This review focuses on recent insights into the mechanisms and the biological functions of the proteasome. This large ATP-dependent proteolytic complex is the main site for protein degradation in mammalian cells and catalyses the rapid degradation of ubiquitinated proteins, and is the source of most antigenic peptides used by the immune system to screen for viruses and cancer. ATP is required to unfold globular proteins to open the gated channel into the 20S proteasome and to facilitate protein translation into it. Inhibitors of its proteolytic activity are widely used as research tools and have proven effective in cancer therapy.
It is a real pleasure and an honour to be involved in this birthday celebration of the Biochemical Journal, and to have the opportunity to tell you about some of our exciting new insights into the molecular mechanism of the proteasome and its critical roles in protein degradation and in immune surveillance, and how we have been able to use this knowledge to develop a novel cancer drug.
Since my student days, almost 40 years, I have been investigating the physiological significance, regulation and mechanisms of intracellular protein breakdown. When I started to study this area, the importance of proteolysis in cell regulation was not appreciated and the biochemical pathways were not known . It was taught that proteins did not turn over in bacteria and that the only site for protein breakdown in mammalian cells was the lysosome [1,2]. However, for those few of us interested in protein turnover, it was hard to understand how that membrane-enclosed bag of proteases could explain some of the key features of this process . For example, Robert Shimke and co-workers showed that enzymes in mammalian cells had widely different half-lives, and that these rates were regulated. Also, as a graduate student, I showed that overall rates of protein breakdown in muscle were carefully regulated by hormones and physical activity, and in the early 1970s we established that the key function of protein breakdown in bacterial and mammalian cells was to rapidly eliminate misfolded or mutant proteins, whose accumulation might be toxic. (This process is now of major importance in human disease and biotechnology.) This exquisite selectivity didn't fit with the lysosome playing a key role, especially since bacteria lack such organelles. An important clue to the existence of another degradative process was our finding in 1972  that this rapid degradation of misfolded proteins required ATP in prokaryotes, as in mammalian cells . An ATP requirement was very surprising, since proteolysis is thermodynamically favoured and no protease was known to require ATP [4,5]. So it had to indicate a novel proteolytic machinery. Much of what we learned and are still learning about this process emerged from efforts to understand this requirement.
A key development occurred exactly 30 years ago, when a postdoctoral fellow, Joseph Ettlinger, and I  demonstrated the existence in reticulocytes of a novel proteolytic system that, unlike the lysosome, was not membrane-enclosed, functioned at neutral pH and required ATP. Moreover, this system seemed responsible for the rapid hydrolysis of misfolded polypeptides, and probably also of short-lived enzymes in intact cells.
The ubiquitin–proteasome pathway
The first major development in explaining the role of ATP in this system was the very important discovery a few years later by Avram Hershko, Aaron Ciechanover and Ernie Rose of the role of ubiquitin conjugation in marking proteins for degradation [5,7], for which they were awarded the Nobel Prize in Chemistry in 2004. They showed that ubiquitination is an energy-requiring process, in which proteins are marked for degradation by covalent attachment of the small protein, ubiquitin, through isopeptide bonds to lysine residues on the protein [4,5]. This process requires the initial ATP-dependent activation of ubiquitin's terminal carboxy group by the ubiquitin-activating enzyme, E1. Then the activated ubiquitin is transferred to one of the 20 to 40 ubiquitin carrier proteins, or E2s, present in our cells. Some selectivity for substrates occurs in this step, but the exquisite selectivity of intracellular proteolysis comes from the ubiquitin ligases, or E3s, which recognize one or a set of protein substrates and catalyses the transfer of the activated ubiquitin from the E2 to the substrate, forming processively a chain of ubiquitins [5,7]. The attachment of a chain composed of four or more ubiquitins leads to rapid degradation of the protein. There are 500 to 1000 different ubiquitin ligases in our genomes, and despite recent progress, we probably know the function of only several dozen, which is unfortunate since they play critical roles in the regulation of cell cycle, gene transcription and metabolic pathways.
