Reconciling data from transgenic mice that overexpress IGF-I specifically in skeletal muscle

Transcription

1 Growth Hormone & IGF Research 15 (2005) 4 18 Review Reconciling data from transgenic mice that overexpress IGF-I specifically in skeletal muscle Thea Shavlakadze a, *, Nadine Winn b, Nadia Rosenthal a,b, Miranda D. Grounds a a School of Anatomy and Human Biology, the University of Western Australia, Crawley, Perth, Western Australia 6009, Australia b European Molecular Biology Laboratories, Mouse Biology Programme, via Ramerini, Monterotondo (Rome), Italy Received 26 May 2004; revised 1 November 2004; accepted 1 November Abstract Transgenic mice that overexpress insulin-like growth factor-1 (IGF-I) specifically in skeletal muscle have generated much information about the role of this factor for muscle growth and remodelling and provide insight for therapeutic applications of IGF-I for different pathological states and ageing. However, difficulties arise when attempting to critically compare the significance of data obtained in vivo by using different genetically engineered mouse lines and various experimental models. Complications arise due to complexity of the IGF-I system, since multiple transcripts of the IGF-I gene encode different isoforms generated by alternate promoter usage, differential splicing and post-translational modification, and how IGF-I gene expression relates to its diverse autocrine, paracrine and endocrine modes of action in vivo has still to be elucidated. In addition, there are problems related to specification of the exact IGF-I isoform used, expression patterns of the promoters, and availability of the transgene product under different experimental conditions. This review discusses the factors that must be considered when reconciling data from cumulative studies on IGF-I in striated muscle growth and differentiation using genetically modified mice. Critical evaluation of the literature focuses specifically on: (1) the importance of detailed information about the IGF-I isoforms and their mode of action (local, systemic or both); (2) expression pattern and strength of the promoters used to drive transgenic IGF-I in skeletal muscle cells (mono and multi-nucleated); (3) local compared with systemic action of the transgene product and possible indirect effects of transgenic IGF-I due to upregulation of other genes within skeletal muscle; (4) re-interpretation of these results in light of the most recent approaches to the dissection of IGF-I function. Full understanding of these complex in vivo issues is essential, not only for skeletal muscle but for many other tissues, in order to effectively extend observations derived from transgenic studies into potential clinical situations. Ó 2004 Elsevier Ltd. All rights reserved. Keywords: Insulin-like growth factor-1; Transgenic mice; Skeletal muscle 1. Complexity of IGF-I isoforms 1.1. Factors defining complexity of IGF-I isoforms The insulin-like growth factor-1 (IGF-I) gene gives rise to several isoforms of unprocessed (precursor) * Corresponding author. Tel.: ; fax: address: sthea@anhb.uwa.edu.au (T. Shavlakadze). IGF-I, which differ by the length of the amino-terminal leader (signal) peptide and structure of the so-called extension peptide (E-peptide) on the carboxy-terminal end (discussed in details below). These unprocessed polypeptides undergo post-translational protease cleavage to remove the leader sequence and the E-peptide and to yield a 70-amino acid long (mol wt 7649 Da) single chain mature IGF-I polypeptide with three intrachain disulfide bridges. IGF-I shares 62% homology with proinsulin and also contains B amino-terminal /$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi: /j.ghir

2 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) domain, A domain and C domain. However, unlike proinsulin, IGF-I polypeptides also contain a D carboxyterminal domain [1,2]. Mature IGF-I peptide is sequenced and described in many mammalian species, including marsupials (kangaroo), poultry, frogs and fish. The primary structure of IGF-I is highly conserved in placental mammalian species: canine [3], bovine [4], ovine [5], porcine [6] and human [1] IGF-Is are identical, whereas rat [7] and mouse [8] IGF-Is differ from human by 3 and 4 amino acids, respectively, and there is only one amino acid difference between these rodent IGF- Is. Amino acid sequence of kangaroo (marsupial) IGF-I differs from human sequence by 6 amino acids [9]. Apart from the full-length IGF-I, two other protein products are thought to be produced by post-translational cleavage of IGF-I precursor protein and these have been identified in the human brain (reviewed in [10]). A truncated IGF-I isoform (-3N:IGF-I) lacks the first three amino acids from the NH 2 -terminal end of the B domain, has low-binding affinity to IGF-binding proteins (IGFBPs) and displays enhanced biological (neurotrophic) effects, which are mediated through the IGF-I receptor [11,12]. Another product thought to result from a post-translational modification of IGF-I precursor protein is a tripeptide glycyl-prolyl-glutamate (GPE), which corresponds to amino-terminal end of mature IGF-I [10]. In the central nervous system, GPE modulates neurotransmitter release via N-methyl-Daspartate type glutamate receptor. The gene encoding IGF-I has been cloned and characterized in several mammalian species: rat [7], mouse [8], human [13], sheep [14], pig [15] and cattle [16]. The structure of the IGF-I gene is highly conserved in mammals and within this review we will mainly focus on the rodent IGF-I gene and its expression pattern with some reference to the human gene. The rodent (rat) IGF-I gene consists of six exons separated by five introns and spans >70 kb of genomic DNA [7,17] (Fig. 1(a)). Exons 1 and 2 (termed as leader exons) contain multiple transcription start sites and encode distinct, mutually exclusive 5 0 -untranslated regions (5 0 -UTRs) of IGF-I precursor mrnas as well as different parts of leader (signal) sequences of IGF-I precursor polypeptides [7,18 20]. Exon 3 encodes 52 amino acids: the initial 27 amino acids precede a start of a B domain of mature IGF-I and comprise the terminal part of the signal peptide; the