PERSPECTIVES
`
`O P I N I O N
`
`mTOR and cancer: insights into a
`complex relationship
`
`David M. Sabatini
`
`Abstract | mTOR (mammalian target of rapamycin) has come a long way since its
`humble beginnings as a kinase of unknown function. As part of the mTORC1 and
`mTORC2 complexes mTOR has key roles in several pathways that are involved in
`human cancer, stimulating interest in mTOR inhibitors and placing it on the radar of
`the pharmaceutical industry. Here, I discuss the rationale for the use of drugs that
`target mTOR, the unexpectedly complex mechanism of action of existing mTOR
`inhibitors and the potential benefits of developing drugs that function through
`different mechanisms. The purpose is not to cover all aspects of mTOR history and
`signalling, but rather to foster discussion by presenting some occasionally
`provocative ideas.
`
`In response to growth factors and nutrients
`mTORC1 (mammalian target of rapamy-
`cin complex 1) regulates cell growth by
`modulating many processes, including
`protein synthesis, ribosome biogenesis and
`autophagy (reviewed in REF. 1). mTORC1
`is a heterotrimeric protein kinase that
`consists of the mTOR catalytic subunit and
`two associated proteins, raptor (regulatory-
`associated protein of mTOR) and mLST8
`(also known as GβL) (BOX 1). The molecular
`mechanisms that regulate mTORC1 kinase
`activity are still poorly understood, but it is
`increasingly clear that many if not most can-
`cer-promoting lesions activate the mTORC1
`pathway (FIG. 1). Most dramatically, the TSC1
`(tuberous sclerosis 1, also known as harmar-
`tin)–TSC2 (also known as tuberin) tumour
`suppressor complex — the inactivation of
`which causes the tumour-prone syndrome
`tuberous sclerosis complex (TSC) and the
`related disease lymphangio leiomyomatosis
`(LAM) — has emerged as a key negative
`regulator of mTORC1 (REFS 2,3). The
`TSC1–TSC2 heterodimer is a GTPase-
`activating protein for Rheb (Ras homologue
`enriched in brain)4–8, a GTP-binding
`protein that activates mTORC1, most
`probably by binding to it9. TSC1–TSC2
`and Rheb also have important roles in the
`
`activation of mTORC1 that occurs when
`cells lose the PTEN (phosphatase and tensin
`homologue), NF1 (neurofibromatosis 1),
`LKB1 (also known as serine–threonine
`kinase 11) or p53 tumour suppressors10–15. In
`all cases, inactivation of the tumour suppres-
`sor triggers a pathway that eventually leads
`to inhibition of TSC1–TSC2. For example,
`the loss of PTEN activates Akt (also known
`as protein kinase B), which then directly
`phosphorylates and inhibits TSC1–TSC2,
`whereas the loss of LKB1 suppresses AMPK
`(AMP-activated protein kinase)16,17, which
`normally mediates an activating phosphory-
`lation of TSC1–TSC2 (REF. 18).
`The mTORC1 pathway regulates
`growth through downstream effectors,
`such as the regulators of translation 4EBP1
`(eukaryotic translation initiation factor 4E
`binding protein 1) and S6K1 (ribosomal S6
`kinase 1) (reviewed in REF. 19). In addition
`to its role in promoting protein synthesis,
`S6K1 represses the phosphatidylinositol
`3-kinase (PI3K)–Akt pathway by inhibit-
`ing IRS1 (insulin receptor substrate 1) and
`IRS2 expression20–24. Therefore, an active
`mTORC1 pathway can suppress PI3K–Akt
`signalling, helping to explain the non-
`aggressive nature of the tumours that are
`found in TSC25,26. The opposite is also true:
`
`inhibition of mTORC1 activates PI3K–Akt
`signalling and, as described below, the
`activation of PI3K–Akt that is caused by
`mTORC1 inhibitors might significantly
`diminish the anti-tumour activity of such
`molecules.
`mTORC2 also contains mTOR and
`mLST8 but, instead of raptor, it contains
`two proteins, rictor (rapamycin-insensitive
`companion of mTOR) and mSin1 (also
`known as mitogen-activated-protein-
`kinase-associated protein 1), that are not
`part of mTORC1 (BOX 1). This second
`mTOR-containing complex is less under-
`stood than mTORC1 but recent work
`indicates that it should be considered part
`of the PI3K–Akt pathway as it directly
`phosphorylates Akt27,28 on one of the two
`sites that are necessary for Akt activation in
`response to growth-factor signalling (FIG. 1).