A very different explanation of this ATP requirement emerged also in the early 1980s from our studies of protein degradation in Escherichia coli, where we discovered a new type of enzyme – large multisubunit proteolytic complexes that cleave proteins and ATP in linked reactions . The first enzyme found was protease La (now called lon), and later Ti (now called ClpAP), HslUV, and the membrane-associated protease, FtsH, whose mechanisms and roles are presently major topics of study . These enzymes either recognize substrates directly or do so together with molecular chaperones. Bacteria lack ubiquitin, which first appeared in eukaryotes apparently to provide greater specificity for protein degradation. Thus initially there were two distinct explanations for the ATP requirement – one for marking proteins for degradation and one to drive an unusual proteolytic complex. Then, Keiji Tanaka and Lloyd Waxman in my laboratory  were able to show that ATP was also required in the reticulocyte system to degrade the ubiquitinated proteins and some non-ubiquitinated ones. Subsequent work in Martin Rechsteiner's and my laboratories  showed 20 years ago that this degradative process was catalysed by a very large ATP-dependent proteolytic complex with many of the properties of the bacterial enzymes [8,9]. We later named this structure the 26S proteasome.
This molecular machine (2.4 MDa) is composed of nearly 60 subunits and comprises approx. 2% of cell proteins [11,13]. Our present understanding of its function has emerged from the work of many laboratories, especially those of Martin Rechsteiner, George DeMartino, Keiji Tanaka and Daniel Finley, and the very important structural work of Wolfgang Baumeister [11,12]. Their electron microscopic tomography work [11,12] greatly clarified its different components and complexity. Protein hydrolysis takes place within its central cylindrical core particle, called the 20S proteasome (Figure 1). Its structure is composed of four superimposed rings and was solved by the highly informative X-ray diffraction studies of Michael Groll and Robert Huber. The 20S proteasome is a three-chambered structure, and in the central chamber, formed by its two central β-rings, substrate proteins are cleaved to small peptides by the combined actions of six active sites, two of which preferentially cut after hydrophobic residues, two after basic residues and two after acidic ones [11,13]. Because its subunits interact tightly, substrate proteins can enter and peptide products can emerge only through small gated openings in each of its outer α-rings. This architecture thus ensures that only proteins targeted for degradation and translocated into this hollow particle are degraded.
The 26S proteasome binds the ubiquitin chain, using ATP to unfold and translocate the substrate into the 20S core particle for degradation
At either or both ends of the 20S proteasome are the 19S regulatory particles (also called PA700), which bind the poly-ubiquitinated proteins, remove and disassemble the ubiquitin chain, unfold the substrate, and translocate it into the 20S core particle. Unfolding the substrate is a key step, since the narrow entry channel into the 20S core particle's outer ring will not allow transit of a globular protein [11,13,14]. All these functions of the 19S particle require ATP, and at the base of the 19S particle adjacent to the 20S core particle are six homologous ATPases, members of the AAA family of ATPases, which play multiple essential roles in proteasome function . Finally, small peptides emerge from the 20S that range from 2 to 25 residues in length  (Figure 1), nearly all of which are quickly hydrolysed to amino acids by cytosolic peptidases, but in higher eukaryotes some function in antigen presentation to the immune system (see below).
Roles of ATP in proteasome function
Because of the complexity of the 26S proteasome, and the requirement for ubiquitin conjugation to the substrates, it is not conducive for many types of mechanistic studies. Therefore we chose to investigate in depth the homologous, but simpler, proteasome from archaea [14,15]. Although these prokaryotes lack ubiquitin, they do contain 20S proteasomes that, like the mammalian 20S core particles, are composed of four superimposed rings, each containing seven subunits. However, the archaeal particles contain only one type of α-subunit and one type of β-subunit, which contain its active sites, all of which are identical [11–14]. Peter Zwickl, a postdoctoral fellow in my lab, was able to find in the archaeon Methanococcus jannaschii an ATPase complex that appears to be the evolutionary ancestor of the 19S mammalian complex. It is a member of the AAA family of hexameric ring ATPases, and it is highly homologous [14,15] to the six ATPases in the 19S regulatory particle complex. We named it PAN, for proteasome-activating nucleotidase [14,15].