rest of the 25 amino acids comprise the B domain of mature IGF-I. The remainder of the coding information for the mature IGF-I peptide (rest of domain B and domains C, A and D) resides in exon 4. Rat exon 4 also contains 16 codons corresponding to the amino-terminal portion of the E-peptide. Exons 5 and 6 each encode distinct E-peptides, termination codons and 3 0 -UTRs [7]. The IGF-I gene gives rise to a heterogeneous pool of mrna transcripts (Fig. 1(b)). Such heterogeneity of the mrnas results from several events (or combination of these events): use of alternative transcription start sites located in leader exons (exons 1 and 2) [18 20]; alternative post-transcriptional exon splicing [8,21,7,18 20]; and use of different polyadenylation sites [22,23]. These multiple IGF-I mrnas transcripts encode different isoforms of IGF-I precursor peptide (Fig. 1(c)), which undergo post-translational cleavage to release the biologically active mature (70 amino acid long) IGF-I. Using RNAse protection and primer extension procedures at least four main transcription start sites have been identified in exon 1 [18 20]: a minor site is located at 382 bp (start site 1), two major sites located at 343 and 245 bp (start site 2 and start site 3, respectively) and a minor start site locate at 40 bp position (start site 4) upstream of the 3 0 -end of exon 1 [18 20]. Simmons et al. [19] have identified two additional minor transcription start sites located at the 361 and 353 bp positions upstream of the 3 0 -end of exon 1, however their use is thought to be insignificant. In exon 2, the most upstream and the most rarely used start site is located at 770 bp (start site 1) upstream to its 3 0 -end. Two downstream start sites are located at 50 bp (start site 2) and 70 bp (start site 3) upstream to the 3 0 -end of the exon 2. In rat liver, 75% of total IGF-I synthesis is accounted for by exon 1 transcription [24], from which the vast majority of the transcripts (70% of total liver IGF-I mrna synthesis) initiate from the start sites 2 and 3 and the minority of the transcripts (5% of total liver IGF-I mrna synthesis) initiate at the start site 4 [24]. Abundance of transcripts initiated at the most upstream promoter (start site 1) of exon 1 is extremely low [24], however they can be easily amplified by RT-PCR (Winn unpublished observations). Almost 90% of exon 2 transcripts initiate from the second start site in exon 2 [24]. During postnatal development and in adult rats expression of exon 2 transcripts is more growth hormone-dependent than expression of exon 1 transcripts [25,26]. Exon 2 transcripts appear later in postnatal development than exon 1 transcripts, increasing especially at the onset of GH-dependent linear growth, and their appearance coincides with that of circulating IGF-I [24]. Expression of exons 1 and 2 transcripts has also been studied in extrahepatic tissues of adult rats [27] and, similar to liver, in kidney and brain most of the exon 1 transcripts initiate from start sites 2 and 3. However in testes, lung, muscle and stomach, transcription initiates primarily from start site 3. Transcription in heart occurs only from start site 3. Presence of the transcripts initiated at start site 4 are also detected in all of these tissues, however at low levels [27]. Exon 2 transcripts are detected in testes, lung, kidney and stomach, however at much lower absolute levels than in liver. Exon 2 transcripts are undetectable in heart, brain and muscle [27].

3 6 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) 4 18 Fig. 1. Structure of the rodent IGF-I gene (a) (modified from [84]). Coloured boxes represent the open frame, which codes for the precursor IGF-I peptide. Both exons 1 and 2 contain multiple transcription start sites (horizontal arrows). Translation initiation codons (AUG) are located at exons 1, 2 and 3. These exons code for the signal peptide of precursor IGF-I. Seventy amino acids of mature IGF-I are coded for by exons 3 and 4 (indicated by red boxes). Exon 4 also contains code corresponding to the amino-terminal portion of the extension peptide (E-peptide). Exons 5 and 6 each encode distinct portions of the E-peptides, termination codons and 3 0 -untranslated regions. Polyadenylation sites located on exon 6 are shown by vertical arrows. The IGF-I gene gives rise to multiple mrna transcripts (b) and the diagram illustrates differential splicing of 5 0 and 3 0 ends of precursor mrna. Consequently, isoforms of unprocessed IGF-I translated from the various IGF-I mrna slice variants differ by the length of the signal peptide (48aa, 32aa and 22aa) and by the C-terminal E-peptide (Ea = 35aa, Eb = 41aa and the third 63aa E peptide, which has recently been identified in rodents and corresponds to human Eb peptide) (c). Heterogenity of IGF-I gene transcripts is further determined by differential splicing of exons at 5 0 and 3 0 ends of precursor mrna (Fig. 1(b)). Transcripts initiated at exon 1 are spliced to produce mrna, which contains exon 1 but lacks exon 2, whereas transcripts initiated at exon 2 give rise to mrna which contains exon 2 but lacks exon 1 [18 20]. In addition, transcripts initiated at start sites 1 and 2 of exon 1 can undergo a post-transcriptional splicing (exon 1 spliced variant), which results in removal of a 186 base long sequence (including a major transcription site 3) and introduces further complexity in heterogeneity of IGF-I mrnas [24]. In rodents about 20% of liver IGF-I transcripts are subject to this modification [24]. Transcripts initiated at exons 1 and 2 are also referred to as Classes 1 and 2 transcripts, respectively, and hereafter we will also use these terms. Data obtained by using in vitro translation techniques confirm that three distinct signal peptides can be generated from Class 1 (exon 1) and Class 2 (exon 2) mrnas transcribed from the IGF-I gene [19,20] (Fig. 1(c)). Major transcription start sites 2 and 3 of exon 1 are located upstream to Met-48 translation initiation codon (also located in exon 1) and give rise to

