Nanomechanics of new materials – AFM and computer modelling studies of

Trichoptera silk

(short title: Nanomechanics of Trichoptera silk)

J. Strzelecki a)

, J. Strzelecka b)

, K. Mikulska a)

, M. Tszydel c)

, A. Balter a)

, W. Nowak a)

a)

Institue of Physics, Nicolaus Copernicus University, Grudziądzka 5, 87-100 Toruń, Poland,

jast@fizyka.umk.pl; b)

Medical Physics Department, Oncology Center, dr I. Romanowskiej 2, 85-796 Bydgoszcz, Poland c)

Department of Ecology and Vertebrate Zoology, University of Łódź, Banacha 12/16, 90-237 Łódź, Poland

1. Introduction

Living systems provide excellent inspiration for design of engineering constructions,

pharmaceutical formulae, and new materials [1]. Silks, due to their extraordinary mechanical

properties, such as high tensile strength and extensibility, are a particularly interesting target

for biomimetics. Both natural Bombyx mori (BM) silkworm and synthetic silks are widely

used in textile industry. Very promising applications of silk, due to its biocompatibility are

biomaterials, like scaffolds in regenerative medicine [2]. Production of nanofibers from silk

attracts attention of engineers [3]. Silk fibers, produced in spinning glands and spinnerets, are

composed mainly of long proteins. They are produced not only by silkworms but also by

spiders and numerous other species [4] and perform different functions in the world of

invertebrates. Since the amino acid composition of silk differs among various species, new

properties of such “exotic” fibers may be expected. In this paper we present the first

nanomechanical studies of natural fibers produced by caddisfly (Insecta, Trichoptera).

Trichoptera are closely related to the order Lepidoptera and the two orders together comprise

the superorder Amphiesmenoptera. Trichoptera larvae and pupae, in contrast to spiders and

silkworms, live predominantly underwater and use silk fibers to build pupation cases and

tubular case shelters or nets for capturing food [5, 6]. To perform those tasks, the caddisfly

silk has special properties: it has to be highly insoluble in water and allow sticking to wet

surfaces. In Trichoptera the „weaving‟ is done by the elements of the mouthpart and not of the

abdomen, as in the case of spiders, which makes it much more precise. Variety of forms and

versatility of silken products is comparable in Trichoptera and Araneinae. Thus, Trichoptera

silk can provide material for fibers excelling in underwater tasks, and in particular, water

resistant biocompatible glue. Even though caddisflies, in opposition to spiders are not

cannibals and are easier to rear, their fibers are not long enough for industrial applications.

Thus, a better understanding of composition and structure of Trichoptera silk is necessary to

synthesize it with bioengineering and electrospinning techniques, as it was recently done for a

spider [7].

With optical and electron microscopy it was possible to determine, that similarly to silkworm

silk, individual fibers consist of two filaments, giving them a ribbon shape (Fig. 1). This

construction is spun by twin silk glands. Individual fibers of caddisfly silk can be a couple of

centimeters long and 0.5-7 micrometers wide, with a rather uniform height of 0.8-1.0

micrometers. Ribbons are composed of bundles of 15–25 Å filaments of H-fibroin and L-

fibroin [8, 9]. Additionally, both ribbons are embedded in the peripheral layer rich in

oligosaccharides [8]. The amino acid sequence of the most abundant proteins – fibroins –

present in Trichoptera silk is known – PDB accession number of Fib-H: A5A6G5, Fib-L:

A5A6G6.

It is expected that the detailed composition of these proteins, together with the proteins fold,

determines mechanical properties of the silk. The bulk strength of individual fibers and

protein glands were determined using a tensile tester [10-12]. The obtained data indicate, that

Trichoptera silk has a tensile strength similar to spider or BM silk, with a strong dependence

on humidity. However, the nanomechanical properties of Trichoptera silk have not been

studied yet. Here we report the first Atomic Force Microscope (AFM) imaging of single

molecules from fibers produced by Hydropsyche angustipennis (HA) Trichoptera. The goal of

this study is characterization of the surface of Trichoptera silk, i.e. morphology on the

nanoscale, and checking to what extent HA fibers are distinct from the spider fibers, whose

surface was already imaged with an AFM [13, 14]. The AFM force spectra of individual

molecules from HA fiber surface are also presented and discussed for the first time.