Neither this proteasome on its own, nor PAN, nor the two together can efficiently degrade proteins. However, when ATP is added to the PAN–20S proteasome mixture, they rapidly degrade unfolded proteins , such as casein, and also globular proteins, such as the green fluorescent protein, GFP (provided it is linked to some short unfolded sequence; ). PAN thus facilitates the entry of these substrate proteins into the 20S particle, and together with the proteasome forms an ATP-dependent protease, with certain features similar to those that occur in bacteria (e.g. they have a distinct ATPase domain or subunits, and a proteolytic compartment). Like them, its rate of ATP hydrolysis increases when it binds a protein substrate. Nadia Benaroudj showed that the PAN–20S particle complex consumes approx. 300–400 ATP molecules each time it degrades a protein the size of casein or GFP, which is about one-third as much as the ribosome uses in synthesizing such a protein .
Four functions have been proposed for the proteasomal ATPases: they facilitate substrate binding; mediate opening of the gated channel in the 20S proteasome to allow protein entry; are involved in substrate unfolding; and promote protein translocation . We now know that PAN (and the homologous 19S ATPases) serve all these functions, and we are beginning to elucidate its ATP-dependent mechanisms in detail. Dr Benaroudj and myself were able to follow this unfolding process using GFP as a substrate, since it is green in the folded state . By itself, PAN with ATP is able to catalyse the loss of GFP fluorescence. This ‘unfoldase’ activity requires ATP hydrolysis, and represents a new kind of biochemical process. Another postdoctoral fellow in my group, Ami Navon, was able to show that unfolding occurs on the surface of the ATPase ring , but the mechanisms by which ATP-induced changes in its structure cause substrate denaturation are completely unclear. When the 20S proteasome was also present, PAN translocated the unfolded molecule into the 20S proteasome, where it was digested. Thus unfolding and translocation are normally linked processes.
Recently, in collaboration with Yifan Cheng and Tom Walz, David Smith (a postdoctoral researcher in my laboratory) succeeded in capturing images by cryoelectron microscopy of the PAN–20S proteasome complex, which associates in a transitory manner . PAN appears as a ‘top-hat, two-ring’ structure at either end of the 20S proteasome. Thus its outer ring appears to bind substrates and to translocate them through its central channel in the 20S proteasome's outer ring. Furthermore, we found that images of PAN could be exactly superimposed upon the densities (i.e. the two pseudo-rings) in the base of the 19S component of the eukaryotic 26S proteasome (e.g. as shown in images from Wolfgang Baumeister's laboratory). Thus the proteasome-regulatory ATPases appear to have been highly conserved during evolution . In addition to these ATPases resembling PAN, the 19S complex contains a set of extra subunits in its lid that are homologous with subunits in the signalosome complex, a eukaryotic particle involved in signal transduction . This lid structure apparently evolved when ATP-dependent proteolysis became linked to ubiquitination to enhance the specificity of the proteasome–ATPase complex.
Substrate entry into the proteasome
The entry of substrates into the 20S proteasome is normally prevented by a ‘gate’; that is formed by the eight N-terminal residues of the a-subunits, and deletion of these N-terminal residues allows rapid entry of unfolded polypeptides . The X-ray diffraction analysis of Michael Groll, Daniel Finley and co-workers has defined the structure of this gated entry channel in the yeast 20S proteasome . One clear role of the regulatory ATPases upon binding ATP is gate opening. In fact, one ATPase subunit, Rpt2, and one of the α-subunits, α3, appear to be particularly important in regulating this gating process . Recently, David Smith has obtained further insight into this ability of PAN and ATP to cause gate opening and substrate entry . The N-termini of the archaeal 20S proteasome's α-subunits also form a gate that excludes entry of proteins, but allows free entry of short peptides that are six residues or smaller [17,19]. However, deletion of these N-terminal residues allows peptides of nine residues and unfolded proteins to enter rapidly. Therefore we could use the hydrolysis of a nine-residue peptide as a measure of gate opening and its regulation by PAN. In the presence of ATP, PAN caused gate opening, but ADP prevented this process. ATPγS, but not ADP, also allowed the association of PAN with the 20S proteasome, and this non-hydrolysable nucleotide surprisingly worked even better than ATP, presumably because ATP hydrolysis to ADP closed the gate until a new ATP replaced the ADP.
