Why are some amyloidoses systemic? Does hepatic “chaperoning at a distance” prevent cardiac deposition in a transgenic model of human senile systemic (transthyretin) amyloidosis?

Abstract

In the human systemic amyloidoses caused by mutant or wild-type transthyretin (TTR), deposition occurs at a distance from the site of synthesis. The TTR synthesized and secreted by the hepatocyte circulates in plasma, then deposits in target tissues far from the producing cell, a pattern reproduced in mice transgenic for multiple copies of the human wild-type TTR gene. By 2 yr of age, half of the transgenic males show cardiac deposition resembling human senile systemic amyloidosis. However, as early as 3 mo of age, when there are no deposits, cardiac gene transcription differs from that of nontransgenic littermates, primarily in the expression of a large number of genes associated with inflammation and the immune response. At 24 mo, the hearts with histologically proven TTR deposits show expression of stress response genes, exuberant mitochondrial gene transcription, and increased expression of genes associated with apoptosis, relative to the hearts without TTR deposition. These 24-mo-old hearts with TTR deposits also show a decrease in transcription of inflammatory genes relative to that in the younger transgenic mice. After 2 yr of expressing large amounts of human TTR, the livers of the transgenic mice without cardiac deposition display chaperone gene expression and evidence of an activated unfolded protein response, while the livers of animals with cardiac TTR deposition display neither, showing increased transcription of interferon-responsive inflammatory genes and those encoding an antioxidant response. With time, in animals with cardiac deposition, it appears that hepatic proteostatic capacity is diminished, exposing the heart to a greater load of misfolded TTR with subsequent extracellular deposition. Hence systemic (cardiac) TTR deposition may be the direct result of the diminution in the distant chaperoning capacity of the liver related to age or long-standing exposure to misfolded TTR, or both.—Buxbaum, J. N., Tagoe, C., Gallo, G., Walker, J. R., Kurian, S., Salomon, D. R. Why are some amyloidoses systemic? Does hepatic “chaperoning at a distance” prevent cardiac deposition in a transgenic model of human senile systemic (transthyretin) amyloidosis?

The pathology of the systemic amyloidoses is defined by extracellular Congophilic fibril deposition occurring in locations distant from the site of synthesis of the precursor protein (1). Molecularly, the amyloidoses represent a subclass of protein misfolding disorders in which aberrant conformations are predisposed to aggregate and acquire toxic properties in vivo and in vitro (2). The fibril precursor in the systemic disorders is usually a variant of a normally secreted protein. The extracellular deposition of such aggregates requires that the cellular mechanisms responsible for protecting the synthesizing cell from toxic aggregates function well enough to allow potentially (or minimally) misfolded molecules to engage the secretory mechanism faster or to a greater extent than they can be shunted into the endoplasmic reticulum-associated degradation (ERAD) machinery or other proteolytic pathways (3). Alternatively, the precursors may be secreted from the cell in the company of another molecule, effectively being “chaperoned” through the secretory pathway and into the circulation. If these mechanisms are inadequate, intracellular aggregation and subsequent dysfunction occur in the synthesizing cell (4, 5). After clearing the secretory hurdle, the systemic deposition of any amyloid can occur by a variety of possible mechanisms at a distance from the site of synthesis.

In the human transthyretin (TTR) amyloidoses familial amyloidotic polyneuropathy (FAP), familial amyloid cardiomyopathy (FAC), and senile systemic amyloidosis (SSA), there are no deposits in the liver, the major site of TTR production (6). However, in some instances, local deposition of fibrils composed of mutant TTR has been described at sites of synthesis other than the liver, including the choroid plexus, retinal epithelium, kidney, and gut, indicating that not all the normal sites of TTR synthesis are created equal with respect to local deposition (7, 8). More striking is the fact that in SSA and FAC, there is extensive deposition in the heart, even though TTR synthesis or TTR gene transcription has not been detected in cardiac cells.

We have generated animals transgenic for ∼90 copies of the wild-type human TTR gene expressed under the control of its own promoter (9). The human TTR gene shows appropriate tissue specific expression, and the animals show tissue deposition patterns similar to those in humans. They also show an analogous degree of age dependence with little or no deposition until middle to late adulthood (12 mo) and no histologically discernible deposition in the heart until after 18 mo of age. In the current studies, we have used the analysis of transcription in the major TTR-producing tissue (liver) and the primary target organ (heart) to obtain a global view of the process. The transgenic mice display a phenomenon that has not previously been described, which we have called “chaperoning at a distance.” It is characterized by the expression of genes encoding molecules involved in various aspects of protein homeostasis (3) at the site of TTR synthesis, i.e., the liver, and is associated with the absence of deposition in a distant target organ, the heart. In 2-yr-old mice with cardiac TTR deposition, there appears to be little or no such transcriptional activity in the liver.