Analogous measurements were also done for spider silk from Tegenaria domestica (TD) as a

reference.

The interpretation of AFM spectra is much easier, if data on simulated stretching of

biopolymers are available [15, 16]. The increase of the contour length during mechanical

protein unfolding can be successfully linked with the length of extended domains calculated

from simulations. Thus, the simulations provide a necessary tool for proper interpretation of

force spectra of individual biopolymers. Therefore, we performed modelling of important,

short components of Fib-H HA species molecules. Selected fragments of this protein were

stretched using the all-atom steered molecular dynamics approach. Interestingly, we have

found that the mechanical strength of this major protein of Trichoptera silk critically depends

on the presence of calcium ions in the fiber. Our study may help to develop a new water

resistant glue or new synthetic fibers useful for sutures or tissue engineering [17].

The composition of the paper is as follows: In the first part experimental technique and

computational protocols are described. In the second part we present results of AFM studies

of HA fibers and results of computational stretching of single molecules of Fib-H fragments.

In the same part an analysis of nanomechanics of HA silk is presented. Conclusions

summarize the paper.

1.1 Experimental Techniques

Silk net collected from laboratory bred HA Trichoptera was untangled with sharp needle to

isolate single fibers. The fiber was glued in elongated form to a glass substrate with nail

polish. TD spider silk was prepared in similar manner. An AFM contact mode imaging of dry

fibers was performed with Bioscope 2 Veeco AFM, using Veeco MSNL-D probes. AFM fast

scan axis was set along the fiber length. Deflection error and lateral force images were

flattened by subtraction of first-order polynomial plane fit.

Force spectroscopy measurements were performed in deionized water using a homemade

AFM one dimensional puller, inspired by [18]. A Veeco MSNL-C cantilever was calibrated

using the thermal tune method [19], placed over fiber using an inverted optical microscope

and lowered on its surface with a micrometer screw. After touching the surface, a piezostack

was used to move the cantilever away and towards the sample within the limit of 15

micrometers. Single molecules were picked by the untreated cantilever tip and stretched at

400 nm/s with a 2-4 nN loading force.

1.2 Computer Modelling Protocols

The main protein present in HA nanofibrils is fibroin. Its 3D structure is not known yet. In

order to investigate the effects of phosphorylation and the presence of Ca2+

ions in the silk we

have constructed 8 models of short, frequently occurring modules of a heavy form of

Trichoptera HA fibroin (Fib-H, accession number: A5A6G5). Particular attention was

devoted to serine-rich motives SXSXSX which may strongly modulate mechanical properties

of silk (see Fig. 2).

The Molecular Builder implemented in the Discovery Studio suite of codes (Accelrys Inc.,

2008) was used. Two spatial models of the Fib-H fragment were created. For the first model

polypeptide (M_b), 31 amino acids long, we had assumed secondary structure in a form of

two antiparallel beta strands. In the second model, having the same sequence, a form of two

dextrorotatory alpha-helices was applied (M_a).

All Steered Molecular Dynamics simulations (SMD), were performed using the software

package NAMD 2.6 [20] with the all-atom CHARMM27 force field [21]. In the SMD method

an external virtual force is applied to investigate the elasticity of a long molecule. This

computational technique has been often used in an interpretation of the AFM single molecule

experiments [16, 22, 23]. The Visual Molecular Dynamics (VMD) program (version 1.8.7)

[24] and home-made scripts were used to prepare input files and analyze output trajectories.

There are numerous models of water available. Their practical utility depends on the question

asked [25]. We have selected the TIP3P model, very popular in CHARMM simulation of

proteins, in order to ensure compatibility with our earlier results [22, 26]. All protein models

were solvated using a 0.8 nm layer of TIP3P waters. Cutoffs of 12 Å for non-bonded

interactions were applied, with a switching distance of 8 Å and pair list distance of 14 Å. The

Langevin dynamics and the Langevin piston algorithm were used to maintain the temperature

at 300 K and the pressure of 1 atm. The multiple time step method was employed, with time

steps of 1 fs for bonded, 2 fs for short-range non-bonded and 4 fs for long-range electrostatic

forces. We performed 0.2 ns of water equilibration, 10000 steps of minimization, 0.35 ns of

heating from the 0 K up to 300 K and 0.65 ns equilibration at 300 K of the whole system

before starting each main SMD simulation. A constant velocity SMD method was used to

stretch fragments of Fib-H. In this technique a virtual harmonic force is applied to one end of

the protein (here the Cα atom of the C-terminus) having simultaneously the other end fixed