4 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) IGF-I precursor polypeptides with 48 amino acid long signal sequence [19,20]. Transcription start site 4 is located downstream to Met-48 codon and mrnas initiated from this site encode a 22 amino acid long signal peptide derived from exon 3 (which encodes a translation initiation codon corresponding to Met-22) [20]. Translation of the mrnas initiated at transcription start site 1 has not been studied, since the amount of these transcripts in rat liver is negligible [20]. Human exon 1 contains an additional AUG codon at a 25 position relative to the IGF-I coding region that is translated into a 25 amino acid signal peptide [28]. In exon 2, all three transcription start sites are located upstream of the Met-32 translation initiation codon (located in exon 2), therefore exon 2 mrnas are translated into a precursor IGF-I peptide initiated at a Met-32 and Met-22 codons [19,20]. In mrnas that contain two translation initiation sites, the preference is given to the upstream one [20]. In summary, in rodents precursor IGF-I peptides translated from Classes 1 and 2 mrnas can have 48, 32 and 22 residue long amino-terminal sequences, which function as signal peptides and are cleaved during posttranslational processing [20]. The length and or the structure of 5 0 -untranslated sequences affects the efficiency of translation initiation [20] and provides for different post-translational modification (e.g., glycosylation) of IGF-I precursor peptide [19]. In particular, translational efficiency of IGF-I mrna in vitro and in vivo is inversely proportional to the length of the untranslated region [20]. It appears that in vitro the IGF-I precursor with the 48-residue signal peptide does not get glycosylated, and a precursor with the 32-residue signal peptide is glycosylated to a lesser extent than a precursor with the 22-residue signal peptide [19]. The alternate splicing and differential promoter usage at the 5 0 -end of IGF-I transcripts that leads to the generation of different signal peptides has also been reported to affect the precise N-terminal cleavage site of the signal peptide [29]. It has been hypothesized that alternate signal peptides could lead to cleavage at a position three amino acids downstream of the usual cleavage site to produce a N-terminal truncation of the B-domain of mature or pro-igf-i [30,10]. Alternative splicing of exons at the 3 0 -end of mrna precursors introduces further complexity in the variety of IGF-I transcripts and IGF-I isoforms translated from these transcripts. In particular, exon 4 6 and exon splice variants have been identified in rat and mouse liver [8,21,7]. When exon 4 is spliced to exon 6, 19 amino acids are added to the common 16 amino acids encoded by exon 4, thus generating a 35 amino acid long E-peptide, termed Ea-peptide (Fig. 1(c)). Inclusion of a 52 base fragment of exon 5 results in addition of 17 amino acids and a frame shift in the reading frame, which introduces an earlier in-frame stop codon in the exon 6 sequence. Hence, the extension peptide encoded by exon splice variant is termed Eb-peptide, which contains 41 amino acids (16 amino acids from exon 4, 17 amino acids from exon 5 and 8 amino acids from exon 6) (Fig. 1(c)). Since in murine species exon 4 6 and exon mrna splice variants differ only by the absence or presence of a 52 base insert, exon 6 codes for 3 0 -UTRs of both of these splice variants. Until recently, only exon 4 6 and exon splice variants have been described in rodents (discussed later). In humans, three mrna variants with alternatively spliced 3 0 -end have been identified. Similarly to rat, splicing of human exon 4 to exon 6 yields an mrna sequence which encodes the 35 amino acid Ea-peptide [28], which shares 91% homology with the mouse Ea-peptide [8]. The second mrna splice variant identified in human liver is an exon 4 5 splice variant, which encodes an extension peptide termed Eb-peptide [13,31]. Human Eb-peptide contains additional 61 amino acids encoded by exon 5 [31], giving rise to an E-peptide with the total length of 77 amino acids. The third mrna splice variant identified in human liver contains exon 4, 49 bp of exon 5 and then exon 6 (exon 4 5 6) and yields an extension peptide with a total predicted length of 41 amino acids, termed as Ec-peptide [32]. This human Ec-peptide shares 73% homology with rat Eb-peptide and is considered to be its counterpart [32]. Possible existence of the exon 4 5 splice variant, which codes for Eb-peptide in humans, was also predicted in rats [7], but since it was not detected in liver using northern blot technique, it was assumed that this splice variant does not exist in rodents. However, Winn et al. recently succeeded in cloning the exon 4 5 splice variant from mouse liver mrna, which is present only in exon 2 mrnas, suggesting that it potentially codes for a circulating isoform (Winn et al. unpublished observations). Mouse exon 5 encodes a TAG stop codon 141 bp downstream to its start and as predicted from the sequence, contains four AATATA polyadenylation signals, 458, 569, 2888 and 5178 bp downstream to the exon 5 start, respectively, while no AATAAA polyadenylation signals are predicted. Further characterization is needed to identify the functional polya addition sites. The mouse exon 4 5 splice variant encodes a 63 amino acid sequence of E-peptide (16 amino acids encoded by exon 4 and 47 amino acids encoded by exon 5), while the human counterpart (Eb-peptide in humans) encodes a 77 amino acid E-peptide (16 amino acids from exon 4 and 61 amino acids from exon 5) and there is only 57% homology between these two sequences. In mouse liver, this IGF-I splice variant is expressed to a low extent compared to other IGF-I splice variants and it is not detectable in any other mouse tissues by RT-PCR. Most importantly the mouse sequence does not contain a nuclear and nucleolar localization signal (Winn unpublished observations), which has been described