MATERIALS AND METHODS

The experimental groups included animals transgenic for a construct consisting of ∼90 copies of a wild-type human TTR gene with all its known regulatory sequences being expressed in an appropriate tissue-specific manner with immunohistochemically documented TTR deposits in heart, kidneys, or both; transgenic animals in which the same tissues were immunohistologically negative for human TTR deposition with the same antibody (Dako A0002; Dako Corp., Carpenteria, CA, USA; ref. 9); and nontransgenic animals of the same background strain as the transgenics were used as controls. The endogenous murine TTR genes were intact. The serum concentrations of human TTR in the two groups of TTR transgenic mice are shown in Table 1.

Table 1.

Serum transthyretin concentrations (mg/ml) in transgenic mice

Parameter	OTGND	OTGWD	OTGND	YTG	OTGWD	YTG
Values (n)	9	7	9	7	7	7
Mean	0.82 ± 0.38	1.23 ± 0.54	0.82 ± 0.38	0.76 ± 0.46	1.23 ± 0.54	0.76 ± 0.46
P	0.05		0.8736		0.1098

OTGND, old (24 mo) transgenic with no cardiac deposition; OTGWD, old transgenic with cardiac depositon; YTG, young (3 mo) transgenic.

RNA was extracted from frozen hearts and livers that had previously been examined histologically from 3- and 24-mo-old transgenic and control C57BL/6 × DBA/2 F1 male mice (15/group; ref. 9). Total RNA was extracted from the tissues using TRIzol (Invitrogen, Carlsbad, CA, USA). Two equal groups of 3-5 animals were utilized as biological replicates for each experimental condition. Equal amounts of RNA from each member of the replicate groups were pooled to generate 2 independent RNA pools for each experimental condition. Duplicate chips were used for each RNA pool (technical replicates). cRNA preparation and hybridizations were carried out using standard Affymetrix (Santa Clara, CA, USA) protocols.

Our initial cardiac experiments were performed with a custom chip containing 36,000 nonredundant genes that was produced collaboratively by the Functional Genomics Institute of the Novartis Foundation and Affymetrix, and represented a prototype of the Affymetrix Mu 430A chip that we used for the bulk of the analyses (10). The two chips had similar, but not identical, genomic representation. Analyses were performed using Affymetrix software. The levels of transcription of the 36,000 probe sets varied over 4 orders of magnitude. We noted redundancy of probes for a number of genes, although the proportion of redundant probes in the total array was not precisely determined. In all but two instances, the multiple probes showed similar significant differences in the same direction, adding confidence to the findings for those transcripts and the integrity of the methods. In the rare case when there were discrepancies among the different probes for the same gene, that gene was dropped from the analysis.

In the analysis, we avoided utilizing any arbitrary fold-change threshold to assess the relevance of transcriptional differences in the comparisons between different experimental states. We have instead depended on a preset level of statistical significance, coupled with concurrent changes in functionally related genes, to establish biological significance. We focused on transcripts with differences between tissues of young transgenic and young nontransgenic animals (transgene effect on 3-mo-old animals, before tissue deposition can be detected), old transgenic animals with and without tissue deposition (TTR deposition effect in animals 24 mo old), and young and old transgenic animals (the latter with and without cardiac deposition). The differentially expressed transcripts with values of P ≤ 0.005 using a 2-sample t test with a random variance model were selected for subsequent analysis. An exact multivariate permutation test (based on 1000 available permutations, allowing a maximum of 10% false positives) was also used. Any genes not identified after these filters were applied were assumed to be equally transcribed in both members of the comparison.

Separately, the entire normalized gene data set was also analyzed using a weighted networks tool that seeks groups of genes, or modules, that are related in a given class comparison by their statistically significant tendency to vary in concert (11, 12). The purpose of this approach is to identify connected molecular networks within large gene expression data sets in a manner that is not influenced by prior knowledge of their molecular functions. The genes comprising these modules are based on connectivity as the metric. These were compared with the distributions in our functional analyses based on differential gene expression. Functional networks were further defined using Ingenuity software (Ingenuity Systems, Redwood City, CA, USA).