(Cα of N-terminus). The vector of the virtual stretching force was aligned with the line

connecting the Cα atoms of the N- and C-terminal residues. It was calculated after water

equilibration. The force probed by a virtual spring was calculated using the formula F = k(vt −

(x0 − x)), where k is the spring constant, v is the velocity, t is time, and x and x0 are the

positions of the pulled atom and the spring‟s opposite end, respectively. Structures were

stretched at a constant speed of 0.025 Å/ps with a “virtual” spring constant of 4 kcal/mol/1Å2

(278 pN/Å).

We have run twenty SMD simulations of Fib-H fragment for each models M_a and M_b

(total simulation time 137.5 ns). For each model we have made four variants where the types

of ions present in the solution (150 mM NaCl or 8 ions of Ca2+

) and/or the phosphorylation of

serines (Ph) were changed. Initial structures are presented in Fig. 3.

2. Experimental Results and Discussion

Fig.4 shows topography, deflection error and friction images obtained by AFM imaging of the

silk surface. In agreement with previous optical microscope and SEM observations, the fiber

consists of two joined filaments, with a distinct groove along their length. However, less

ordered features, in particular dense 150 nm bumps could also be seen. Such structures were

not observed in SEM imaging. Similar surface pattern, albeit less dense and showing fibril

structure was however observed in spider silk AFM images [14]. Large features indicated by

arrows on the friction image (c) have a friction response different than the surrounding area.

Most probably these are foreign matter debris attached to the silk surface. Repetitive 150 nm

bumps seem to be, on the other hand, structural features of silk surface.

As this pattern shows increased friction (see Fig. 4 (c)), we hypothesize that it is created

during silk spinning to provide surface optimized for better attachment. Trichoptera spin layer

upon layer of silk to create the pupa case. The resulting weave sticks to very diverse material

used to reinforce the case. Additionally, silk fibers are joined together to form a dense mesh in

webs. Finally, the web has to hold caught prey. In all these situations nanometer scale

roughness of silk provides better sticking surface.

AFM force spectroscopy (Fig. 5) of spider and caddisfly silk exhibits a large variety of

recorded force curves. For small extensions step like curves could be observed for both silks

as seen in Fig. 5 (a, b). Such curves are typically associated with a rapid detachment and

unpeeling of strands from substrate surface, as was reported in case of amyloidal fibers [27]

and cellular tethers [28]. At higher extensions of 4 micrometers some structures were

stretched at forces of 200 pN as seen in Fig. 5 (c). Data obtained from spider silk not only

shows similarity, but can be directly superimposed as in Fig 5 (d). Relaxation curve follows

the same trace as stretching and no hysteresis can be seen. However, long extension curves in

Fig 5(e) that show refolding with significant hysteresis were seen in caddisfly silk. While we

observed such curves only in caddisfly silk, they bear significant resemblance to curves

reported for catch silk of orb weaving spiders [29]. The existence of hysteresis indicates

successful energy dissipation during stretching. It is a mechanism useful both for the web

resisting water current and holding the prey caught in it, as well as in maintaining the integrity

of the case protecting the insect during pupation. At last, a saw-like patterned class of force

curves was observed solely in caddisfly silk at small (600 nm) extensions (f). Those resemble

a pattern seen when unfolding globular proteins [30] or amyloidal beta structures [27] and

cloned nanofiber proteins from spiderwebs [31]. Such a pattern was also observed in

secretions of microorganisms that stick to underwater surfaces [32, 33].

3. Results of Calculations

SMD computed force-extension curves are presented in Fig. 6 (for M_b) and Fig. 7 (M_a).

The mechanical strength of each studied fragment clearly depends on the simulation

condition. In a normal saline solution, the unphosphorylated peptide may be fully stretched

under a force of 500 pN, irrespective of the assumed secondary structure of the model. Using

Ca2+

ions in the medium instead of NaCl only slightly affects unfolding forces: they rise by

some 100-200 pN (see Figs. 6 and 7). There are some linking hydrogen bonds present in Ca2+

solution and absent in the NaCl solution – this effect is weak, but noticeable.