`This finding makes mTORC2 a key part of
`the pathway that activates Akt and, like PDK1
`(3-phosphoinositide-dependent protein
`kinase 1) and PI3K, a potential drug target for
`cancers in which there is Akt deregulation.
`The Akt-activating function of mTORC2
`sets up the intriguing situation in which
`mTOR, as part of two distinct complexes, is
`potentially both ‘upstream’ and ‘downstream’
`of itself. mTORC2 has other functions besides
`activating Akt, such as regulating the cyto-
`skeleton29,30, but the implications for cancer of
`these roles are still unknown.
`
`What does rapamycin do to the mTORCs?
`mTOR was discovered in the early 1990s
`in studies into the mechanism of action of
`rapamycin (also known as sirolimus), which
`is a macrolide that was originally found as
`an antifungal agent and was later recog-
`nized as having immunosuppressive and
`anticancer properties. Even today, exactly
`how rapamycin perturbs mTOR function
`is not completely understood. The complex
`of rapamycin with its intracellular receptor
`FKBP12 binds directly to mTORC1 and, at
`least in vitro, suppresses mTORC1-mediated
`phosphorylation of the substrates S6K1
`and 4EBP1. Rapamycin also weakens the
`interaction between mTOR and raptor31,
`which is a component of mTORC1 that
`can recruit substrates to the mTOR kinase
`domain32–34. It is not known if mTORC1 has
`
`NATURE REVIEWS | CANCER
`
` VOLUME 6 | SEPTEMBER 2006 | 729
`
`© 2006 Nature Publishing Group
`
`Ex. 1105-0001
`
`

`
`P E R S P E C T I V E S
`
`Box 1 | The mTORC1 and mTORC2 complexes
`
`mTORC2
`rictor
`mSin1
`mLST8
`mTOR
`
`mTORC1
`
`raptor
`mLST8
`mTOR
`
`mTOR (mammalian target of rapamycin) is a
`large protein kinase that nucleates at least two
`distinct multi-protein complexes — mTORC1
`and mTORC2 (REFS 29–32,92,93). The first
`evidence for the existence of the two
`complexes came from work in budding yeast, in which two related proteins TOR1p and TOR2p
`can both participate in complexes that are analogous to those found in mammals93.
`mTORC1 has three components — the mTOR catalytic subunit and two other proteins, raptor
`(regulatory-associated protein of mTOR) and mLST8 (also known as GβL)29–32,92,93. mTOR contains a
`serine–threonine protein kinase domain near its C terminus and there is no evidence that
`mTORC1 contains any other enzymatic function besides kinase activity. Raptor and GβL are
`evolutionarily conserved but their functions are still poorly understood. Raptor might have roles
`in mTOR assembly, recruiting substrates to mTOR, and in regulating mTOR activity. The strength
`of the association between mTOR and raptor is regulated by nutrients and other signals that
`regulate the mTORC1 pathway, but how this translates into regulation of the mTORC1 pathway is
`unknown. The small GTP-binding protein Rheb (Ras homologue enriched in brain) binds to the
`mTOR kinase domain and seems to have a key role in activating it.
`mTORC2 also contains mTOR and mLST8 but instead of raptor two other proteins, rictor
`(rapamycin-independent companion of mTOR) and mSin1 (also known as mitogen-activated-
`protein-kinase-associated protein 1). Both rictor and mSin1 are necessary for the phosphorylation
`of Akt (also known as protein kinase B) on its C-terminal hydrophobic motif and this function is
`conserved in Drosophila. Compared with raptor and mLST8, rictor, and particularly mSin1, are
`poorly conserved at the amino-acid-sequence level. Recent work indicates that mTORC2 exists in
`several distinct forms that are defined by different alternatively spliced isoforms of mSin1. Unlike
`the raptor–mTOR association, the interaction between mTOR and rictor does not seem to be
`regulated by upstream signals. However, growth factors do stimulate the mTORC2 kinase activity
`but the mechanism of regulation is not yet understood. In vitro, rictor is required for mTORC2 to
`be able to phosphorylate Akt.
`
`mTORC2 might help explain why the
`cellular effects of rapamycin vary among
`cancer cell lines. Moreover, in a tumour this
`inhibition might have the beneficial effect
`of preventing the activation of Akt, through
`inhibition of S6K1 (FIG. 1), that rapamycin
`would otherwise cause.