TLDA microfluidic cards (384-well; Applied Biosystems, Foster City, CA, USA) were used to validate differential gene expression of 63 genes and 1 endogenous housekeeping control (18s ribosomal RNA) each in liver and heart. For reverse transcriptase quantitative polymerase chain reaction (RT-qPCR), cDNA was transcribed from 2 μg total RNA using the cDNA archive kit (Applied Biosystems). All transcript assays consisted of 2 unlabeled PCR primers and a FAM dye-labeled TaqMan MGB probe, prespotted on the TLDA card. All amplifications were done in triplicate, and threshold cycle (C_t) and scores were averaged to calculate relative expression. The C_t scores were normalized against 18S ribosomal RNA controls by the RQ Manager 1.2 software (Applied Biosystems).

qPCR reactions using SYBR Green dyes were used to quantify a smaller series of mRNAs of particular interest, including that encoding human and mouse TTR and those encoding Xbp1 spliced (Xbp1s) and unspliced (Xbp1u). Cells treated with staurospaurine served as positive controls (13). The primers for those analyses are shown in Supplemental Table S1.

RESULTS

Comparison of transcription patterns of the livers of TTR transgenic and nontransgenic littermates at 3 mo of age show almost 400 nonredundant transcripts that differ in abundance (Table 2). Of these, 157 identified and annotated genes are more highly transcribed in the nontransgenic livers, while a bit more than twice as many (327) are increased in the transgenics. The classes of genes showing increases in the young transgenic animals include those involved in intracellular and secretory trafficking and those encoding proteins characteristic of inflammation/immunity (Table 2, Fig. 1). The increases are statistically significant when calculated on the basis of their contribution to the total number of transcripts increased in the transgenics relative to the nontransgenic mice. Genes encoding extracellular matrix molecules were also more abundant, but the increases were not statistically significant (see Supplemental Table S2 for genes included in each functional class). There are also increases in the number of genes encoding proteasome constituents and the ERAD elements Derlin 1 and 2 that contribute to the total increase in differentially transcribed genes in the transgenics. In the livers of the age- and sex-matched nontransgenic mice, only the number of transcripts encoding solute carrier proteins and transcription factors are significantly greater than in the transgenics.

Table 2.

Differences in functional groups of hepatic transcripts between experimental cohorts

Parameter	Young			Old
	YNTG > YTG	YTG > YNTG	P	OTGWD > OTGND	OTGND > OTGWD	P
Total transcript differences	230	392		121	75
Total identified transcript differences	171	355		106	68
Total identified and annotated differences	157	327		105	64
Inflammation	10	43	0.118	13	7	NS
Chaperones	0	2	NS	0	8	0.001
Proteasome	5	15	NS
Ubiquitin system	2	5	NS	3	1	NS
Autophagy	1	2	NS
Antioxidants	0	4	NS	7	2	NS
Apoptosis	5	9	NS	4	0	NS
Mitochondrial	5	15	NS	5	1	NS
Transcription factors	12	17	0.08	7	3	NS
Kinases	6	12	NS	6	1	NS
Extracellular matrix molecules	1	8	NS	0	1	NS
Ribosomal	0	1	NS
Mitochondrial ribosomal	1	2	NS
Proteases	2	0	NS
Channels	8	5	0.008
Secretory apparatus	4	23	0.17	5	5	NS
Peroxisomes	0	1	NS

YNTG, young (3 mo) nontransgenic; YTG, young transgenic; OTGWD, old (24 mo) transgenic with cardiac depositon; OTGND, old transgenic with no cardiac deposition; NS, not significant.

Figure 1.

Distribution of hepatic transcripts encoding genes of the identified functional classes in 3-mo-old TTR transgenic and nontransgenic mice, neither of which show tissue deposition. A) Transcripts that are more abundant in young nontransgenic mice relative to age-matched transgenic animals are represented as the proportion of the total number of increased transcripts that could be assigned to those functional groups (see Table 1). B) Distribution among transcripts that are relatively increased in the transgenic animals.

Comparison of livers from 24-mo-old age and sex-matched transgenic animals with and without cardiac deposition show many differences, many of them concentrated in a group of genes involved in proteostatic function (P=0.001; Table 3 and Fig. 2). These include the HSP 90 cofactor Aha1; HSP 70 family member HSP 110; the transcription factor Xbp1; the small heat-shock proteins DNA J (HSP 40) subfamilies A and B, which serve as HSP 70 cochaperones; β3 crystallin; and 2 members of the ubiquitin system. The livers from animals with cardiac deposition show no relative increases in transcripts encoding known chaperones or cochaperones, but relative increases in those encoding tribbles 3 and ubiquitin peptidase 18.

Table 3.