The open-gate conformation induced by PAN–ATP also allowed entry of unfolded or denatured proteins. Thus, while binding and hydrolysis of ATP are necessary for degradation of GFP, if this protein was first denatured, only nucleotide binding (not hydrolysis) was sufficient to allow its entry and degradation by the proteasome . Therefore energy is used for protein unfolding on the ATPase, but the subsequent translocation of the unfolded protein can occur without ATP hydrolysis. Thus, if PAN is in its ATP-bound form, protein entry can occur by facilitated diffusion through the ATPase ring and the entry channel in the 20S proteasome. Presumably, in the 20S polypeptide, binding to the active sites and nucleophilic attack on the protein functions as a Brownian ‘ratchet’ that blocks retrograde diffusion of the polypeptide out of this proteasome.
To summarize, ATP serves multiple roles in proteasome function [14,19]: (1) ATP binding allows PAN–20S association (or 19S–20S association in eukaryotes) and (2) opens the gated channel for substrate entry into the 20S proteasome; (3) ATP enhances the proteasome's affinity for proteins, and their binding; (4) protein substrates stimulate ATP hydrolysis, which allows protein unfolding on the ATPase surface; and (5) ATP binding allows diffusion of the unfolded polypeptide into the degradative chamber. We do not know how peptides exit the proteasome, although this perhaps occurs by triggering gate opening in the 26S complex. This multistep mechanism (in the 26S proteasome) must involve further ATP-dependent steps, such as disassembly of the ubiquitin chain.
Exciting progress in understanding the ATP-dependent gate-opening mechanism has been achieved recently by David Smith, myself and others (unpublished work), who showed that the three C-terminal residues of PAN contain a highly conserved HbYX motif, which is essential for gate opening. This sequence is conserved in most of the ATPases in the mammalian 26S ATPase, and we showed that it docks into pockets located between the α-subunits of the 20S proteasome. Furthermore, short 7–10-residue peptides that contain this C-terminal HbYX motif, by themselves, can dock into these same pockets and cause gate opening. Thus the ATPases’ C-termini, in their ATP-bound form, function like a ‘key in a lock’ to cause gate opening. Our collaborators, Yifan Cheng and Julius Rabl, have succeeded in using cryoelectron microscopy to visualize these short peptides bound to the intersubunit pockets by comparing densities with or without the gate-opening peptides. Their remarkable images have also provided new insights into the open and closed forms of the gate, and how the N-terminal residues prevent substrate entry. They also demonstrate that the ATPases' C-termini trigger conformational changes in the α-ring that stabilize the gate in its open conformation to allow substrate entry.
Proteasome inhibitors and cancer therapy
These basic advances in understanding the mechanisms and biological importance of the proteasome have already had practical applications in cancer therapy and in expanding knowledge in many areas of biology. About 15 years ago, we undertook the development of inhibitors of proteasome function that could reduce protein degradation in intact cells. To establish collaboration with chemists, biologists and physicians, we formed a small biotechnology company. Our initial goal was to develop drugs that might reduce muscle wasting, as occurs in many major diseases, ranging from spinal injuries, to cancer cachexia, to renal failure. We had shown that loss of muscle mass in these various conditions was due primarily to the excessive protein breakdown and activation of the ubiquitin–proteasome pathway . This major medical need enabled us to obtain venture capital funding. (I also recognized that such small-molecule inhibitors could greatly advance our understanding of the physiological roles of the ubiquitin–proteasome pathway, but this goal would not motivate investors.)
Effective proteasome inhibitors were identified rapidly not through screens of large chemical libraries, but based on simple biochemical insights . When this work was initiated in the early 1990s, most of what I described alone concerning proteasome structure and function was not known, but we did know through the early work of Wilk and Orlowski that the eukaryotic 20S proteasome contained three types of active sites [11,13]. One was chymotrypsin-like, cleaving preferentially after large hydrophobic residues; one was trypsin-like, cleaving mainly after basic residues, and the third, which we now refer to as ‘caspase-like’, cleaves preferentially after acidic residues . Genetic studies suggested that the chymotrypsin-like site was the most important in protein degradation, and we focused on building aldehydes of its hydrophobic peptide substrates, and the recognition that hydrophobic peptides might best enter cells. The first types of inhibitor developed were peptide aldehydes, such as MG132. This widely used inhibitor is an aldehyde derivative of three leucine residues with a blocked N-terminus. Initially, the proteasome was believed to be an anomalous serine or cysteine protease, and peptide aldehydes are transition-state inhibitors of both, but the later X-ray diffraction analysis by Huber, Baumeister and colleagues showed that the proteasome was actually a novel type of protease, whose active site nucleophile was the hydroxy group of an N-terminal threonine. Julian Adams, who led the chemistry effort at our company, then introduced boronate as ‘warheads’, which were known to inhibit serine proteases strongly. This modification dramatically enhanced proteasome inhibition, and subsequent medicinal chemistry efforts finally yielded the very potent derivatized dipeptide boronate, termed PS341 or Bortezomib, but commercialized under the name Velcade® (Figure 2).