The phosphorylation of all serines (Ph) in the model has also some impact on the mechanical

resistance of the fragment studied. In M_b the effect of phosphorylation is stronger than in

M_a (see Figs. 6 and 7). The most spectacular observation is a dramatic increase of maximum

unfolding force for models with phosphorylated serines and with the presence of Ca2+

ions in

the medium. The maximum forces in M_a are about 2500 pN and for M_b are even higher, at

the level of 3000 pN. The values of all maximal forces observed in computer stretching

experiments for Ph M_b and M_a, in the presence of Ca2+

, are collected in Tab. 1 and Tab. 2.

Table 1. Maxima of unfolding forces for M_b (model D) in five distinct simulations.

Fmax sim1 sim2 sim3 sim4 sim5

t/ns F/pN x/nm t/ns F/pN x/nm t/ns F/pN x/nm t/ns F/pN x/nm t/ns F/pN x/nm

max1 1.47 1417 2.42 1.63 2123 2.5 1.65 2322 2.49 1.54 1689 2.45 1.44 1421 2.3

max2 1.91 2446 3.13 - - - 1.89 1867 3.25 1.83 1703 3.18 1.94 2158 3.3

max3 2.12 2316 3.67 2.13 2592 3.58 2.24 2396 3.94 2.27 2342 4.03 2.31 2867 3.94

max4 2.86 1846 5.69 3.22 2871 6.21 3.15 2599 6.1 3.14 2254 6.23 3.09 2640 5.98

max5 3.43 2148 7.03 3.58 3084 7.07 3.34 2194 6.75 3.44 2585 6.87 3.6 3120 7.05

Table 2. Maxima of unfolding forces for M_a (model D) in five distinct simulations.

Fmax sim1 sim2 sim3 sim4 sim5

t/ns F/pN x/nm t/ns F/pN x/nm t/ns F/pN x/nm t/ns F/pN x/nm t/ns F/pN x/nm

max1 0.39 725 1.04 0.42 793 1.09 0.42 788 1.07 0.38 695 1.06 0.42 729 1.06

max2 0.94 1913 1.94 1.04 2741 1.88 0.94 1944 1.92 1.023 2717 1.82 1.04 2845 1.86

max3 1.36 931 3.35 1.56 1130 3.79 1.29 784 3.2 1.8 2288 3.94 1.39 1142 3.36

max4 2.2 2424 4.87 2.15 2087 4.89 2.19 2611 4.78 2.24 2710 4.92 2.08 1820 4.79

The SMD simulations reveal the nature of observed maxima in force-extension spectra. The

sequence of events related to these maxima, for the M_b (Ca2+

, Ph) unfolding process, is

shown in Fig. 8. The Ca2+

ions bind β strands together by strong interactions with negatively

charged phosphorylated serines. At 1.6 ns the structure of the model is substantially stretched

(Fig. 8 a-e) and the force increases drastically. Thus, the interactions between the first two

phosphorylated serines and Ca2+

break (see Fig. 8 e-f) and the force drops down. That‟s why

we observe the first maximum (Fig. 6). The next three maxima have a similar origin: they are

also connected with the loss of interactions between some of phosphorylated serines and Ca2+

(see Fig. 8 f-g, h-i, i-j). The number of hydrogen bonds between β strands (and α helices as

well) changes during simulations (data not shown) but it doesn‟t correlate with maxima in

force spectra.

A similar effect of Ca2+

and Ph is observed also for the α-helical model M_a (see Fig. 7), but

the nanomechanical resistance of such an architecture is 30 % smaller than that for β strands.

Thus, the mechanical strength of Fib-H molecules, and probably the whole HA Trichoptera

nanofibers as well, critically depends on the presence of phosphorylated serines in Fib-H and

on the sufficient supply of Ca2+

ions. Due to our modelling, the molecular origin of this

resistance is clear: numerous hydrogen bonds and electrostatic interactions between negative -

PO42-

groups and divalent Ca2+

cations stabilize SXSXSX fragments very strongly.

A similar mechanism has been already discussed in the literature [34]. One should note that in

a real fiber probably not all serines will be phosphorylated. The concentration of calcium,

though found to be very high (between 55 to 123 nmol) in Trichoptera silk [34], may be not

high enough to induce such strong effect as observed here.