5 8 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) 4 18 for the human Eb-peptide and has been shown to direct this human IGF-I isoform to the nucleolus [33]. It is thought that in humans the exon splice variant (termed Ec-peptide) occurs by use of a cryptic IGF 633 donor splice site located 49 bp downstream from the 5 0 end of exon 5, which deviates from the vertebrate 5 0 -donor splice site consensus. Failure to use this cryptic IGF 633 donor splice site results in exon 4 5 splice variant (termed Eb-peptide) [34]. When comparing the 3 0 exon5:5 0 intron5 boundary to the vertebrate 5 0 -donor splice site consensus AG:AGTAAGT, the rat sequence matches by 5 out of 6 bases, the mouse sequence matches by 4 out of 6 bases, and the human sequence matches by only 3 out of 6 bases (Fig. 2). These polymorphisms might alter the strength of the donor splice site and influence the splicing machinery. Because the rat 5 0 -donor splice site shows the highest match to the vertebrate 5 0 -donor splice site consensus, it is more likely to be efficiently used by the splicing machinery. In addition, the rat intron sequence following the 52- bp exon contains 4 purine-rich repeats (GGAAG) within 300 bp downstream of the 5 0 -donor splice site, which have been shown to enhance splicing in the bovine GH gene [35]. Finally, the sequence downstream of the donor splice site contains only one AATATA polyadenylation signal, which might not be strong enough to compete with the 5 0 -donor splice site. In addition to the differences in amino acid composition, an important difference between Ea- and Ebpeptides in rodents (mouse and rat) is the presence of two potential N-linked glycosylation sites in the Ea variant, which are absent from the Eb variant [21]. An in vitro study showed that both glycosylation sites are N-glycosylated and that both sites are probably used [36], suggesting that this post-translational modification is involved in a biological action of the IGF-I isoform. Human Ea-peptide contains only the first glycosylation site [2]. Existence of at least 4 polyadenylation sites in the 3 0 -UTRs of the rodent and human IGF-I mrnas, encoded by exon 6 [22,7] adds another level of posttranscriptional regulation. The use of several different polyadenylation sites creates size heterogeneity of rodent IGF-I mrnas and gives rise to three main IGF-I transcript sizes: , , and 7.5 kb [23]. The largest 7.5 kb IGF-I mrnas have a long ( kb) 3 0 -UTR, whereas the smaller and kb IGF-I mrnas have a shorter ( and kb) 3 0 -UTR. It has been shown in vitro and in vivo that the half-life of IGF-I mrnas is inversely proportional to the length of their UTRs with the longest transcript of 7.5 kb having the shortest half-life Expression of IGF-I isoforms in skeletal muscle The predominant IGF-I mrna variant expressed in skeletal muscle is initiated at exon 1 (Class 1) and represents an exon 4 6 spliced variant, which encodes an IGF-I isoform containing an Ea-peptide [37]. Confusion exists in the literature regarding nomenclature of different isoforms of IGF-I. Since mrna encoding the Class 1 IGF-I Ea isoform is expressed locally in muscle (and other tissues) and also represents a predominate variant expressed in liver, some investigators refer to this particular isoform as a local muscle specific isoform (migf-i) [37,38] and others as a muscle liver type isoform [39 41]. In muscles subjected to damage or exercise the exon splice variant is upregulated and the isoform encoded by this mrna has been called mechano growth factor (MGF) [39 41]. Although the signal peptide included in MGF has not been specified, MGF presumably corresponds to IGF-IEb in the rat and IGF-IEc in humans. Thus, in rats MGF contains a 52 bp insert and in humans a 49 bp insert, contributed by the inclusion of exon 5. Expression of MGF has also been detected by RT-PCR in resting muscles of rats and humans, however at a much lower levels than migf-i [42,41]. A study on C2C12 myogenic cells suggests different roles for migf-i and MGF: MGF prolongs proliferation of myogenic cells, whereas migf-i promotes their fusion [43]. While both isoforms increase myoblast proliferation, MGF stimulates higher proliferation within the first 48h [43]. It is important to point out that these tissue culture experiments employed a predicted synthetic MGF peptide [43] and as yet, there is no evidence for the presence of MGF protein in vivo. However, involvement of MGF in myoblast proliferation is also supported by in vivo studies on humans [41] and rats [44]. Robust upregulation of MGF mrna (in the range from 2% to 864%) was observed within 2.5 h after high resistance exercise in young men, with no significant changes in migf-i mrna levels [41]. Rapid upregulation of MGF was Fig. 2. Comparison of the 3 0 exon5:5 0 intron5 boundary of rat, mouse and human IGF-I gene to the vertebrate 5 0 -donor splice site consensus. Shaded nucleotides represent deviations from the vertebrate 5 0 -donor splice site consensus.

6 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) seen in rat muscles after stretch or bupivacaine injury, where peak of MGF mrna level preceded that of migf-i. However, it should be pointed out that in these experiments the absolute mrna levels of migf-i were always much higher (10- to fold) then those of MGF [41,44]. Overall, the available techniques do not allow for identification and differential quantitation of different IGF-I isoforms at the protein level, and existing studies provide information on mrna rather than protein levels. Definitive analysis of the complexity of IGF-I expression in vivo awaits the development of appropriate antibodies to distinguish isoform production and distribution. 2. Transgenic mice with skeletal muscle specific IGF-I overexpression 2.1. Different transgene constructs Three mouse strains have been developed that overexpress IGF-I (Fig. 3) specifically in skeletal muscle using different transgenic constructs. They exhibit quite distinct phenotypes, which should be considered when comparing experimental data reported for these different transgenic strains (Table 1). Also the age and gender of the animals should be taken into the account, as discussed below. The SK733 IGF-I 3 0 SK transgenic mouse strain (Fig. 3(a)) was developed on the FVB mouse background strain, and carries the human IGF-I gene expressed in skeletal muscle under the avian skeletal a-actin promoter [45]. Human IGF-I cdna that has been used to generate this transgenic construct presumably corresponds to Class 1 IGF-I Ea isoform (migf-i) [28]. To increase the stability of the transgenic transcripts, 1.8 kb of the a-skeletal