Structure of the boronate derivative NF-κB inhibitor PS-341 (Bortezomib; Velcade®), a successful drug for multiple myeloma that is now in clinical trials for other cancers
Meanwhile, a key discovery made through a collaboration with Tom Maniatis, Vito Palombella and co-workers  led to a major alteration in the company's focus. Their studies demonstrated an essential role of the 26S proteasome in the activation of NFκB (nuclear factor κB), the transcription factor critical in the immune and inflammatory responses and important in cancers. During activation of NFκB, the ubiquitin–proteasome pathway catalyses the rapid degradation of its inhibitory protein, IκB (inhibitory κB), and processing of the p105 precursor to NFκB's p50 subunit. Consequently, proteasome inhibitors had real potential as anti-inflammatory agents (for which they were initially developed) and as anticancer agents.
PS341, or Bortezomib, had very promising biochemical and pharmacological properties. At low nanomolar concentrations, it inhibited NFκB activation, stabilized critical cell cycle regulators (p53 and p27), and slowed cell cycle progression, especially of transformed cells. It was given to the NCI (National Cancer Institute) for testing against many cancer models, where it was found to have strong anticancer effects distinct from those of known agents, and with the support of NCI, PS341 entered human trials to treat diverse cancers. In Phase I trials (which normally only evaluate toxicity), two of the patients had multiple myeloma and they showed clear responses. Therefore the Phase II trials focused on myeloma, and approx. 35% of the very ill patients, whose cancers no longer responded to any other drugs, showed significant responses and a few had near complete remissions. Consequently, the FDA approved this treatment in 2003, even without Phase III trials. It is now approved as a second line therapy, and has been taken by approx. 50000 patients worldwide. It also shows clear efficacy against related haematological malignancies, Waldenstrom's macroglobulinaemia and mantle cell lymphoma. Although, in animals, Velcade® showed potent synergy against several solid tumours, similar efficacy has not been demonstrated thus far in human trials. In preclinical studies, proteasome inhibitors do show potential promise in other therapeutic areas, including in inflammatory diseases and stroke, as well as muscle atrophy and tuberculosis (since mycobacteria have proteasomes).
We believe we understand the main reasons for the special sensitivity of multiple myeloma cells to proteasome inhibitors, which has been clearly established in cultured myeloma cells by Ken Anderson's group. First, in these cells, NFκB plays a particularly important role. In most cells, and especially in cancers, it is a major inhibitor of apoptosis. In myeloma cells, NFκB transcribes genes for IL-6 (interleukin-6), a critical growth factor for myeloma cells, for adhesion factors that are necessary for myeloma growth to adhere to the bone marrow, and for production of angiogenic factors, e.g. VEGF (vascular endothelial growth factor). Secondly, myeloma is a cancer of the plasma cell, and these cells actively produce large amounts of abnormal immunoglobins. Myeloma cells must be very active in degrading such misfolded proteins, a key role of the proteasome. Abnormal proteins in the ER (endoplasmic reticulum) also undergo retrograde translocation back into the cytosol, where they are degraded by the proteasome by a process often termed ERAD (ER-associated degradation) . This quality control system is very active in plasma cells , and proteasome inhibitors block this capacity to eliminate misfolded immunoglobins. The accumulation of such abnormal proteins has been shown to activate stress responses, the ‘unfolded protein response’ and JNK (c-Jun N-terminal kinase), and to cause apoptosis. These mechanisms seem to account for the therapeutic benefits of Velcade®, but in themselves do not explain why proteasome inhibition does not have a dramatic toxicity on other cell types.
