Comparative Vibrational Spectroscopy of Intracellular Tau and Extracellular Collagen I Reveals Parallels of Gelation and Fibrillar Structure

The N-terminal tau (2-19) peptide undergoes gelation, syneresis and aggregation over a period of years. These changes may be approximated on a shorter time scale by agitation and partial dehydration. The anomalously enhanced (229 nm.) ultraviolet resonance Raman (UVRR) imide II band reveals a common structural feature for gels of non-dehydrated tau (2-19) and collagen I, and insoluble paired helical filaments (PHFs) and collagen I, of weak hydrogen bonding at proline carbonyls. Anomalous UVRR enhancement of amide bands at 229 nm. results from gel structure, as demonstrated by increased amide absorption at the red edge for tau (2-19) gel, and implies the involvement of water in gel structure. In aged, dehydrated tau (2-19) gel, proline carbonyls becomes nonbonded and tyrosine becomes deprotonated, as demonstrated by UVRR spectroscopy. The Fourier transform infrared (FTIR) amide I band shows that antiparallel β sheet structure increases with syneresis in the tau (2-19) hydrogel. The comparison of FTIR results for PHFs with collagen I gel and polyproline demonstrates that the secondary structure of PHFs is polyproline II. One implication of this assigment is that hydrophilic tau fibrillization is thermodynamically driven by the entropy gained as hydrogen-bonded water is freed, as for collagen I. The FTIR results also show that peptide domains culled from a longer protein do not necessarily fold into identical secondary structures. A pathological, sequential mechanism of gelation, syneresis and fibrillation for tau in AD is suggested, and supported by the observation of amorphous tau plaque development and fibrillation in vivo .


Introduction
The cytomatrix is a living, insoluble, thixotropic gel (1) that is dynamic because its structure may change from fluid (sol) to elastic (gel) and back again; the filamentous network is constantly changing (2). Its constituent, organized structures are polymeric microtubules, actin microfilaments and intermediate filaments (2). The assembly of this gel network and the composite fibrils is assisted through interactions with a number of other proteins. Within the set of microtubule assembling proteins, tau is of particular interest.
In healthy neurons, tau, through its control of the reversible polymerization of microtubules, participates in the extension of both axons and neuritic processes, also known as neural growth cone development, and in the development of neural polarity (3). In a posttranslationally modified state, however, tau is responsible for several neurodegenerative diseases or tauopathies, one of which is Alzheimer's Disease (AD), characterized by intraneuronal inclusions of phosphorylated tau that are initially amorphous. Over time, the intraneuronal inclusions of tau isoforms (4-6) self-assemble into paired helical filaments (PHF's) (7,8); these are responsible for the neurofibrillary tangles found in AD.
Preparation of Collagen Samples. Insoluble bovine collagen type I was suspended in Tris buffered saline at a concentration of 25 mg/ml. A bovine skin gelatin gel was prepared by dissolution in deionized water and in D 2 O at a 50 mg/ml concentration at ca. 60°C with stirring.
The shoulder at 1218 cm -1 is an unassigned amide mode that has also been found in the UVRR spectrum (229 nm.) of neat N-methylacetamide at -30°C (Juszczak, unpublished data).
The difference spectrum shown in Fig. 1, spectrum b clarifies the amide band positions for the five-year old, cured tau gel results in Fig. 1, spectrum a. It also reveals a residual peak at 1600 cm -1 , which is attributed to deprotonation of some unspecified population of tyrosines within the gel rather than to the single phenylalanine, because of the predominance of tyrosinate peaks in Fig. 1, spectra d,e and the presence of a weak Y9a' band at 1168 cm -1 (42,43). The presence of tyrosinate also suggests that the 1579 cm -1 band contains a tyrosinate Y8b' by guest on March 24, 2020 http://www.jbc.org/ Downloaded from Vibrational Spectroscopy of Tau and Collagen I 10 component (42,43). 1 The 1218 cm -1 shoulder of Fig. 1, spectrum a is revealed in the difference spectrum, b of Fig. 1. Amide I bands, which appear at or above a frequency of ca. 1630 cm -1 are not seen (37,44) .
The 1555 cm -1 band is attributed to amide II. The 1583 cm -1 band may also contain a weak amide II component although a three component curve fit (data not shown) to this high frequency cluster of bands (1555-1602 cm -1 ) yielded a reduced χ 2 < 1, suggesting that both amide II and Y8b' bands do not contribute.