The numerical values of maximum forces obtained from the SMD modelling don‟t necessary

correspond to the maxima observed in AFM experiments [22] due to the mismatch in pulling

rates and limitations of force fields used in the simulations. However, simulation results can

still help to explain force spectroscopy results. As the simulation shows, a hidden length of

approximately 10nm is set free if all bonds break in M_a or M_b model structures, resulting

in a decrease of force (Figs. 6 and 7). Since during the AFM stretching many such structures

unfold, we should expect a rapid decrease of force and increase of the contour length in

multiplies of 10 nm. Such situation can be observed in the force curve presented in Fig. 5f.

Force curves of this type show a peak to peak length of 80 nm within the 40-250 nm range.

The experimental fold of H-Fib is not known yet, but based on the recent paper [35] one may

expect that a hidden length in the whole H-Fib structure backbone (Fig. 2) is approximately

230 nm. Thus, we hypothesize, that saw-like force curves result from breaking many beta (or,

to a lesser extent, α-helical) structures localized in parallel stacks. A similar situation was

already observed during breaking of many beta turns in amyloid fibers [36]. However, it is

necessary to note that due to a random nature of AFM molecule pick-up process the

attachment points are scattered and this estimate is approximate.

The present modelling data support the hypothesis that both phosphorylation and

availability of Ca2+

ions contribute substantially to mechanical and elastic properties of silk.

This is particularly true if proteins are rich in amino acids prone to phosphorylation. Thus,

based on this observation, the composition of artificial, biocompatible nanomaterials may be

planned in a better way.

4. Conclusions

Our AFM images revealed for the first time a characteristic structure of HA silk fibers

surface. We hypothesize, that this pattern is created during spinning to optimize silk floss

sticking. AFM force spectroscopy at the fiber surface shows a diverse mechanical behavior.

The mechanical strength of individual molecules, probably fibroins, of HA is high (200 pN).

Force spectra indicate the existence of a sacrificial bonds mechanism that can serve as a

mechanical buffer facilitating the dissipation of energy. Both the surface structure and the

mechanical behaviour indicate, that HA Trichoptera silk shows similarities to TB spiderweb.

Those similarities suggest that the same proteins can be present, even though both of those

organisms represent distinct orders. With a denser silk surface pattern caddisfly silk is

however better adapted to sticking and binding diverse materials. With our SMD simulation

we have also found that mechanical strength of nanofibers critically depends on the presence

of calcium ions in the silk. Ions act as a kind of glue that increases the nanomechanical

strength of model systems by a factor of 5. Independent studies show [34] that the content of

calcium in HA fibers is very high indeed. Therefore these ions should be considered as an

important component of future protein based nanomaterials.

5. Acknowledgments

Janusz W. Strzelecki and Joanna Strzelecka acknowledge “Krok w Przyszlosc” Edition I

(JWS, JS), II (JS) & III (JWS) from Kujawsko-Pomorskie Voivodeship and EU.

Authors thank Dr. Julita Templin at Department of Invertebrate Zoology, Institute of General

and Molecular Biology, Nicolaus Copernicus University for a generous gift of TE spiderweb.

[1] C. Stevens, Industrial Applications of Natural Fibres: Structure, Properties and Technical Applications

(Wiley, 2010)

[2] N. Kasoju, R. Bhonde, U. Bora, J. Tissue Eng. Regen. Med. 539, 3 (2009)

[3] H. Zhao, X. Feng, H. Gao, Applied Physics Letters 073112, 90 (2007)

[4] C. Craig, Spiderwebs and silk: tracing evolution from molecules to genes to phenotypes (Oxford

University Press, USA, 2003)

[5] V. Nepomnyashchikh. (1992), vol. 265,

[6] J. Edington, A. Hildrew, A revised key to the caseless caddis larvae of the British Isles with notes on their ecology (Publication, 1995)

[7] M. Heim, D. Keerl, T. Scheibel, Ang. Chem. Int. Ed. 3584, 48 (2009)

[8] M. Engster, Cell Tissue Res. 77, 169 (1976)

[9] N. Yonemura et al., Biomacromolecules 3370, 7 (2006)

[10] M. Tsukada, M. Khan, Int. J. Biol. Macromol., (2009)