actin 3 0 -UTR was substituted for the native 3 0 -UTR of the human IGF-I gene. Even though this substitution led to less than a 2-fold increase in higf-i mrna levels, it might also influence the function of the transgene in a yet unidentified way. However, overexpression of this IGF-I transgene leads to hypertrophy of fast (type 2B), intermediate (type 2X or 2A) and slow (type 1) myofibres in the muscles (superficial gluteus) of adult female and male mice and induces a shift towards more oxidative myofibres [45] (Table 1). Unfortunately, this initial report did not indicate the age of the mice used for myofibre measurements and presented data for very few animals (n = 2 for male and n = 1 for female transgenic and control mice). Muscle mass increase was also reported in adult (6 9 months old) male SK733 IGF-I 3 0 SK mice [46] and this increase was equally well pronounced in predominately fast as well as predominately slow muscles: e.g., gastrocnemius mass was increased by 21.5%, tibialis anterior (TA) by 15%, soleus by 20.5% and extensor digitorum longus (EDL) by 11.6%. A recent very detailed characterization of male and female SK733 IGF-I 3 0 SK mice showed that hypertrophy of fast and slow type muscles occurs at different ages: hypertrophy of predominately fast muscles (EDL, TA, gastrocnemius) is detected from 5 to 10 weeks of age, whereas hypertrophy of soleus was seen much later, at 20 weeks of age in females and at 32 weeks of age in males [47]. The MLC/mIGF-I transgenic mouse strain (also developed on the FVB background) carries a transgene constructed using the rat Class 1 IGF-I Ea isoform (migf-i) driven by the MLC promoter (Fig. 2(b)) to restrict expression only to differentiated skeletal muscle cells [48,37,49]. Skeletal muscle hypertrophy in these mice was first detected at day 10 and increased into adulthood. At 2, 6 and 20 months of age mice, exhibited hypertrophy of muscles rich in fast type fibres: e.g., in 6 months old MLC/mIGF-I mice gastrocnemius mass was increased by 56% and EDL mass by 77% and there was a modest shift of myofibre types towards an increased number of type 2B myofibres in EDL muscle (Table 1) [37]. Fig. 3. IGF-I isoforms presumably overexpressed from transgenic constructs in different transgenic mouse strains. (a) The SK733 IGF-I 3 0 SK strain carries human transgenic cdna, which codes for exon 1, exon 4 6 splice variant (Class 1 IGF-I Ea). (b) The MLC/mIGF-I mouse carries transgenic rat cdna, which also codes for the exon 1 exon 4 6 splice variant. In rodents this splice variant can be translated into the IGF-I Ea peptide with 48aa or 22aa signal sequence, whereas in humans an additional 25aa signal sequence can be predicted. (c) The Rska-actin/hIGF-Imouse carries a transgenic cdna coding for mature IGF-I peptide and a signal sequence from the somatostatin gene.

7 10 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) 4 18 Table 1 Characterization of different transgenic mouse strains with skeletal muscle specific overexpression of IGF-I Presence of transgene in blood Systemic IGF-I level Hypertrophy Shift in myofibre types Transgene expression pattern Strain Promoter Transgenic IGF-I Promoter expression pattern SK733 IGF-I 3 0 SK Avian skeletal Class 1 IGF-I Ea Differentiated Type 2B myofibres Fast and slow myofibres More oxidative No change Yes [45,47] a-actin skeletal muscle cells Heart Differentiated cardiomyocytes Spleen MLC/mIGF-I [37] Rat MLC1/3 Class 1 IGF-I Ea Differentiated High in fast myofibres Type 2B myofibres More glycolytic No change Not defined skeletal muscle cells Low in slow myofibres No heart hypertrophy (type 2B) Rska-actin/hIGF-I Rat skeletal Mature IGF-I Differentiated Not defined Male gonads Not defined Increase Yes (Chris Goddard, a-actin skeletal muscle cells No muscle hypertrophy personal Differentiated cardiomyocytes Other organs not defined communication) Hypertrophy of soleus muscle was not seen at any age (2, 6 and 20 months) examined [37]. A recent report on the same 3-month-old male MLC/mIGF-I mice confirmed the absence of hypertrophy in their soleus muscles (Schulze et al. manuscript submitted), however one study showed increased myofibre cross sectional area in soleus muscles of 4 month MLC/mIGF-I mice, which became even more pronounced after physical exercise [50]. The third strain of transgenic mice (produced on the C57Bl6J background) overexpresses the human IGF-I gene driven by rat skeletal a-actin and hereafter are referred to as Rska-actin/hIGF-I mice (Fig. 3(c)). A description of this strain has not yet been published and the data presented within this review are obtained by personal communication (Chris Goddard, South Australia) and our experimental observations (Shavlakadze et al. unpublished observations). The fragment of the rat a-skeletal actin gene used to create a chimaeric construct in Rska-actin/hIGF-I transgenic mice contains 730 bp of the 5 0 -flanking sequence of exon 1, followed by exon 1, the entire first intron, and the 5 0 -end of exon 2 excluding the translation initiation site located in this exon. Instead, a synthetic translation initiation signal (135 bp) from the rat somatostatin gene was inserted between exon 2 of the rat skeletal a-actin gene and the higf-i cdna via a linking sequence. A SV40 poly A signal was inserted after the higf-i cdna to promote a high level of transcription. The construct was designed in this way because the signal sequence for the IGF-I gene was not definitely known when this construct was made in The sequence of higf-i used in the construct codes for the mature IGF-I. In spite of increased levels of IGF-I in skeletal muscles and circulation (discussed later), these mice do not exhibit hypertrophy of fast or slow type myofibres (Shavlakadze et al. unpublished observations): the reason for this have yet to be investigated Local versus systemic availability of transgenic IGF-I It is essential to know whether the overexpressed IGF-I protein is available only to the muscle cells that produce it (autocrine effects) or can act locally on adjacent cells (paracrine effects) or can enter the circulation (endocrine effect). Even if the action of the IGF-I isoform is restricted to an autocrine mode, it may be difficult to resolve whether the IGF-I leaves the cell to bind IGF-I receptors on the surface of the same cell or instead uses an alternative pathway to activate a signalling cascade as suggested for the IGF-I Eb isoform [43]. Whether specific IGF-I isoforms expressed in the skeletal muscle are normally released into the circulation is unclear. Mice with the IGF-I gene deleted in the liver maintain 25% of the total circulating IGF-I and the level of free (i.e., not bound to IGFBPs) circulating IGF-I is not changed (discussed in [51]). The source of this