`
`Anticancer uses for mTOR inhibitors
`Rapamycin and its analogues can inhibit
`several processes that are relevant to the anti-
`tumour properties that these molecules exert
`in pre-clinical cancer models, including cell
`proliferation, cell survival and angiogenesis
`(reviewed in REF. 35). Exactly how mTORC1
`inhibition mediates all these varied effects
`is not well worked out and the potential
`for rapamycin to inhibit mTORC2 and Akt
`provides additional mechanisms to consider.
`A case in point is the effects of rapamycin on
`apoptosis, which vary depending on which
`cell line is tested. There are many reports
`of rapamycin promoting pro-apoptotic
`stimuli39–44 but there are also reports of it
`promoting cell survival45. As rapamycin uni-
`versally inhibits the mTORC1 pathway, its
`effects on apoptosis might correlate with its
`varying effects on Akt, a well-known regula-
`tor of cell survival. In cells in which the drug
`inhibits mTORC2 and Akt it might promote
`apoptosis, as has been shown36. On the
`other hand, when the drug does not inhibit
`mTORC2, so that mTORC1 inhibition leads
`to Akt activation, the drug might protect
`against apoptosis. As induction of apoptosis
`rather than cytostasis is increasingly consid-
`ered a prerequisite for an effective anticancer
`agent, it will be crucial to understand when
`rapamycin has such effects and where it does
`not, and to learn how to trigger apoptosis
`with additional therapies.
`Despite the substantial pre-clinical
`data indicating that rapamycin and its
`analogues have anti-tumour effects and that
`mTOR participates in many cancer-related
`pathways, these molecules have not shown
`universal anti-tumour activity in early clini-
`cal trials. Response rates vary among cancer
`types from a low of less than 10% in patients
`with glioblastomas46,47 or advanced renal-cell
`cancer48 to a high of around 40% in patients
`with mantle-cell lymphoma (MCL; an
`aggressive non-Hodgkin lymphoma with a
`poor prognosis)49. Many in the community
`have found these results disappointing, but
`until we understand why rapamycin ana-
`logues do have significant anti-tumour effects
`in certain patients it is too early to draw a
`conclusion on the utility of inhibiting mTOR
`in cancer treatment. Clearly, we require
`more information on which combination of
`
`functions that depend on its kinase activity
`but are not sensitive to rapamycin, so it is
`still unclear if a molecule that directly inhib-
`ited the mTORC1 kinase domain would
`have different biological effects to those
`of rapamycin. Analogues of rapamycin,
`such as CCI-779 (also known as temsiro-
`limus; Wyeth), RAD001 (also known as
`everolimus; Novartis) and AP23573 (Ariad
`Pharmaceuticals), are likely to be the first
`mTOR-perturbing molecules to be approved
`for anticancer use in humans (reviewed in
`REF. 35). These molecules inhibit mTORC1
`through the same mechanism of action as
`rapamycin, but have different pharmaco-
`kinetic and solubility properties that increase
`their desirability for clinical use.
`In contrast to mTORC1, FKBP12–
`rapamycin cannot bind directly to mTORC2
`(REFS 29,30), suggesting that the effects of
`rapamycin on cellular signalling are due
`to inhibition of mTORC1. A potentially
`important wrinkle in this seemingly closed
`story has recently emerged36. It turns out
`that prolonged treatment with rapamycin
`— clearly a situation that is relevant to its use
`in patients — perturbs mTORC2 assembly
`and, in about 20% of cancer cell lines, the
`drop in intact mTORC2 levels is sufficient
`to strongly inhibit Akt signalling (FIG. 2). The
`binding of FKBP12–rapamycin to mTOR
`seems to block the subsequent binding of
`
`the mTORC2-specific components rictor36
`and mSin1 (REF. 37) but it is unknown why
`in certain cell types rapamycin only partially
`inhibits mTORC2 assembly. No absolute
`correlation exists between the tissue of origin
`of a cell line and the sensitivity of mTORC2
`formation to rapamycin, although many cell
`lines with this property are derived from
`the haematological system. Recent work
`provides the first evidence that mTORC2
`function can be rapamycin-sensitive in
`patients. In more than 50% of patients with
`acute myeloid leukaemia, rapamycin or an
`analogue inhibited Akt phosphorylation in
`primary leukaemic cells and the inhibition
`correlated with the loss of intact mTORC2
`(M. Konopleva, personal communication).