Differences in functional groups of cardiac transcripts related to age and presence of TTR deposition

Parameter	Young			Old
	YNTG > YTG	YTG > YNTG	P	OTGWD > OTGND	OTGND > OTGWD	P
Total transcripts that differ	294	257		533	139
Total identified transcripts	255	224		516	125
Transcripts identified and annotated	228	198		490	120
Inflammation	4	42	0.0001	24	61	<0.0001
Chaperones	2	0	NS	17	1	0.03
Proteasome/ubiquitin system	7	1	0.03	25	2	0.01
Autophagy	4		0.052	3	1	NS
Anti-oxidants	3	3	NS	13	0	0.02
Apoptosis	1	2	NS	21	1	0.01
Mitochondrial	15	4	0.007	64	2	<0.0001
Transcription factors	14	19	NS	7	4	NS
Kinases	2	3	NS	12	3	NS
Extracellular matrix	2	2	NS	1	4	0.02
Ribosomal	3	10	0.09	4	0	NS
Mitochondrial ribosomes	3	0	NS	9	0	0.12
Proteases	3	4	NS			NS
Channels	8	3	NS	7	1	NS
Secretory	11	11	NS	16	1	0.05
Peroxisomal	5	1	0.11	4	0	0.58
Ras-signalling	14	6	0.06	6	1	NS
Cardiac function	24	15	0.08	12	6	NS
Olfactory	3	13	0.02	0	4	0.005

YNTG, young (3 mo) nontransgenic; YTG, young transgenic; otgwd, old (24 mo) transgenic with cardiac depositon; OTGND, old transgenic with no cardiac deposition; NS, not significant.

Figure 2.

Distribution of hepatic transcripts encoding genes of the identified functional classes in 24-mo-old TTR transgenic mice. A) Transcripts that are relatively more abundant in livers of animals that have cardiac TTR deposition, represented as the proportion of the total number of increased transcripts that could be assigned to those functional groups (see Table 1). B) Distribution of hepatic transcripts that are increased in TTR transgenic mice that do not have cardiac TTR deposition. C) Hepatic transcriptional networks evident in the two groups of animals, as defined by Ingenuity software (Ingenuity Systems, Inc.). Genes depicted in red represent the network that is more highly activated in livers of animals in which there is cardiac TTR deposition. Shown in green is a transcriptional network found to be more highly expressed in the livers of TTR transgenic mice in which there is no cardiac deposition.

Approximately 10% of the relatively increased transcripts in livers from 2-yr-old animals with or without cardiac deposition are associated with inflammation or the immune response. In the livers of animals with cardiac deposition, most of the inflammatory genes appear to be interferon driven, suggesting some form of categorical response, while those increased in livers of animals without cardiac deposition encode a small number of diverse molecules (Supplemental Table S2).

In the hearts of 3-mo-old mice, there are ∼550 differentially expressed genes between transgenic and nontransgenic mice, with 426 of those identified and annotated (Table 4). The number of relative increases is evenly divided between the two groups of hearts, but the distribution across functional groups is strikingly different (Fig. 3A). Hearts from the nontransgenic mice have significant relative increases in genes encoding mitochondrial proteins and components of the proteasome apparatus. They have almost significant differences in cardiac structural molecules (P<0.08), transcripts encoding signaling pathways involving Ras-associated genes (P<0.06) and autophagy (P<0.052), transcripts presumably reflecting normal cardiac function at 3 mo of age, which are relatively decreased in the transgenic mice. Although cardiac TTR deposition has never been seen before 15 mo of age in this model, the hearts of the 3-mo-old transgenic mice display a highly significant increase in the number of transcripts associated with the inflammatory response and a lesser, but statistically significant, relative increase in the number of transcripts encoding olfactory receptor genes, albeit they are of relatively low intensity and of unknown significance.

Table 4.

Differential expression of Xbp1s, Xbp1u, and hTTR in the hearts and livers of hTTR transgenic mice with and without cardiac TTR deposition and controls

Sample	Xbp1s		Xbp1u		hTTR
	Fold change	P	Fold change	P	Fold change	P value
Heart
Positive control vs. YTG	84.00
Positive control vs. OTGWD	353
Positive control vs. OTGND	521
YTG vs. OTGWD	−5.6	0.14	−1	0.76	49.048105	0.42
YTG vs. OTGND	−5.7	0.27	−2.5	0.36	26.037647	0.48
OTGWD vs. OTGND	−1	1	−2.4	0.37	−1.883738	0.62
Liver
YTG vs. YNTG	−7.4	0.23	−4.4	0.006	−104267	4E-07
YTG vs. OTGND	−8.5	0.05	−3.1	0.001	−5	0.0003
YTG vs. OTGWD	−16	0.02	−2.4	0.44	−1.8	0.011
OTGND vs. OTGWD	−1.9	0.15	−1.3	0.8	−2.6	0.001

YTG, young (3 mo) transgenic; OTGWD, old (24 mo) transgenic with cardiac depositon; OTGND, old transgenic with no cardiac deposition; YNTG, young nontransgenic.