8 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) remaining IGF-I is yet to be established, however it is speculated that skeletal muscle and adipose tissue might secret systemic IGF-I [52]. Some data suggest that IGF-I is released by damaged muscle. It was demonstrated that in humans IGF-I is released from exercising muscle into the blood stream [53] and in dystrophic mdx mice, where skeletal muscles undergo continuous necrosis and regeneration, increased levels of muscle IGF-I is concomitant with increased IGF-I in the blood, but not in the liver [54]. These results indicate that muscle can be a source of circulating IGF-I, however whether muscle releases IGF-1 containing Ea- or Eb-peptides and/or simply mature IGF-I has not been demonstrated. When evaluating the effects of IGF-I in transgenic animals, it is important to consider the expression pattern of the transgene and availability of the factor beyond the target cell. In transgenic mice, with tissuetargeted overexpression of IGF-I [45,55 57,37] the extent to which the transgenic IGF-I enters the circulation is unclear. However, restricting the effects of transgenic IGF-I to the targeted tissue seems preferable to avoid adverse side effects such as hypertrophy of the heart [58] or visceral organs [47]. Also, it is expected that dramatic changes in IGF-I expression would cause perturbations in the endogenous IGF-I system locally (e.g., tissue levels of endogenous IGF-I, IGF-I receptor and IGFBPs), or have an effect on the whole somatotrophic axis, especially in transgenic mice where excess IGF-I is released into the circulation. The sensitivity of the somatotrophic GH/IGF-I axis in response to changes in circulating IGF-I levels has been demonstrated in mouse models with complete or tissue-specific IGF-I gene deletion [59,52,60,61]; however in transgenic animals with skeletal muscle specific IGF-I overexpression, these issues remain to be addressed. Originally, using a radioimmunometric assay technique, Coleman et al. [45] reported 45-fold increase of skeletal muscle total (endogenous and transgenic higf-i) IGF-I protein content (from a total 6 ng/g in control mice up to 270 ng/g in transgenic mice) and no changes in IGF-I serum levels in adult male SK733 IGF-I 3 0 SK mice. Unfortunately, the report did not indicate the exact age of the mice. Another study performed on the same transgenic male 6 9-month-old mice, using a radioimmunometric assay (however a different kit), reported much higher levels of higf-i (1800 ng/g) in gastrocnemius muscle [46], which is about 7-fold higher than the originally reported value [45]. The same study showed no correlation between gastrocnemius mass and levels of higf-i protein in the muscle and suggested that there was a threshold for IGF-I levels above which no further enhancement of muscle growth occurred [46]. A lack of any circulating higf-i has been widely cited in studies using these transgenic mice [46,62]. However, recent measurements clearly show low (up to 100 ng/ml) higf-i in their plasma, although the overall plasma level of IGF-I (maximum 600 ng/ml) [47] was similar to that already reported [45]. In these same mice [47], muscle levels of higf-i were reported as high as 1000 ng/g at 5 weeks of age and 2000 ng/g at 20 weeks, which is about 4- and 7-fold higher then total IGF-I in muscle reported in the original paper [45] and is in agreement with data published by Criswell et al. [46]. While the two later reports [46,47] state that mice used in their study were derived from the line originally described by Coleman et al. [45], no details are given and it seems likely that either the original study used mice at a much younger age (less than 5 weeks old) or the transgenic lines have diverged with perhaps selection of mice with very high IGF-I expression being used in more recent work [46,47]: such marked differences in expression levels make it difficult to critically compare the various studies. In MLC/mIGF-I transgenic mice constructed using the rat Class 1 IGF-I Ea isoform driven by the MLC promoter to restrict expression only to differentiated skeletal muscle cells [37,49], total levels of muscle IGF- I protein are almost doubled (16 ng/g versus 9 ng/g) at 3 months of age compared with non-transgenic controls [49], whereas serum levels of IGF-I protein remain unchanged (800 ng/ml in both transgenic mice and their non-transgenic littermates) (Shavlakadze et al., unpublished observations). Available evidence indicates that transgenic migf-i remains within skeletal muscle in the transgenic mice and does not enter the circulation, however this cannot be confirmed since antibodies currently available for detection of IGF-I protein concentrations do not distinguish between the mouse (endogenous) and rat (transgenic) IGF-I. While the migf-i may not enter the blood stream in MLC/mIGF-I transgenic mice, this situation is complicated by recent observations that overexpression of this transgene upregulates other endogenous IGF-I isoforms that may potentially enter the circulation [38]. In particular, there is a pronounced upregulation of Class 1 IGF- I Eb (possibly corresponding to MGF) and lesser amounts of Class 2 containing isoforms within muscles [38]: the extent to which these elevated endogenous isoforms may enter the circulation or have a paracrine effect is not known. Clearly, such a situation raises the possibility of indirect systemic effects of migf-i overexpression in transgenic mice. In the Rska-actin/hIGF-I transgenic mice where higf-i overexpression was driven by the promoter for rat skeletal a-actin and a signal sequence from human somatostatin gene was included (Goddard et al., personal communication), total plasma IGF-I was elevated (60% increase from the native level of 140 ng/ml). The consequence of elevated circulating levels of human IGF-I in these transgenics is demonstrated by a high incidence of enlarged seminal vesicles in male mice aged more than 24 months (Shavlakadze et al., unpublished