`So, rapamycin and its analogues are
`universal inhibitors of mTORC1 and S6K1,
`and cell-type specific inhibitors of mTORC2
`and Akt. As the inhibition of mTORC2 by
`rapamycin is time and dose dependent36,38,
`Akt activity in tumours will vary with
`the length of rapamycin treatment and
`the dosing regimen (FIG. 2). It is important
`to keep in mind that, because inhibi-
`tion of mTORC1 and mTORC2 will not
`always occur at the same time, markers
`of mTORC1 inhibition, such as loss of
`phosphorylated S6, will not necessarily
`reflect mTORC2 activity. As discussed
`below, the capacity to sometimes inhibit
`
`730 | SEPTEMBER 2006 | VOLUME 6
`
` www.nature.com/reviews/cancer
`
`© 2006 Nature Publishing Group
`
`Ex. 1105-0002
`
`

`
`molecular lesions is likely to make a tumour
`susceptible to mTOR inhibition.
`As discussed below, good scientific
`reasons are emerging as to why rapamycin
`might benefit particular tumour types, and
`the hope is that with the proper insights this
`drug or other mTOR inhibitors might be
`used for patient benefit.
`
`TSC. A strong scientific rationale exists
`for the use of rapamycin and its analogues
`in the treatment of TSC. Rapamycin sup-
`presses the molecular consequences of
`TSC1–TSC2 loss on the mTORC1 pathway
`and, in cultured cells and model organ-
`isms, the drug also reverses the increase
`in cell size that is a hallmark of the disease
`
`DNA damage
`
`NF1
`
`LKB1
`
`p53
`
`Ras
`
`AMPK RSK
`
`ERK
`
`Energy
`depletion
`Hypoxia
`HIF1α
`
`Redd
`
`TSC1/2
`
`Amino
`acids
`
`Rheb
`
`Insulin
`receptor
`
`NF1
`
`IRS1
`
`Ras
`
`TK
`receptors
`
`PI3K
`
`PTEN
`
`mTORC2
`
`PDK1
`
`Stress
`
`mTORC1
`
`FKBP12-
`rapamycin
`
`?
`
`PKCα RAC1
`
`Akt
`
`S6K1
`
`?
`
`S6K2 4EBP
`
`CLIP-170
`
`MDM2
`
`GSK3
`
`Foxo
`
`eIF4E
`
`p53
`
`Survival
`
`Metabolism
`
`Proliferation
`
`Autophagy
`
`Translation
`
`Ribosome
`biogenesis
`Figure 1 | Circuitry of the mTORC1 and mTORC2 pathways and their relationships to the PI3K
`pathway. The main points of contact between the mTORC1 (mammalian target of rapamycin complex
`1) and PI3K (phosphatidylinositol 3-kinase)–Akt (also known as protein kinase B) pathways are empha-
`sized. A main function of the mTORC1 pathway is to regulate the accumulation of cell mass by activat-
`ing mRNA translation and ribosome biogenesis and by inhibiting autophagy. mTORC1 directly phos-
`phorylates and activates S6K1 (ribosomal S6 kinase 1), which is an important regulator of cell size.
`Phosphorylation of S6K1 by PDPK1 (3-phosphoinositide-dependent protein kinase 1) is also important
`for its activation, but for clarity this connection is not shown. S6K1 inhibits IRS1 (insulin receptor
`substrate 1) by directly phosphorylating it, a connection that in the mTOR field is frequently called ‘the
`feedback loop’ and is responsible for the inhibition of Akt that is caused by high mTORC1 activity. By
`phosphorylating the 4EBP (eukaryotic translation initiation factor 4E binding protein) family of proteins
`mTORC1 represses their capacity to inhibit the mRNA cap-binding protein eIF4E (eukaryotic initiation
`factor 4E). Less is known about how mTORC1 activates S6K2 and CLIP-170 (cytoplasmic linker protein
`170, also known as restin) and it is likely that many direct substrates of mTORC1 remain to be discov-
`ered. The TSC1 (tuberous sclerosis 1)–TSC2 heterodimer is a key negative regulator of mTORC1 that
`functions by suppressing Rheb (Ras homologue enriched in brain), a small GTP-binding protein that
`activates mTORC1. Mammals contain two Rhebs, RHEB1 and RHEB2, which can both activate mTORC1
`signalling. Insulin and other growth factors, energy status and DNA damage signal to TSC1–TSC2 by
`regulating kinases that directly phosphorylate TSC2. Hypoxia induces the expression of REDD1
`(regulated in development and DNA-damage responses 1) and REDD2, which activate TSC1–TSC2
`through an unknown mechanism. It is unknown how osmotic and heat-shock stress, as well as amino
`acids, signal to mTORC1 and it might be that mechanisms apart from the Redds are involved in the
`regulation of mTORC1 by hypoxia. mTORC2 directly phosphorylates Akt on the hydrophobic site in
`the C-terminal tail, which together with the PDK1-mediated phosphorylation of the activation loop
`is necessary for full Akt activation. How mTORC2 is regulated is unknown but its activity does respond
`to growth factors; this is mediated through tyrosine kinase (TK) receptors. mTORC2 can be considered
`upstream of mTORC1 because by activating Akt it leads to the inhibition of TSC1–TSC2, which causes
`the activation of Rheb and mTORC1. AMPK, AMP-activated protein kinase; ERK, extracellular signal-
`regulated kinase; FKBP12, intracellular receptor for rapamycin; Foxo, Forkhead box; GSK3, glycogen
`synthase kinase 3;HIF1α, hypoxia-induced factor 1α; LKB1, serine–threonine kinase 11; MDM2, mouse
`double minute 2; NF1, neurofibromatosis 1; PKCα, protein kinase Cα; PTEN, phosphatase and tensin
`homologue; RAC1, Ras-related C3 Botulinum toxin substrate 1; RSK, ribosomal protein S6 kinase.