Figure 3.

Distribution of cardiac transcripts encoding genes of the identified functional classes in TTR transgenic mice. A) Relative abundance of transcripts increased in young TTR transgenic and nontransgenic control mice. Blue bars represent the proportion of transcripts increased in nontransgenic mice relative to those carrying the human wild-type TTR gene. Red bars represent the functional groups of transcripts showing relative increases in young transgenic relative to nontransgenic mice. Transgenic mice do not show cardiac deposition at this time. B) Green bars represent the functional groups of transcripts that are increased in hearts with deposition. Yellow bars represent the relatively increased transcripts, according to functional group, in transgenic animals that have no detectable cardiac TTR deposition. Numbers in parentheses in A and B indicate the number of transcripts that are included in the analysis (see Table 1). C) Major networks (as defined by Ingenuity analysis) expressed in the hearts with (red) and without (green) cardiac deposition.

The hearts of the 24-mo-old transgenic animals with documented cardiac TTR deposition have increases in almost 500 transcripts relative to age-matched hearts without deposition. They show significant increases in transcripts encoding mitochondrial genes, genes for proteasome and ubiquitin system elements, proteins of the secretory pathway, redox enzymes, and molecules involved in apoptosis. The hearts without deposition have relative increases in 120 of the identified and annotated genes. Sixty-six percent of those represent inflammation-associated genes, almost all of which are characteristic of the innate immune response, including the cytokines IL1β and lymphotoxin B, with a minor portion related to T-cell activation (Fig. 3B). The inflammatory genes increased in the hearts with deposition show evidence of a polyclonal Ig response and cytokine receptor transcripts, but no Ig response encoding cytokines per se. The number and proportion of inflammatory genes transcribed in the hearts of animals with deposits are significantly lower than in the comparably aged animals without deposits (P<0.0001), as well as being reduced relative to the number increased in the younger transgenic animals before any deposition occurs (P<0.0001).

We selected 2 sets of 64 transcripts, one hepatic, one cardiac, of varying degrees of abundance for validation. In the cardiac data set, ∼70% of the changes were validated in both the direction and magnitude of the fold change. In the liver comparisons, 90% were similarly validated. It is likely that the larger number of apparent false positives in the cardiac experiments was related to the number of nonannotated and falsely annotated probes on the early-generation chips. Nonetheless, the fact that we based our analyses on functional groups of genes rather than single genes minimized the interpretive consequences of the lower validation rate among the cardiac transcripts.

Our microarray analyses using murine-specific arrays could not detect human TTR transcripts in the transgenic animals. We chose experimental animals to have comparable serum levels of human TTR in an attempt to ensure that the levels of the hepatic transcripts of human TTR were similar. However, as shown in Table 1, differences between the old animals with and without cardiac deposition approached significance. In an effort to determine whether the unfolded protein response (UPR) was activated in the tissues of the mice, we measured the levels of mRNA encoding Xbp1 and Xbp1s, the latter being the transcription factor that is pathognomonic for activation of the UPR (13). qPCR of cardiac and hepatic RNAs from the various experimental groups were performed using primers for murine TTR, human TTR, unspliced Xbp1u, and spliced Xbp1s.

The cardiac RNAs from the nontransgenic animals gave no signal above background when amplified with primers specific for human or murine TTR, using probes for 18S ribosomal RNA as controls (Table 4). Consistent with the lack of transcription of the TTR gene of either species in the heart, using the same primers, there were no significant differences in the comparison of cardiac murine and human TTR RNAs among any of the experimental groups. When we compared mice with and without cardiac TTR deposition, we found no significant differences in cardiac transcription of Xbp1u or Xbp1s. Similar studies of liver RNAs indicated that Xbp1s mRNA was more abundant in the 24-mo-old animals without cardiac deposition, although the difference was not statistically significant. However, the level of Xbp1s mRNA in the livers of the 24-mo-old animals was significantly lower than in the 3-mo-old transgenic mice. Both human and murine TTR RNAs were less abundant in the livers of the mice without cardiac deposition.

DISCUSSION

One caveat in our analysis is that we are analyzing tissues that comprise a variety of cell types. While the dominant cell in the respective tissues may be a cardiomyocyte, or hepatocyte, there are also fibroblasts, endothelial cells, vascular smooth muscle cells, and a mixed population of cells derived from the circulation, including mast cells, monocytes and macrophages, neutrophils, lymphocytes, and dendritic cells. Although we lose a degree of cellular precision, we believe this approach strengthens the analysis since complex biological systems are composed of networks of cells, as well as interacting molecular pathways. Thus, we have chosen to use the whole organ as the unit of response and to interpret the experiments in the context of published data, and our observations in isolated cell populations, intact tissues and whole organisms.