An important question raised by these findings is why the proteasome inhibitors are not more toxic (as most ‘experts’ had predicted they would be). The recent studies by Alexei Kisselev and Alice Callard in my laboratory indicate the likely explanation . When only the chymotrypsin sites in the 20S particle are inhibited, as occurs with therapeutic doses of Velcade®, the rates of protein degradation decrease to only a limited extent (by, at most, <40%), since the remaining trypsin-like and caspase-like sites still function. To inhibit proteolysis markedly, at least two types of proteasomal active sites have to be blocked, which occurs at higher concentrations used by investigators in vitro . Also, in patients, this inhibitor is given intermittently, and thus there is only a limited inhibition of proteolysis for only a limited period, which is sufficient to cause apoptosis of the myeloma cells but is of little consequence to normal cells.
Proteasomes and immune surveillance
Proteasome inhibitors have proven to be very valuable tools in elucidating the mechanisms of a wide variety of physiological and pathological processes, ranging from inflammation and cell cycle to neurodegeneration [13,26]. One area where these agents have had an enormous impact is in our understanding of antigen presentation to the immune system . With the evolution of the immune system in higher vertebrates, both the proteasome and the lysosome have assumed a new role in helping the immune system screen for the foreign materials in the extracellular and intracellular spaces . Extracellular proteins taken up by endocytosis and degraded in lysosomes are the source of the antigenic peptides that are presented on MHC class II molecules and elicit antibody production. During the degradation of intracellular proteins by the proteasome, the great majority of the peptides generated are hydrolysed by cytosolic peptidases to amino acids within seconds of release. However, a small fraction escape this fate and are transported by the TAP transporter into the ER, where these peptides bind to MHC class I molecules and are delivered to the cell surface for presentation to circulating lymphocytes. If a non-native peptide, as may arise due to the presence of viruses, intracellular pathogens or cancer, is presented, the cytotoxic cells destroy the cell.
Using proteasome inhibitors, my long-term collaborator, Kenneth Rock, myself and colleagues were first able to demonstrate this key role of the proteasome in antigen presentation, and to clarify subsequent steps in this process . To bind to most MHC class I molecules, peptides have to be eight to nine residues long and have to have hydrophobic or basic C-termini . Approx. 70% of the peptides that emerge from the proteasome, however, are smaller than this length. Although proteasomes can generate the mature epitopes, most often they appear to produce longer peptides with N-terminal extensions. These precursors are trimmed further, most often within the ER, where there is a novel aminopeptidase, ERAP1 (ER aminopeptidase 1), recently identified by Nilabh Shastri's group and Tomo Saric in my laboratory. It has the remarkable ability to trim away the additional residues, and then stops at eight or nine residues, the precise length for binding to MHC molecules . In immune cells and most cells during inflammation, interferon-γ stimulates antigen presentation by inducing the TAP transporter, ERAP1, MHC molecules, the PA28 proteasome activator  and a specialized form of the proteasome now termed the immunoproteasome . It differs from normal 26S or 20S proteasomes in containing three alternative active sites with slightly different specificities from the constitutive proteasome. As a result, although these structures digest proteins at rates similar to those for constitutive particles, the 20S core particles generate a higher fraction of peptides with the appropriate C-termini and length to serve in antigen presentation.
I hope I have convinced you that the proteasome is an elegant molecular machine that has evolved to digest certain proteins efficiently while not harming most cell constituents. The development of this area also illustrates the unpredictable benefits of basic research – how efforts to understand a biochemical curiosity, the ATP requirement for proteolysis, can uncover novel biochemical mechanisms and a broad range of new biology, and even lead to important new human therapies.
Literature, Legacy, Life … Biochemistry for the 21st Century: Biochemical Journal Centenary Symposium held at the SECC, Glasgow, Scotland, U.K., 24 July 2006. Organized by G. Banting (University of Bristol, U.K.), P. Parker (Cancer Research UK, London, U.K.) and G. Salvesen (Burnham Institute, La Jolla, CA, U.S.A.). Edited by Dr Clare Sansom (Teaching Fellow, Birkbeck College, London, U.K.). Sponsored by the Biochemical Journal, Portland Press Ltd, GE Healthcare, BBSRC (Biotechnology and Biological Sciences Research Council), Cold Spring Harbor Laboratory Press, Alexis® Biochemicals & Axxora® Platform, DiscoveRx, Stratagene, Monsanto and EMBO (European Molecular Biology Organization).
These findings were made possible by the hard work of many members of my laboratory and by research grants from NI General Medical Sciences and the Ellison Foundation. The author is grateful to Mary Dethavong for expert assistance in the preparation of this manuscript.