`
`P E R S P E C T I V E S
`
`(reviewed in REF. 35). In a recent clinical trial
`rapamycin reduced the sizes of the astrocy-
`tomas that are frequently seen in patients
`with TSC50, providing the first human data
`that supports the widely held expectation
`that rapamycin will be a useful drug for TSC.
`Because mTORC1 is at least two molecules
`downstream of the TSC1–TSC2 complex
`it is unlikely that rapamycin will reverse all
`TSC-associated phenotypes. It is known that
`TSC1–TSC2 has targets besides Rheb51,52
`and it is likely that Rheb has targets in addi-
`tion to mTORC1 (REFS 53–57), potentially
`allowing TSC1–TSC2 loss to cause many
`mTORC1-independent sequelae. There is
`already evidence that this is the case as rapa-
`mycin cannot reverse the dendritic-spine
`elongation that is seen in neurons that lack
`TSC2 (REF. 58), or the resistance of TSC2-null
`fibroblasts to hypoxia-induced apoptosis59.
`The role of mTORC1-independent pathways
`in disease pathogenesis remains to be
`determined but will surely be a topic of great
`interest if not all clinical features of TSC
`prove sensitive to mTORC1 inhibitors.
`
`Tumours with activated PI3K–Akt signal-
`ling. Data from cancer cell lines in vitro and
`from xenografts indicate that a strong cor-
`relation exists between the antiproliferative
`effects of the rapamycin analogues and the
`loss of PTEN60,61. Although this correla-
`tion is not perfect, work in mouse models
`bolsters the idea that rapamycin might
`be particularly effective against tumours
`with an activated PI3K–Akt pathway.
`Rapamycin or an analogue blocked both
`prostate intraepithelial neoplasia62 and the
`lymphoproliferative disease63 that is caused
`by expression of an activated allele of Akt.
`These findings indicate that tumorigenesis
`that is driven by a hyperactive PI3K–Akt
`pathway requires the activation of mTORC1
`by Akt. Unfortunately, the situation is not
`as straightforward in patients because
`rapamycin analogues have not shown good
`anti-tumour activity against tumours that
`are known to have high Akt activity, such as
`glioblastomas46,47 and breast cancers64.
`An interesting hypothesis is emerging as
`to why this might be. As described earlier,
`by inhibiting mTORC1 rapamycin and its
`analogues are expected to strongly activate
`Akt, a prediction that has now been observed
`in many cancer cell lines in vitro65,66 and in
`tumours in patients66. Inhibition of PI3K
`signalling blocks rapamycin-mediated activa-
`tion of Akt in cancer cells65–67, suggesting
`a possible strategy for boosting the anti-
`tumour efficacy of mTORC1 inhibitors65–67.
`Consistent with this idea, the combination of
`
`NATURE REVIEWS | CANCER
`
` VOLUME 6 | SEPTEMBER 2006 | 731
`
`© 2006 Nature Publishing Group
`
`Ex. 1105-0003
`
`

`
`P E R S P E C T I V E S
`
`rapamycin and an inhibitor of IGF1R
`(insulin-like growth factor 1 receptor) pre-
`vents Akt activation in various human cancer
`lines and has a greater antiproliferative effect
`than rapamycin alone66. Similar antiprolifera-
`tive effects occur in multiple glioma cell lines
`that have been treated with PI-103 (REF. 67),
`which is a molecule that inhibits the kinase
`activity of both mTOR and PI3K p110α (the
`isoform of the PI3K catalytic subunit that
`
`activates Akt in response to insulin68,69). Even
`though PI-103 inhibits mTORC1, mTORC2
`and PI3K p110α it has anti-tumour activity
`in mice without overt toxicity67.