9 12 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) 4 18 observations). In muscles of these transgenic mice, the total IGF-I concentration was increased about 15-fold (200 ng/g compared with 14 ng/g in controls), although endogenous mouse IGF-I levels were halved (7 ng/g). While this level of higf-i transgene expression (200 ng/g) is about 10-fold less (2000 ng/g) than reported in skeletal muscle by others [47], the striking (60%) increase in total serum IGF-I in the former (Chris Goddard et al., personal communication), is in marked contrast with little or no change in total serum IGF-I in the latter [47]. While a promoter used in this transgenic strain is similar to one used in SK733 IGF-I 3 0 SK strain [45] other critical aspects (e.g., signalling sequence) of the transgene should account for the release of the factor into the circulation. It is difficult to reconcile these disparate patterns of higf-i expression and such results emphasize the complexity of comparing different lines of the transgenic mice with the skeletal muscle specific overexpression of IGF-I. The situation is further complicated by striking differences between circulating levels of IGF-I reported in normal mice, which were used as background strains for the transgenic strains. For example, ng/ml of IGF-I is reported in serum of FVB mice (used as a background strain for SK733 IGF-I 3 0 SK and MLC/ migf-i transgenic lines) [45,47] in contrast to 140 ng/ml of IGF-I reported in plasma of C57Bl/6J mice (used as a background strain for the Rska-actin/hIGF- I transgenic line) (Chris Goddard, personal communication). In C57Bl10 mice at 8 10 weeks of age, one study reported 700 ng/ml of IGF-I in plasma [54], which is only 1.5-fold higher then that detected by Fiorotto et al. [47] in 10 weeks old FVB mice. However, the former study reports an increase of circulating IGF-I level to up to 1200 ng/ml at 20 weeks in C57Bl/10 mice [54], whereas similar levels of circulating IGF-I are reported at 10, 20 and 32 weeks of age in FVB mice [47]. These discrepancies in the absolute numbers reported for circulating IGF-I levels by different groups might be partly attributed to the use of different strains and the method of IGF-I protein measurement used in the studies. Two important aspects should be taken into consideration when performing IGF-I measurements: (i) if the method does not allow complete separation of IGF-I from serum IGFBPs, the numbers obtained will be spuriously high because the assay will measure IGFBPs as an IGF peptide and (ii) IGF-I standards are notoriously variable, even in commercially available assays, so if the same standard is not used in each assay compared, the absolute numbers become meaningless (Chris Goddard, personal communication) Effects of different promoters on transgene expression Two main promoters have been used to drive overexpression of the various IGF-I isoforms specifically in skeletal muscle: the promoters for skeletal a-actin [63] and myosin light chain (MLC1/3) [64]. To compare various studies reporting effects of IGF-I on myogenesis and regeneration, it is crucial to know exactly which cells activate the promoter that controls overexpression of the IGF-I transgene. The promoter for the skeletal a-actin gene is active in differentiated skeletal and cardiac muscle cells [63]. Skeletal a-actin is the major actin isoform in adult skeletal muscle [63] and, while greater expression of the transgene is seen in skeletal muscle compared with the heart [45,58], heart hypertrophy is noted in SK733 IGF-I 3 0 SK mice with the avian skeletal a-actin promoter [58]. In tissue cultured C2C12 myogenic cells, transcripts of skeletal a-actin are not detected in proliferating or undifferentiated myoblasts and accumulate only after differentiation [65] and, during early embryogenesis in mice cardiac actin is the major form of sarcomeric actin expressed initially whereas skeletal actin starts to accumulate in more differentiated myotomal cells [66]. These in situ and in vivo data indicate that when driven by the skeletal a-actin promoter, the IGF-I transgene is not expressed within quiescent or replicating (undifferentiated) myoblasts and thus this transgene (expressed only by differentiated muscle cells) can be expected to have only a paracrine or endocrine effect on satellite cell (myoblast) replication: in vivo this could occur only if exogenous transgenic IGF-I is available from another source (e.g., the circulation or adjacent myofibres) to act upon satellite cells. This is similar to the situation for the MLC/mIGF-I construct [37,49] that is also not expressed in undifferentiated myoblasts. In SK733 IGF-I 3 0 SK mice, despite the normally ubiquitous expression of endogenous skeletal a-actin in all muscle types, expression of the higf-i transgene in skeletal muscle was restricted to type 2B myofibres [47]. Yet these transgenic mice are characterized by increased size of all types of myofibres [45] indicating some paracrine action of the transgene (in the absence of expression in all myofibres). The soleus muscle, which does not contain type 2B myofibres was unaffected in young mice, but increased in mass in old mice (at 20 weeks of age in females and at 32 weeks of age in both females and males) [47]. Enhanced proliferation of myogenic cells following denervation reported in SK733 IGF-I 3 0 SK mice [62] can similarly only be accounted for by a paracrine, systemic or else an indirect (upregulation of other factors) action of the transgenic IGF-I, since the promoter for the skeletal a-actin gene in skeletal muscle is expressed only in differentiated muscle cells [66,63]. A report that cultured myoblasts from SK733 IGF-I 3 0 SK mice show elevated levels of higf-i in the supernatant [67] supports secretion of higf-i but the result is unexpected since this promoter is normally not expressed in undifferentiated myoblasts.