[11] J. Bilska, I. Krucinska, FIBRES & TEXTILES in Eastern Europe 54, 13 (2005)

[12] S. Brown, G. Ruxton, S. Humphries, J. North Am. Ben. Soc. 771, 23 (2004)

[13] A. Schäfer et al., Eur. Biophys. J. 197, 37 (2008)

[14] S. Gould et al., Int. J. Biol. Macromol. 151, 24 (1999)

[15] H. Lu, K. Schulten, Chemical Physics 141, 247 (1999)

[16] W. Nowak, P. Marszalek, Molecular Dynamics Simulations of Single Molecule Atomic Force Microscope J. Leszczynski, Ed. (World Scientific Pub Co Inc, Singapore, 2005)

[17] O. Hakimi et al., Compos. B Eng. 324, 38 (2007)

[18] F. Kühner et al., Rev. Sci. Instrum. 75105, 78 (2007)

[19] J. Hutter, J. Bechhoefer, Rev. Sci. Instrum. 1868, 64 (1993)

[20] J. Phillips et al., J. Comput. Chem. 1781, 26 (2005)

[21] A. MacKerell Jr et al., J. Phys. Chem. B 3586, 102 (1998)

[22] G. Lee et al., Biophys. J. 1456, 87 (2004)

[23] J. Strzelecki et al., Acta Phys. Pol. A S156, 116 (2009)

[24] W. Humphrey, A. Dalke, K. Schulten, J. Mol. Graph. 33, 14 (1996)

[25] W. Jorgensen et al., J. Chem. Phys. 926, 79 (1983)

[26] Z. Lu et al., J. Am. Chem. Soc 9033, 126 (2004)

[27] M. Kellermayer et al., J. Biol. Chem. 8464, 280 (2005)

[28] M. Benoit et al., Nat. Cell Biol. 313, 2 (2000)

[29] N. Becker et al., Nature Mater. 278, 2 (2003)

[30] M. Rief et al., Science 1109, 276 (1997)

[31] E. Oroudjev et al., Proc. Natl. Acad. Sci. U. S. A. 6460, 99 (2002)

[32] T. Dugdale et al., Biophys. J. 4252, 89 (2005) [33] A. Mostaert et al., J. Biol. Phys. 393, 32 (2006)

[34] R. Stewart, C. Wang, Biomacromolecules, (2010)

[35] S. Keten et al., Nature materials 359, 9 (2010)

[36] T. Fukuma, A. Mostaert, S. Jarvis, Tribology Letters 233, 22 (2006)

Figure captions

Fig. 1. Morphology of HA Trichoptera silk as seen with a scanning electron microscope

(courtesy of H.Wrzosek). Double-ribbon structure can be seen.

Fig. 2. A sequence of HA Fib-H chain with motives affecting mechanical properties of

protein [34]. The fragment selected for computer modelling is highlighted.

Fig. 3. Initial structures of model fragments of Fib-H: M_b (A-D) and M_a (E-H).

Fig. 4. Contact mode AFM images of HA silk fibers surface. Topography (a), deflection error

(b) and friction (c) images show a groove where filaments join together as well as an uneven

surface with 150 µm bumps. Arrows in the lateral image denote features with distinctive

friction behaviour.

Fig. 5. AFM force spectroscopy at HA and TB silk fiber surface. A diverse array of force

curves was obtained. A similarities of mechanical behavior can be observed (a,b), with

features identical in both species (d). A hysteresis in stretching/relaxation, providing good

energy dissipation during stretching is also present. Moreover, saw like pattern, reported in

many biological systems with mechanical function was also observed (f). Histogram of peak

to peak distances is shown in the inset.

Fig. 6. Mechanical unfolding of the beta sheet model of Fib-H fragment (model M_b). Black

curve NaCl solution, Red – NaCl+phosphorylated serines, Blue – Ca2+

solution, Green - Ca2+

solution+ phosphorylated serines. Annotated circles indicate maxima related to events

discussed in the text.

Fig. 7. Mechanical unfolding of the helical model of Fib-H fragment (model M_a). Black

curve NaCl solution, Red – NaCl+phosphorylated serines, Blue – Ca2+

solution, Green - Ca2+

solution+phosphorylated serines.

Fig. 8 The sequence of events in the mechanical unfolding path of M_b (Ph, Ca2+

) for sim2.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8