`In cancer cells in which rapamycin
`inhibits both mTORC1 and mTORC2, the
`drug inhibits Akt instead of activating it36.
`This phenomenon seems to occur in only a
`minority of cancer lines36 and perhaps many
`of the tumours that do respond to rapamycin
`
`a Cells with mTORC2 assembly that is completely sensitive to rapamycin
`
`No rapamycin
`
`Rapamycin for 1 hour
`
`Rapamycin for 24 hours
`
`S6K1
`
`Akt
`
`S6K1
`
`Akt
`
`S6K1
`
`Akt
`
`b Cells with mTORC2 assembly that is partially sensitive to rapamycin
`
`No rapamycin
`
`Rapamycin for 1 hour
`
`Rapamycin for 24 hours
`
`S6K1
`
`Akt
`
`S6K1
`
`Akt
`
`S6K1
`
`Akt
`
`mTORC2
`
`FKBP12–
`rapamycin
`
`mTORC1
`
`Dissociated mTORC2
`
`Figure 2 | Two models to explain the varying effects of long-term rapamycin treatment on Akt
`activity. a | In this scenario, the assembly of mTORC2 (mammalian target of rapamycin complex 2) is
`completely sensitive to rapamycin treatment — 24 hours after rapamycin addition no intact mTORC2
`remains in the cell. Therefore, Akt (also known as protein kinase B) phosphorylation does not occur and
`its activity drops. After 1 hour of rapamycin treatment the drug inhibits only mTORC1. This eliminates
`the inhibitory signal that is normally mediated by S6K1 (ribosomal S6 kinase 1) to IRS1 (insulin recep-
`tor substrate 1), which suppresses the activity of the PI3K (phosphatidylinositol 3-kinase)–Akt pathway.
`Therefore, Akt activity increases with short rapamycin treatment times but is inhibited by prolonged
`treatment. b | In this scenario, Akt activity also increases after 1 hour of treatment but mTORC2 assem-
`bly is not completely sensitive to rapamycin, so some mTORC2 remains intact even with prolonged
`treatment and Akt activity remains at increased (shown) or at baseline (not shown) levels. Only about
`20% of cancer cell lines seem to have mTORC2 assembly that is completely sensitive to rapamycin. The
`size of the icons that represent S6K1 and Akt depicts their activity at different times after rapamycin
`treatment. FKBP12, intracellular receptor for rapamycin.
`
`monotherapy have drug-sensitive mTORC2
`activity and depend on PI3K–Akt signalling.
`To test this hypothesis it will be necessary
`to develop biomarkers that predict in which
`tumours rapamycin will inhibit Akt and to
`understand the molecular mechanisms that
`confer this phenotype. Because rapamycin-
`mediated inhibition of mTORC1 activates
`the PI3K–Akt pathway, the relative strength
`of this activation versus the degree of Akt
`suppression that is caused by inhibition of
`mTORC2 assembly might set the ultimate
`levels of Akt activity in a rapamycin-treated
`cell. Of course, it is probable that the insensi-
`tivity of certain tumours to rapamycin does
`not depend on the inherent sensitivity of
`mTORC2 assembly to the drug. Rather, as
`yet unidentified mutations in tumour cells
`might determine how important mTORC1
`signalling is to the proliferation and survival
`of a particular cancer cell.
`
`Tumours with VEGF-driven angiogenesis
`or VHL loss. As mentioned earlier, rapa-
`mycin suppresses angiogenesis70, indicating
`that mTOR inhibition might be useful
`in tumours in which this is an important
`component of the pathogenesis. A striking
`example is the regression caused by rapa-
`mycin of Kaposi sarcoma71,72, a tumour in
`which VEGF (vascular endothelial growth
`factor)-driven angiogenesis is a prominent
`feature (reviewed in REF. 73). Recent work
`indicates that rapamycin inhibits activated
`Akt signalling in endothelial cells and sup-
`presses the angiogenesis that is promoted by
`the expression in vivo of constitutively active
`Akt74. It is feasible that the anti-angiogenic
`effect of rapamycin is the combined result of
`both mTORC1 and mTORC2 inhibition. By
`inhibiting the mTORC1-dependent transla-
`tion and activity of HIF1α (hypoxia-induced
`factor 1α)75–77, rapamycin decreases VEGF
`production by cancer cells. In endothelial
`cells the drug also inhibits VEGF-driven
`proliferation70 and promotes apoptosis78.
