. It also implies that HIFβ (ARNT) binds the 3′ GTG half-site, whereas its interaction partner binds the 5′ half-site. In the case of the dioxin receptor, the 5′ half-site is TNG (56), whereas HIFα preferentially binds NAC. However, Fig. 1 suggests that additional 5′ nucleotides confer specificity and probably also activity to the HRE. Indeed, A at the −2 position is associated with a decrease in HRE activity in the endothelin-1 and Flt-1 genes (57).

Whether and how HIF-1 and HIF-2 distinguish between target genes is still an open question. A recent report showed that in VHL-defective cells, different sets of genes are regulated by HIF-1 or HIF-2 (58). Few data are available on the specific interaction of HIF-2 with the core HRE. The VEGF receptor Flk-1 is apparently regulated specifically by HIF-2 (59). However, no obvious difference in the core HRE composition compared to that of HIF-1 target genes could be observed (Table 4 and Fig. 1). Intriguingly, even though HIF-1 initially was thought to be the transcription factor that activated erythropoietin transcription, it turned out that HIF-2 rather than HIF-1 regulates erythropoietin expression in the kidney (60, 61). Thus, the distinction between HIF-1 and HIF-2 target genes does not seem to depend on core HRE composition, but rather on cooperation with other transcription factors and tissue-specific HIFα expression patterns.

The Myc-binding palindromic CACGTG sequence, or E-box, is the second most common motif in human gene promoters (62). Even if HIF binding does not typically occur at the palindromic E-box, 22 out of 107 known core HREs listed in Table 4 are in fact E-boxes. This implies that more core HRE-containing, potential HIF target genes exist than have so far been recognized. However, there are many more potential core HRE sequences in mammalian gene promoters than are actually used by HIF to regulate gene expression. So, what allows HIF to distinguish between functional and nonfunctional HREs when the core HRE sequence is identical? One explanation might be the occurrence of a HIF-1 ancillary sequence (HAS) 8 base pairs distant from the core HRE (63). Because the HAS is not well conserved and cannot be found in most HIF target genes, it is likely that other transcription factors, which bind nearby, cooperate with HIF in a tissue- and gene-specific manner.

Epigenetic Mechanisms Regulating HIF Binding to the Core HRE

Apart from the requirement for cooperation among transcription factors, epigenetic effects may also reduce the large number of putative core HREs to relatively few functional HREs. Methylation of the CpG dinucleotide (containing 5-methylcytosine) plays an important regulatory role in mammalian gene expression, contributing to X-chromosome inactivation and genomic imprinting, as well as tissue- and developmental stage-specific transcriptional regulation. CpG dinucleotides are underrepresented in the mammalian genome and are usually methylated if located outside of GC-rich "CpG islands." The consensus core HRE contains a CpG dinucleotide, thus lowering the number of actual HREs found in the genome compared to the frequency of a random tetranucleotide sequence. Methylated CpG interferes with transcription factor binding to DNA through both direct steric hindrance and the binding of repressor proteins. Indeed, HIF binding to the core HRE is blocked by 5-methylcytosine (64). The promoters of many HIF-dependent target genes thus overlap with methylation-free CpG islands, as confirmed by the high mean GC content (64.2%) of the core HREs listed in Table 4. Because HREs are usually identified experimentally by reporter gene or gel-shift assays on nonmethylated bacterial or synthetic DNA, the data presented in Table 4 should be interpreted cautiously. Even the best in vitro HRE is worthless if its endogenous counterpart is CpG methylated in the chromatin.

It has been shown for the erythropoietin gene that CpG methylation of the 3′ HRE abrogates both HIF DNA binding and reporter gene activation, and that erythropoietin expression is inversely related to the degree of CpG methylation of the erythropoietin 3′ HRE (64, 65). In contrast to most HIF-1–dependent genes, in which the HREs are located in methylation-free CpG islands, the erythropoietin 3′ HRE is present in a locus with average GC content and suppressed CpG dinucleotide frequency. Such loci are usually methylated and hence "silenced" in mammalian genomes. Thus, there must be a selective pressure to keep this site methylation free. This could be the constitutive binding of a transcription factor to the HRE. Because HIF activity can only be detected under hypoxic conditions, the specific constitutive binding of ATF-1 and CREB-1 (66, 67) to the HRE might prevent CpG methylation. In support of this model, in vivo footprinting shows that the HRE is also occupied in normoxia in the absence of HIF (68).

Oxidative DNA damage caused by ROS might represent another epigenetic modification regulating HRE accessibility. Intriguingly, ROS induced in response to hypoxia oxidize particular bases within specific DNA sequences of the HIF target gene VEGF. The most frequently modified nucleotide is the terminal guanine of the VEGF core HRE (ACGTGGG). HREs with oxidative base modifications are associated with increased binding of HIF-1 and Ref-1 (also known as Ape1) transcription factors, leading to increased reporter gene expression (69). Another hot spot for oxidative DNA damage was detected in the HRE of the HIF target gene phosphoglycerate kinase 1 (PGK1) (70). Preferential cleaveage of the sequence RTGR (where R is G or A) has been suggested to result from sequence-specific Fe2+ association and subsequent localized Fenton reaction with H2O2 (71).Taken together, these results suggest another mechanism whereby oxygen-dependent ROS production could modulate the HRE. Because the base modifications occur close to but not directly at the consensus core HRE, different HREs are likely to be unequally affected by ROS, providing a means of variable modulation of the efficiency of gene-specific transcriptional induction.

Conclusion

Figure 2 provides a simplified scheme of the most important oxygen-dependent and oxygen-independent signaling pathways involved in HIF-regulated gene expression. The determination of a consensus core HRE and a detailed knowledge of its regulation will be useful for at least two therapeutic strategies: First, the HRE can be used to amplify the efficiency of gene therapy by targeting hypoxic tissue with HRE-driven expression vectors (72). Second, precise identification of the core HRE sequence is essential for the development of drugs that target endogenous core HRE sequences to impair specifically target gene expression (73).

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 191.
  192. 192.
  193. 193.
  194. 194.
  195. 195.
  196. 196.
  197. 197.
  198. 198.
  199. 199.
  200. 200.
  201. 201.
  202. 202.
  203. 203.
  204. 204.
  205. 205.
  206. 206.
  207. 207.
  208. 208.
  209. 209.
  210. 210.
  211. 211.
  212. 212.
  213. 213.
  214. 214.
  215. 215.
  216. 216.
  217. 217.
  218. 218.
  219. 219.
  220. 220.
  221. 221.
  222. 222.
  223. 223.
  224. 224.
  225. 225.
  226. 226.
  227. 227.
  228. 228.
  229. 229.
  230. 230.
  231. 231.
  232. 232.
  233. 233.
  234. 234.
  235. 235.
  236. 236.
  237. 237.
  238. 238.
  239. 239.
  240. 240.
  241. 241.
View Abstract

Stay Connected to Science Signaling

Navigate This Article