A second limitation of microarray analysis is its inability to track functionally critical events that do not require transcription, e.g., post-translational modifications, although transcription in one or many genes occurring as the result of protein modification may be identified and the original stimulus may be inferred or identified by other technologies. Hence differences in phosphorylation states of various components of the UPR would not be seen in our comparisons, although the transcription of Xbp1 and its spliced form Xbp1s should be evident (see below). Unfortunately, because of technical issues related to tissue storage, we were unable to perform the protein studies, which would have allowed us to absolutely confirm the functionality of the transcripts observed to be increased.

The strength of the microarray methodology is its breadth and its lack of bias, allowing the quantitation of expression of every gene being transcribed in a given tissue under a given set of conditions. The analysis encompassed transcription in an in vivo steady-state condition rather than one reflecting an acute perturbation. We also focused on changes in the expression of groups of functionally related genes, gaining insights that might not be available from the detailed examination of a single gene.

While every cell has the capacity to chaperone proteins in the interest of their own well being, mammalian hepatocytes and plasma cells synthesize large numbers of secreted proteins required for organismal survival. Hence, they have highly developed ER networks and mechanisms for assisting protein folding or degrading misfolded end products (14, 15). Quantitatively, because a certain fraction of every protein has the potential to misfold and aggregate, the absolute misfolded protein load may exceed the capacity of the cells' “chaperome” (15). When misfolded, potentially fibrillogenic proteins are secreted, and they represent a risk to other organs and tissues unless there is a postsecretory mechanism to maintain their stability, e.g., ligand binding or protein-protein interactions that provide chaperone-like functions in the extracellular space. Other mechanisms for disposal, perhaps uptake and/or proteolysis mediated in some immunological or inflammatory, manner may also be involved.

A third consideration is the apparent requirement for considerable overexpression of the human TTR to obtain a pathological phenotype. The only reported successful TTR deposition models in which the normal pattern of tissue expression has been retained have been in mice in which 30 or more copies of the human gene are expressed or animals in which either the endogenous TTR gene or the hsf1 gene or both have been silenced (16, 17). The need for substantial transgene overexpression is likely to be related to the formation of mixed murine-human TTR heterotetramers. The murine tetramer is much more kinetically stable than the human form, and the presence of one murine subunit is sufficient to stabilize the heterotetramer (18). Hence, the tetramer dissociation yielding monomers, which subsequently misfold and aggregate, only occurs in pure human tetramers. To overcome the human/mouse stoichiometry, one must overexpress the human protein, silence the endogenous murine gene, or disable the normal chaperoning mechanisms, i.e., knock out the gene for hsf1. We have assumed that the mechanisms governing host defenses against protein misfolding will be the same in the case of overexpression of a relatively stable protein in which only a small proportion misfolds as exists in mice with lesser levels of expression of a highly thermodynamically unstable protein where most of the protein dissociates to yield the aggregation prone precursor. This may not be true and requires further investigation in models in which the murine TTR gene has been silenced and lower levels of the human transgene are expressed.

An additional reservation regarding the data is that, in truth, we cannot discern with certainty whether the changes we observed are responsive or enabling. While we can draw some conclusions from the nature of the changes in the young and old mice, a definitive conclusion awaits more detailed temporal studies, particularly of the very early events.

The hepatic transcription patterns of young transgenic animals did not differ very much from those of nontransgenic mice. There were significant differences in mitochondrial gene transcripts and those encoding components of the secretory apparatus, perhaps reflecting greater energy utilization and secretory activity in response to overexpression of the TTR transgene.

Most striking was the observation that relative to livers of old transgenics with cardiac deposition, livers of 2-yr-old animals that did not display cardiac deposition showed active transcription of genes encoding chaperones and ERAD elements (3). In these livers, Xbp1u transcription increased, but the amount of spliced Xbp1 mRNA was also increased, suggesting an active UPR. Although the difference between the livers of animals with and without cardiac deposition only showed a statistical trend (P=0.15), the difference between young transgenic mice and those without deposition was significant, perhaps suggesting a diminution with age but with a residual functional level capable of driving the UPR. Transcription of the human TTR transgene and the endogenous mouse TTR gene was lower in the livers of animals without cardiac deposition, a phenomenon presumably related to rapid mRNA degradation that has been proposed as being part of proteostatic response but perhaps independent of the UPR (19). This may also account for the lower mean serum TTR concentrations in the mice without cardiac deposition. However, more specific evidence for transcriptional activation of the UPR, such as increased eif2α, CHOP, ATF4, or ATF6 transcripts, was lacking, perhaps reflecting the chronic nature and mixed-cell populations being studied in our model. Because most of the features of the UPR have been defined in acutely perturbed homogeneous cell populations in tissue culture, it is possible that analysis of in vivo steady state transcription patterns is not sufficiently sensitive to detect the full panoply of UPR-associated changes. However, the transcriptional data are consistent with tissue culture studies in which mutant TTR genes have been overexpressed (4, 20).