`Given the important role of Akt in these
`processes (reviewed in REF. 79), it is not
`unreasonable to hypothesize that the known
`capacity of rapamycin to inhibit Akt in
`endothelial cells36 might be important for
`its anti-angiogenic properties. In support of
`this idea, rapamycin suppresses angiogenesis
`in mice at high concentrations but not at
`the low concentrations that are sufficient
`to inhibit mTORC1 (REF. 70). Lastly, it is
`interesting to note that in Kaposi sarcoma
`Akt hyperactivation in endothelial cells is
`essential for tumorigenesis80.
`The HIF1α transcription factor stimulates
`VEGF production, and renal cancer cells that
`
`732 | SEPTEMBER 2006 | VOLUME 6
`
` www.nature.com/reviews/cancer
`
`© 2006 Nature Publishing Group
`
`Ex. 1105-0004
`
`

`
`lack the tumour suppressor VHL (von Hippel
`Lindau), which normally inhibits HIF1α, are
`particularly sensitive to rapamycin in culture
`and in tumour xenografts77. The expression
`in the cancer cells of a HIF1α mRNA that is
`engineered to make its translation resistant
`to rapamycin can largely eliminate the
`sensitivity of the cells to the antiproliferative
`effects of the drug77. In this case, rapamycin is
`clearly functioning through its action on the
`cancer rather than endothelial cells.
`
`Tumours with cyclin D1 overexpression.
`Tumours with cyclin D1 overexpression
`deserve their own special mention because
`this characteristic might underlie one of the
`more promising indications for rapamycin.
`In clinical trials rapamycin slowed the
`progression of nearly 40% of advanced
`MCLs49. A hallmark of MCL is the transloca-
`tion-induced overexpression of cyclin D1
`(reviewed in REF. 81). The mTORC1 pathway
`positively regulates cyclin D1 transcription,
`translation and stability in many types of
`cancer cell82–86. Despite the clear rationale
`for the use of rapamycin in MCL, in tis-
`sue culture experiments the drug had the
`unexpected effect of arresting MCL cells
`without decreasing their high levels of cyclin
`D1 (REF. 87). Recent work reveals that hyper-
`active PI3K–Akt signalling (in some cases
`caused by PTEN loss) occurs in about 50%
`of MCLs88 and it is tempting to speculate that
`rapamycin-mediated inhibition of Akt might
`contribute to the effectiveness of the drug in
`the treatment of some cases of MCL.
`
`Beyond rapamycin
`The likely approval in the near future of a
`rapamycin analogue for an anticancer indica-
`tion is almost certainly only the first foray
`into the oncology arena for mTOR inhibitors.
`Only recently has work begun on trying to
`identify compounds that perturb the mTOR
`complexes through mechanisms other than
`rapamycin and its analogues. Of particular
`interest will be molecules that directly inhibit
`the mTOR kinase domain, the assumption
`being that such molecules will inhibit both
`mTORC1 and mTORC2. Of course, this
`remains to be proven as structural changes
`that are induced by interacting proteins in
`the mTORC1 and mTORC2 kinase domains
`might not permit a single molecule to inhibit
`both while retaining specificity for mTOR.
`mTOR is essential for cell proliferation in
`mice89,90, so the expectation is that a direct
`mTOR kinase inhibitor will have more pro-
`nounced effects than rapamycin, which rarely
`completely arrests or kills cells on its own.
`On the other hand, there is no formal proof
`
`that the cell-essential functions of mTOR
`depend on its kinase activity as mTOR might
`have scaffolding functions that do not require
`an active kinase domain. Still, it is interesting
`to consider the potential anti-tumour activ-
`ity of a hypothetical molecule that directly
`inhibits the kinase function of mTOR so that,
`unlike rapamycin, it suppresses both mTOR-
`dependent pathways in all cancer cells. Such
`a molecule should inhibit both the mTORC1
`growth pathway that is regulated by S6K1 and
`4EBP1, and the mTORC2-dependent Akt
`pathway, and so should prevent the activation
`of Akt that is caused by mTORC1-only inhibi-
`tors. If activation of Akt by rapamycin and
`its analogues explains their ineffectiveness
`in certain tumours, such a molecule should
`overcome this. Currently, this idea cannot be
`tested because all molecules that inhibit the
`kinase domains of mTORC1 and mTORC2
`also inhibit PI3K p110α. Because acute
`inhibition of mTORC2 or PI3K p110α both
`suppress Akt27 it is difficult to distinguish the
`effects of an mTORC2 inhibitor from that
`of a PI3K p110α inhibitor when using Akt
`activity as a read-out. This raises the question
`of whether the greater effectiveness com-
`pared with rapamycin of molecules such as
`PI-103, which inhibit both PI3K p110α and
`mTOR67, depends on their capacity to inhibit
`mTORC2, PI3K p110α or both kinases.