UVRR Spectroscopy of PHFs and Insoluble Collagen I. PHFs represent the endpoint in a
transition from molecular disorder to order. Like the N-terminal tau (2-19) gel (Fig. 2b), they are stained by uranyl acetate; water is incorporated in their structure (7). The UVRR result for PHFs, composed of full-length tau isoforms, is given in Figure 7, spectrum b along with result for insoluble bovine Achilles tendon collagen type I (Fig. 7, spectrum a). Both spectra show a tyrosine Y8a band at 1616 cm -1 and a strong broad, imide band at 1466-68 cm -1 . For full length tau isoforms, the proline content is ~10-12% whereas the proline content of collagen type I isoforms is ~25%; thus the area of the imide II band scales accordingly. An amide II band at 1555 cm -1 is strongly enhanced in the PHF spectrum, and weakly enhanced in the collagen type I results. A weak and broad amide III band at 1305 cm -1 is found in the PHF results (Fig. 7, spectrum b). Lastly, a broad, weak amide S band appears at ~1380 cm -1 in both spectra.
FTIR Spectroscopy of PHFs, Collagen I and Polyproline. FTIR results were acquired for two forms of PHFs; these are presented, along with results for other illustrative proteins, in Figure 8.
Results for freshly isolated PHFs in KBr pellet are given in Fig. 8, spectrum a. The peaks, A small peak at 1746 cm -1 has been magnified for comparative purposes, and will be discussed below. The amide I region for three proteins is given in Fig. 8, spectra b. These include amide I results for PHF's recovered from dissolution in LiSCN, with a broad Lorentzian peak (1631 cm -1 ) corresponding to that for freshly isolated PHFs (Fig. 8, spectrum a); a single broad amide I band at 1643 cm -1 for polyproline and a similar Lorentzian-shaped amide I band at 1653 cm -1 for bovine skin gelatin. The 1746 cm -1 peaks for the recovered PHF and polyproline results have been magnified. The high frequency region of the FTIR spectrum for insoluble bovine Achilles tendon collagen is given in Figure 8, spectrum c. Although this sample was similarly embedded in KBr pellet, the amide I band at ca. 1650 cm -1 is not as sharp as that for the PHFs (Fig. 8, spectrum a); the amide I band for collagen is clearly composed of two Lorentzian peaks, determined from curve fitting and centered at 1645 cm -1 and 1686 cm -1 . Another feature of the bovine collagen FTIR spectrum is the strong, sharp band at 1746 cm -1 , which may be attributed to type I β-turns (45) but has also been found in the FTIR amide I band for other proteins with PII secondary structure (46).

Interpretation of the Tau (2-19) FTIR Results.
The amide I band results for the one week old tau (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) gel, non-stirred process (Fig. 5, spectrum a) and for the stirred, dehydrated tau (2-19) gel ( Fig. 5, spectrum b) demonstrate that within the disordered gel network, reorganization of peptide chains to a recognizable secondary structure is occurring. Gel networks are known to be dynamic structures where reorganization processes may occur over time (61). The more sharply defined peaks for the stirred, dehydrated tau gel (Fig. 5, spectrum b) relative to those for the nonstirred tau gel (Fig. 5, spectrum a) suggest that the former gel possesses greater secondary structural organization. The cloudy physical appearance of the stirred, dehydrated tau gel results from a refractive index difference created by an incomplete phase separation of the peptide into discrete islets in a gelation process known as syneresis (61); these islets of concentrated peptide may facilitate the development of secondary structure.
The two most prominent peaks in the amide I band for the stirred, dehydrated tau gel ( and their splitting (52 cm -1 ), resulting from transition dipole coupling of nearest neighbor peptide groups, strongly suggests the inception of an antiparallel β-sheet structure in the gel (45,62). As FTIR peaks below 1620 cm -1 have been noted for β-sheet (63), the strong shoulder in the amide I results for the stirred, dehydrated gel is also ascribed to β-sheet structure. The remaining disordered fraction of peptide in the gel undoubtedly yields an FTIR signal; the broad, unresolved absorbance region between the 1620 cm -1 and 1672 cm -1 peaks is assigned to this fraction.