Transcripts increased in the livers from animals with cardiac deposition with respect to livers from age-matched animals without cardiac deposition, including a set of interferon-induced inflammatory genes and genes involved in glutathione metabolism, suggesting the presence of oxidative hepatic damage in those mice (although these increases were not statistically significant). There were no increases in transcription of chaperones, little evidence of proteasome activity, and interestingly, a modest increase in the transcription of Tribbles 3, a gene suggested to be a feedback inhibitor of the ATF4 UPR pathway (21–24). The absence of Xbp1s transcripts in these livers by microarray analysis (although present at reduced levels by qPCR analysis) suggests that at this time in these animals, the UPR was not highly induced, was ineffective, or primarily activated the ATF4 apoptotic pathway, and we were sampling the surviving cells without actually observing apoptosis.

The observations suggest that two major physiological processes are involved in or reflect the deposition of the amyloidogenic precursor TTR in the heart. Well before there is any morphological or chemical evidence for tissue deposition of TTR, i.e., at 3 mo of age, the hearts of the transgenic animals showed increased transcription of a large number of genes that are associated with inflammation, despite the absence of morphological evidence of inflammatory infiltrates. The source of the transcripts may be circulating inflammatory cells that are present in numbers that we did not perceive by microscopy as being increased relative to nontransgenic animals, a population of inflammatory cells resident in the heart or the transcription of genes associated with inflammation by cells not usually thought to be involved in the inflammatory response (i.e., the cardiomyocytes). Studies of isolated cardiomyocytes in tissue culture have suggested cell autonomous expression of similar inflammatory transcripts in response to exposure to amyloidogenic TTR (N. Reixach, Scripps Research Institute; unpublished results). It is likely that this represents a tissue-specific rather than a generalized systemic inflammatory response, since the transcriptional profile differs from that of the liver as well as that seen in the kidneys (not shown). The characteristics of the transcripts suggested a predominantly nonspecific response by cells of the monocyte/macrophage and T-cell lineages. However, we also observed significant increases in Rag 1, Tcrg-V4, Tcrb-V8.2, Igk (c-region), and H-chain constant region transcripts for gamma and alpha chains, suggesting a polyclonal B- and T-cell immune response. Small amounts of anti-TTR antibodies have been seen in some patients with TTR mutations, as well as normal controls (Y. Sekijima, Division of Clinical and Molecular Genetics, Shinshu University School of Medicine, Matsumoto, Japan; unpublished results). Thus, it is possible that this represents a clearing mechanism for small amounts of aggregated protein.

In contrast, the hearts of 2-yr-old transgenic mice with histologically apparent TTR deposition when compared with age and sex-matched animals without cardiac deposition show a transcriptional pattern associated with an oxidative stress response, largely driven by Nrf2 (ref. 25 and Fig. 3C). There is enhanced transcription of genes associated with glutathione metabolism, a large increase in the number of transcripts of mitochondrial genes involved in fatty acid oxidative metabolism, but most striking is an increase in the number of transcripts encoding both cytoplasmic and ER chaperones and genes involved in proteasome function and ubiquitination. We found no evidence of cardiac TTR transcription, confirming that in the cells of these hearts, the stress is initiated from outside rather than from within the cell. These findings are consistent with in vivo data from tissues of FAP patients that show evidence of ER stress (26). Cardiac transcription of Xbp1, an early activator of the UPR, showed increased transcription, but quantitative PCR did not demonstrate an increase in the spliced (UPR-active) form of the gene. The role of unspliced Xbp1 is not clear. It has been suggested that it exerts negative feedback on initiation of the UPR via interaction with Xbp1s or inhibits ATF6 activation (27, 28).