10 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) The other promoter used to overexpress IGF-I specifically in skeletal muscle is the MLC1/3 promoter, used in MLC/mIGF-I mice [37]. Activation of this promoter is restricted to differentiated skeletal muscle cells (it is not expressed in cardiac muscle), with high levels of expression in the fast-twitch myofibres (in the order 2B, 2X and 2A) and low levels in slow-twitch (type I) myofibres [68]. Initially, the increased size of only type 2B myofibres was reported in MLC/mIGF-I mice [37]. However, a later study by the same group [49] showed that in mdx mice the expression of MLC/mIGF-I transgene improves the phenotype of many muscles including the diaphragm, which lacks type 2B myofibres [69]. It is now apparent that while histochemical staining shows no type 2B myofibres in diaphragm muscle of mdx mice [69], myosin specific antibodies reveal type 2B myofibres in the diaphragms of older mice, which increase in number most likely due to regeneration following muscle necrosis [70]. In addition, another study reported increased myofibre size in slow-twitch soleus muscles of 4 month old MLC/mIGF-I mice, but the authors suggested that such hypertrophy should be attributed to increased load on neighbouring muscles, rather than a direct effect of the transgene [50], since the MLC/ migf-i transgene is expressed in soleus at very low levels [37]. Recent data showing that the migf-i isoform can upregulate other endogenous IGF-I isoforms raises the possibility of indirect paracrine (or even systemic) effects of this transgene [38]. Moreover, an increased number of Sca1+ stem cells has been reported recently in the bone marrow and skeletal muscle of MLC/mIGF-I mice compared to control littermates following cardiotoxin muscle injury [71], supporting a systemic or indirect effect of local migf-i overexpression Other variables to consider Comparison of the literature is complicated further by the nature of the experimental model used to study muscle responses to overexpression of transgenic IGF- I. Such models include exercise-induced hypertrophy [50], ageing [37], chemical injury [37,71], muscle grafting [72], muscle denervation followed by re-innervation [62], permanent denervation [73], dystrophy [49,70], unloading induced atrophy [46], muscle atrophy following chronic cardiac dysfunction (Schulze et al., manuscript submitted) and angiotensin II induced congestive heart failure [74] (Table 2): these all have distinctive features that need to be taken into account. For example, although both muscle injury and denervation lead to activation of satellite cells, the environment and therefore the signals required for satellite cell activation are very different. Injury leads to massive necrosis of the myofibres, an acute inflammatory response and activation of many satellite cells, with onset of DNA synthesis in myoblasts starting from about 24 h after injury and myotubes formed de novo by about 3 days (reviewed in [75]). In contrast, during the early stages of denervation the morphology of the myofibres is preserved, the classical inflammatory response does not occur and relatively very few satellite cells are activated [76,77]. Over time, denervated muscles decrease in size (atrophy) and this is followed by a steady decline in the number of satellite cells [78]. Following denervation autoradiographic studies in mice show progressive loss of labelled nuclei adjacent to muscle fibres [79]. Thus, in permanently denervated muscles the protective effect of IGF-I overexpression on myofibre atrophy is likely to be due to enhanced survival through increased protein synthesis or decreased protein breakdown. A protective effect of the transgenic migf-i on myofibres is also evident in the transgenic mice overexpressing a mutated version of the human superoxide-dismutase 1 (SOD1) gene, which represents an animal model of amyotrophic lateral sclerosis (ALS) (Dobrowolny et al., manuscript submitted). Similarly, in the mouse model of chronic cardiac dysfunction (Schulze et al. manuscript submitted) and congestive heart failure [74], where overexpression of migf-i prevents muscle atrophy (seen in non-transgenic mice under the same conditions), protection by migf-i is due to a direct effect within differentiated myofibres, involving inhibition of the ubiquitin-proteosome pathway in case of chronic cardiac dysfunction (Schulze et al., manuscript submitted) and decreased activation of pro-apoptotic caspase-3 in congestive heart failure. The nature of the experimental model can also affect the expression pattern of the transgene: for example, denervation (as also occurs in muscle grafting), significantly reduces skeletal a-actin expression [80] and thus might reduce the expression levels of any IGF-I transgene driven by this promoter. In summary, different experimental conditions can affect the expression pattern, expression level and availability of transgenic IGF-I and such variations should be accounted for when reconciling data reported using different experimental models. It is noted that the restricted availability of IGF-I isoforms in transgenic situations may not reflect the normal availability of the same isoform: thus the transgenic situation may produce quite different results from administration (in tissue culture or in vivo) of exogenous IGF-I that is freely available to all cells. Furthermore, the artificial transgenic situation where cdna for a specific IGF-I isoform is driven by the promoter of an unrelated gene, means that IGF-I overexpression is disconnected from the normal cellular signalling that stimulates transcription of the IGF-I gene (along with a cohort of related genes). In light of growing support for the hypothesis of integrated control of gene expression where non-coding regions of DNA also contribute

11 Table 2 Beneficial effect of IGF-I overexpression reported in different experimental transgenic mouse models and some related clinical conditions in humans Experimental model in transgenic mice Strain Effect of IGF-I overexpression Clinical condition in humans Young adult and 22 month old mice [37] MLC/mIGF-I Muscle hypertrophy Age related sarcopenia Protection from age-related myofibre atrophy (sarcopenia) Cardiotoxin injured muscle [71] MLC/mIGF-I Increased muscle regeneration Muscle injury Auto-transplanted whole muscle graft [72] MLC/mlGF-l No effect on early events during regeneration Muscle injury Transplantation Adult dystrophic mdx mouse [49] mdx/migf-i Improved dystrophic phenotype Duchenne muscular dystropy Myofibre hypertrophy Young dystrophic mdx mouse [70] mdx/migf-i Protection of dystrophic myofibres from necrosis Duchenne muscular dystropy Denervated muscle [73] MLC/mIGF-I Protection of myofibres from denervation atrophy Immobilization Spinal cord injury Denervation Disuse Space flight Atrophic muscle in chronic cardiac dysfunction (Schulze et al., submitted) MLC/mIGF-I Protection of myofibres from atrophy Cachexia Atrophic muscle in congestive heart failure [74] MLC/mIGF-I Protection of myofibres from atrophy Cachexia ALS (Dobrowolny et al. submitted) SOD1/mIGF-I Protection of both muscle and nerve from degeneration ALS Spinal cord injury Hind limb unloading [46] SK733 IGF-I 3 0 SK No effect on myofibre atrophy Immobilisation Disuse Space flight Adult mouse [45,47] SK733 IGF-I 3 0 SK Muscle hypertrophy Sarcopenia Sciatic nerve crush followed by re-innervation [62] SK733 IGF-I 3 0 SK Increased re-innervation & improved recovery of muscle mass Muscle denervation Immobilisation Disuse Space flight 14 T. Shavlakadze et al. / Growth Hormone & IGF Research 15 (2005) 4 18