Changes in the Tau
Comparative Spectroscopic Analysis of PHFs. Collagens are a complex, multi-isoform set of generally nonglobular proteins containing stretches of proline residues which may (e.g., collagen I) or may not (e.g., collagen VI) be arranged in a periodic repeating motif, with N-and Cterminal propeptide sequences having no periodic repeat sequence of amino acids, and containing water-bridged interamide hydrogen bonds (64)(65)(66). Collagen I in the fibril state is a coiled coil of three collagen I isoforms with a polyproline II (PII) secondary structure (64,66). Subsequent to proteolysis by procollagen peptidases, which removes frayed sequences at the termini (67,68), the primary structure of the collagen chains consists overwhelmingly of a three amino acid repeat of the type X-Y-Gly, where X is often proline and Y, hydroxyproline (64,66). The PII structure of collagen isoforms is stabilized by stereochemical interactions and does not require intrachain hydrogen bonds (69). X-ray diffraction results show that the interchain interactions, however, are stabilized by both interamide hydrogen bonds and water-bridged hydrogen bonds (64,66); main chain carbonyls are also hydrogen bonded to water molecules (66). Numerous collagen isoforms exist (65), and although domains of PII structure are often proline-rich, the prolines are not  Tau and Collagen I 21 necessarily arranged in a regular sequence pattern as in collagen I (See, for example, the human collagen VI alpha 1 chain). 2 This, of course, is the case for the proline-rich domain of the tau isoforms. Indeed, Ramachandran has shown that all amino acids can be accommodated in the PII conformation (64). Thus, the comparison between the proline-and water-rich structure of collagen I and the tau PHF is valid, and supported both by the similarity of the UVRR results given in Figure 7 and the correspondence of FTIR peaks in Figure 8.
The one-to-one correspondence in UVRR peaks within the 1350 to 1650 cm -1 frequency window (Fig. 7) suggests a similarity in secondary structure for bovine Achilles tendon collagen type I and PHFs. Solid state PHFs, reconstituted from dissolution in LiSCN (Fig. 8, spectrum b), yield a single Lorentzian amide I band when probed by FTIR spectroscopy, as do solid state polyproline and bovine skin gelatin gel. Clearly, a PII structure is indicated for the reconstituted PHFs. Freshly isolated PHFs in the solid state result in an unambiguous amide I peak at 1631 cm -1 as well (Fig. 8, spectrum a), and a second, broader amide I band at 1665 cm -1 . Again, the 1631 cm -1 peak suggests a PII structure for the virgin PHFs; the second amide band at 1665 cm -1 does not necessarily indicate that a second structural element is present, as discussed below, although the FTIR band frequency is consistent with either random coil or α-helical structure (63). The latter assignment seems unlikely as the amide S band in the UVRR results precludes α-helical structure (Fig. 7, spectrum b).
The amide I band for insoluble bovine Achilles tendon collagen, type I (Fig. 8, spectrum c) is also resolvable into two components, yet x-ray diffraction data shows the molecular structure of collagen to be solely PII (64). In their infrared spectroscopic study of undenatured bovine Achilles tendon collagen, Susi et al. similarly recorded an amide I band resolvable into two peaks, which they attributed to two populations of carbonyl groups within the structural repeat unit (70). The 1631 cm -1 peak of the PHFs may similarly arise from heterogeneity in carbonyl binding.
PII Structure predicted for PHFs. The vibrational spectroscopic evidence presented here strongly supports the assignment of a PII structure to PHFs. Phosphotungstate staining results for the PHFs also supports the assignment of PII structure to PHFs as the fibrils have been shown to consist of subunit domains of greater and lesser stain inclusion, corresponding to fibril domains of greater and lesser hydrophilicity (7). The assignment of polyproline II secondary structure to PHFs is not excluded by Congo red staining since dense collagen fibers, known to possess polyproline II secondary structure, are also stained by Congo red (71).
Factors Effecting Secondary Structure. The FTIR spectroscopic results presented here show that peptide domains culled from a longer protein will not necessarily fold into the same secondary structure, and that the physical process applied to gelling peptides will effect the structure (31).
The apparent discrepancy between reported CD results for various tau constructs (72)(73)(74)(75) suggests that solution phase secondary structure can also depend on one or more factors such as protein concentration (76), time, the construct used and ionic strength (73). Also, CD and other spectroscopic techniques have demonstrated that many proteins classified as 'random coil' in fact possess PII structure (46,69,(77)(78)(79)(80). Indeed, the CD spectrum for the tau construct K12, which is reported as similar to results for full length tau, exhibits a strong, negative peak at ~198 nm. and is negative over the 190-250 nm. spectral window (76). A similar CD result has been reported for collagen I (80). polymer science have not been applied. As for collagen I in the extracellular matrix and actin in the cytoskeletal matrix, spontaneous fibrillogenesis or self-assembly is preceded by a gelation phase (23). Prefibrillar amorphous plaques and the masses of tangled tau fibrils seen in vivo, as well as the formation of a tau film from a dissolved PHF solution reported here, suggest a gel precursor for PHFs. Indeed, a reticulated structure, seen in the tau (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) gel electron micrograph in Fig. 2, spectrum b, is also seen in an image of PHFs incubated at 90°C for five minutes (73) and in that of the cytomatrix (2).