We also found increases in genes involved in apoptosis, suggesting that cell death was occurring in the hearts of the mice with cardiac TTR deposition. Inflammatory gene transcription was increased in the hearts with deposition, but the number of increased inflammatory transcripts was lower than in the younger animals without deposition and had some suggestion of specificity, i.e., transcripts encoding Ig H (α and γ), κL, and J chains, all suggesting the presence of mature B cells or plasma cells in the heart, which may have been responding to some conformation displayed by the misfolded or aggregated TTR. Because we observed a greater proportion and a somewhat different distribution of transcribed immune/inflammatory genes in the hearts of animals before there was any evidence of TTR deposition, it is tempting to view the early response as protective and the transcription pattern in the presence of deposition as a failure of that mechanism.

In addition to the transcripts that appeared to represent a direct effect of the cardiac deposits, we also observed patterns of transcription, suggesting physiological adaptation to an insult that may compromise cardiac function. We noted increases in groups of transcripts responsible for proteins involved in mitochondrial activity and β-oxidation of fatty acids. Some of these are regulated by PPARα, which is known to influence genes maintaining fatty acid oxidation and mitochondrial biogenesis (29). Congestive heart failure may be associated with a shift in metabolism from fatty acid oxidation as an energy source to more efficient generation of energy via glucose consumption (30). It is likely that these hearts were not sufficiently functionally compromised to produce that metabolic change. There were also increases in the transcription of a group of genes involved in the cardiac response to hypoxia and may reflect incomplete compensation to a relatively hypoxic state secondary to the tissue TTR deposits (31).

We interpret these results to indicate that despite the relatively late onset of tissue TTR deposition in these animals, molecules with fibrillogenic potential are present early in life and presumably circulate and are recognized by cells in the heart as a potential target organ. We have not yet identified the nature of such circulating molecules. The early cardiac transcriptional response is primarily inflammatory. In the older animals with deposition, the relative number of immune/inflammatory genes showing increases is diminished when compared with the hearts of the younger transgenic mice. The hepatic response appears to involve UPR, chaperone, and proteasome gene transcription. With time, perhaps related to oxidative damage secondary to long-term exposure to high concentrations of potentially (or actually) misfolded TTR species, as evidenced by the enhanced transcription of redox-active enzymes and interferon-stimulated inflammatory genes, hepatic transcription of chaperones and proteasome components are diminished. It is possible that under these conditions, larger amounts of aggregation-prone precursor may be released into the circulation, even though the circulating TTR concentration is reduced. The increased amyloid precursor load overcomes the apparent inflammatory response in the heart, deposits extracellularly, and induces a stress response, as well as a metabolic compensatory state. Hence, it appears that the chaperone activity of the liver determines the presence or absence of extracellular cardiac TTR deposition, i.e., the critical chaperoning function takes place at a distance from the site of deposition. The sequence is depicted in the model shown in Fig. 4.

Figure 4.

Transcriptional responses in an organ synthesizing TTR and an organ in which there is TTR deposition. Does this represent chaperoning at a distance? At 3 mo of age, there are no TTR deposits in the heart, but transcription patterns in both liver and heart differ from nontransgenic controls. At 2 yr of age, mice without patterns of hepatic transcription of elements of the proteostatic network show cardiac deposits and transcription patterns reflecting oxidative stress.

In contrast to the peripheral nerves showing TTR deposits in Hsf1-knockout TTR transgenic mice, the hearts of these animals did not show evidence of RAGE or NF-κB expression (17). In human patients with FAP, limited studies of peripheral nerve and labial gland biopsies have shown increases in the transcription of biglycan and HSP 27 (32, 33). Although we observed increases in small heat-shock proteins and decorin in hearts with deposits, we did not note increased biglycan transcription except in young transgenic animals. These differences may be a function of species or tissue or reflect the more limited nature of the analysis in the human tissues compared with what we were able to see in a systematic examination of TTR transgenic mice with and without tissue deposition.

Are these findings clinically relevant? Approximately 10% of individuals who, because of the universal shortage of liver donors, have received livers from patients with FAP (usually with TTR V30M mutations) in the course of “domino transplantation” treatment for liver failure, have developed peripheral TTR amyloidosis within 8 yr of transplantation. The fibrils bear the mutation of the liver donor (34–36). A more recent systematic study indicated that at 5 yr post-transplant, 35% of domino liver recipients had TTR deposits in rectal or sural nerve biopsies, with no differences in the age or sex distribution between those with and without deposition (37). This is a much shorter period than the 30 yr it takes for carriers of the FAP mutation to develop TTR amyloid deposition. The current study suggests that livers from some patients with FAP may have reduced capacity to chaperone the misfolded precursor of the mutant TTR being synthesized by the transplanted liver. The defect puts the recipient at risk for systemic TTR amyloidosis. At present, there is no clinical test of liver function that measures chaperone or protein homeostatic capacity. While this is a relatively rare circumstance, the murine model suggesting chaperoning at a distance may provide a possible explanation.