`We have known since the early 1980s that
`rapamycin has anti-tumour properties91, but it
`has taken two decades for our understanding
`of mTOR and its connection with cancer-
`related pathways to progress to the point
`where we can begin to consider using mTOR
`inhibitors in a logical fashion. For those of us
`who have wrestled with the maddening com-
`plexity of the mTOR pathway, it is exciting
`that this time is finally here.
`
`David M. Sabatini is at the Whitehead Institute for
`Biomedical Research, MIT Department of Biology,
`9 Cambridge Center, Cambridge,
`Massachusetts, 02142-1479, USA.
`e-mail: sabatini@wi.mit.edu
`
`doi:10.1038/nrc1974
`Published online 17 August 2006
`
`1. Sarbassov, D. D., Ali, S. M. & Sabatini, D. M. Growing
`roles for the mTOR pathway. Curr. Opin. Cell Biol. 17,
`596–603 (2005).
`2. Tapon, N., Ito, N., Dickson, B. J., Treisman, J. E. &
`Hariharan, I. K. The Drosophila tuberous sclerosis
`complex gene homologs restrict cell growth and cell
`proliferation. Cell 105, 345–355 (2001).
`3. Gao, X. et al. Tsc tumour suppressor proteins
`antagonize amino-acid-TOR signalling. Nature Cell
`Biol. 4, 699–704. (2002).
`4. Saucedo, L. J. et al. Rheb promotes cell growth as a
`component of the insulin/TOR signalling network.
`Nature Cell Biol. 5, 566–571 (2003).
`5. Stocker, H. et al. Rheb is an essential regulator of S6K
`in controlling cell growth in Drosophila. Nature Cell
`Biol. 5, 559–565 (2003).
`6. Zhang, Y. et al. Rheb is a direct target of the tuberous
`sclerosis tumour suppressor proteins. Nature Cell Biol.
`5, 578–581 (2003).
`
`P E R S P E C T I V E S
`
`7. Tee, A. R., Manning, B. D., Roux, P. P., Cantley, L. C. &
`Blenis, J. Tuberous sclerosis complex gene products,
`Tuberin and Hamartin, control mTOR signaling by
`acting as a GTPase-activating protein complex toward
`Rheb. Curr. Biol. 13, 1259–1268 (2003).
`8. Garami, A. et al. Insulin activation of Rheb, a mediator
`of mTOR/S6K/4E-BP signaling, is inhibited by TSC1
`and 2. Mol. Cell 11, 1457–1466 (2003).
`9. Long, X., Lin, Y., Ortiz-Vega, S., Yonezawa, K. &
`Avruch, J. Rheb binds and regulates the mTOR kinase.
`Curr. Biol 15, 702–713 (2005).
`10. Inoki, K., Li, Y., Zhu, T., Wu, J. & Guan, K. L. TSC2 is
`phosphorylated and inhibited by Akt and suppresses
`mTOR signalling. Nature Cell Biol. 4, 648–657
`(2002).
`11. Potter, C. J., Pedraza, L. G. & Xu, T. Akt regulates
`growth by directly phosphorylating Tsc2. Nature Cell
`Biol. 4, 658–665. (2002).
`12. Manning, B. D., Tee, A. R., Logsdon, M. N., Blenis, J. &
`Cantley, L. C. Identification of the tuberous sclerosis
`complex-2 tumor suppressor gene product tuberin as
`a target of the phosphoinositide 3-kinase/akt pathway.
`Mol. Cell 10, 151–162 (2002).
`13. Tee, A. R., Anjum, R. & Blenis, J. Inactivation of the
`tuberous sclerosis complex-1 and -2 gene products
`occurs by phosphoinositide 3-kinase/Akt-dependent
`and -independent phosphorylation of tuberin. J. Biol.
`Chem. 278, 37288–37296 (2003).
`14. Johannessen, C. M. et al. The NF1 tumor suppressor
`critically