Pathological Tau Gelation within the
The gel state is characterized by processes that are ongoing beyond the initial gelation event. These may include reprecipitation and separation into solid-liquid or solid-solid phases The microtubules and actin microfilaments act as a single integrated lattice with built-in redundancy such that the load-bearing function of the cytomatrix may be maintained when the functioning of either element is compromised (12,19,86). Indeed, microtubule disassembly alone does not prevent dendrite growth as the use of extracellular matrix as a cell adhesion substrate, in the absence of microtubules, is sufficient to induce dendrite elongation in PC12 cells (86).
Apparently, dendrite growth on extracellular matrix results in less internal cellular tension, diminishing the need for the compressive support of microtubules (86). However, cell spreading is totally halted when microtubule and actin microfilament function is simultaneously compromised (19). Since distal, followed by proximal, dendritic processes are found to retract as neurofibrillary tangles appear (11), both microtubule and actin microfilament function must be compromised in AD. This suggests a nonspecific method of action for the tau plaques.
From a purely biophysical perspective, PC12 dendrites and chick dorsal root ganglion neurons have been found to behave as viscoelastic solids under applied forces (87). Generation of low levels of cellular tension and the ensuing retraction response is the reverse of growth, which requires higher levels of cellular tension (87,88). This cellular tension model has been used to explain the complete retraction of severed dendrites in PC12 and chick sensory cells (87).
The gelation-associated processes of syneresis and fibrillation potentially generate cytoplasmic forces that are responsible for the dendritic retraction seen in AD. It is foreseen that the process is set in motion by the permeation of the cytomatrix gel by phosphorylated tau, which subsequently gels to form an enmeshed network of tau gel and cytomatrix gel. The initial effect of an increase in cytoskeletal viscosity may be to retard the sol-gel transitioning of the cytoskeletal network and therefore, to interfere with signalling pathways. Indeed, cognitive impairment in AD has been shown to precede tau fibril formation (9,10). shown that the force applied regulates microtubule polymerization (14). An unmitigated force is expected to lead to massive microtubule depolymerization, denying specialized cell extensions the necessary mechanical support (89). The stress force generated by tau gel contraction and fibrillation may also destabilize the balance of tensile and adhesion forces across the actintransmembrane integrin-extracellular matrix bonds. The expected outcome is dendritic retraction and cell rounding, as cytoskeletal supports of cell shape will have been compromised. The delicate balance of tensile forces between the cytomatrix and the extracellular matrix is thus destroyed by the matching of the collagenic extracellular matrix with an intracellular collagenlike network of tau. This scenario, where cellular differentiation is reversed and cytomatrix dynamics are fouled, is predicted to lead to neuronal death.

Conclusion
This vibrational spectroscopic comparison of tau PHFs with collagen I and polyproline has demonstrated that the secondary structure of PHFs is polyproline II. This structural comparison also leads to an explanation for the paradox of how an hydrophilic protein such as tau could spontaneously form fibrils: as for collagen I, fibrillation may be thermodynamically driven by the entropy gained as hydrogen-bonded water is freed. The N-terminal tau (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) peptide has been shown to form a hydrogel with distinct spectroscopic markers when aged for hydrogen bonding at the carbonyl of proline residues. Several additional threads of evidence point to a pathological, sequential mechanism of gelation, syneresis and fibrillation for tau in AD. It is hypothesized that phosphorylated tau intercalates into the cytomatrix and enmeshes it by formation of its own gel structure. The resulting tau matrix, beyond the control of the cell, is thought to interfere with the dynamic rearrangement of the actin microfilaments, and related signalling pathways. Subsequent syneresis and fibrillation of the tau gel produces intracellular stress forces, overwhelming the supportive cell scaffolding elements of microtubules and actin microfilaments, along with their integrin-mediated bonds to extracellular collagen I. The retraction of neuronal dendrites, as found in preaptotic neurons, is predicted.