1. Class 13 Two sequencing methods that aim to sequence
single DNA molecules
Pacific Biosciences “zero mode” wave guide
Bayley group nanopore method
What steps in sequencing methods we considered so far
would ability to sequence single molecules avoid?
What technical problem would be eliminated?

2. Pacific Bioscience seq. strategy – single-molecule, real-time
immobilize DNA pol on glass
surface at bottom of very small
wells (~100nm radius); aim for
1 pol molecule/well
add DNA template + primer
use dNTPs with fluor attached
to terminal P so that it is cleaved off during incorporation
(how different from previous fluor-dNTPS we discussed?)
collect sequence in real time (enz can go ~1-100b/s)

3. How long should pulses last?
Why would you expect to see only 1 dye at a time?

4. What steps in previous methods would enzyme-
removal of fluor during synthesis eliminate?
How fast were previous methods (in terms of bases
sequenced/s) given need for chemical steps and
washes between base additions?

5. Challenge for detecting single fluor molecules is usually
not sensitivity, but reducing background
“Zero mode” waveguide (ZMW)– for well diameters << l
w/metallic walls, propagating waves blocked; evanescent
waves of exciting and emitted light decay exponentially.
Detection depth (volume) ~30nm (10-19 liters) for
100nm holes in aluminum. At mM dye conc., <1
dye/detection vol on average, and any such molecule
diffuses out of detection vol in ~100ms (verify: t ~x2/2D,
D= 10-12m2/s) whereas dye on dNTP being incorporated
by DNA pol expected to be retained by DNA pol. for ms.

6. Science 299:683, 2003
E-beam lithography makes
array of <100nm diam. holes in
~ 100nm thick aluminum film
on silica slide
For well diameter << l, I(z) ~ e-kz ;
excitation of fluor also inhibited
by wall proximity; effective
illuminated height theoretically
~ 30nm (~ 10x smaller than TIRF)
vol ~ liters

7. optical set-up to excite and read fluorescence from each well
holographic wave plate
divides input laser beams
into array of beamlets, 1/well
? -> more light/well than
If whole field illuminated
prism diffracts emitted light
to collect diff dye-dNTP
signals in different pixels

8. http://www. sciencemag
93 rows (1mm spacing) x 33 columns (4mm spacing)
Light from each well diffracted laterally for diff. detectors

9. Why do some wells stay illuminated?

10. Why do you want high (~mM) dNTP conc.?
Binding rate = kon [dNTP]
Rate (#/s) at which dNTPs bind each polymerase molecule,
per M conc of dNTP; if kon = 107/Ms, what conc.
of dNTP do you need for pol to synthesize 10b/s?
If you use TIRF, min. vol. of illuminated spot ~ pl2*hatt
~p (500nm)2 100nm = 10-19m3 = 10-16liter
How many dNTPs in this vol. at 1mM? Need <1
ZMW gets you to <1 by reducing illum. vol ~1000-fold!

11. Idea hinges on using dNTPs with dye labels on phosphate
(diff colors for each nt)
so dye will be cleaved off
during incorporation (don’t
need separate chemical step)
-> real-time sequencing
Base-labeled nucleotide
Phosphate-labeled nucleotide

12. Actual dNTPs used have big dyes + 6 phosphates –
They “invented” these dye-NTs and then had to engineer
(mutate) DNA pol to use them efficiently

13. C A T G
Emission spectra - would be easiest to distinguish C from
G, harder to distinguish 1 from 2, and 3 from 4
(affects “substitution” error rates)

14. Start with “simple” ss-template – 150 bases with
alternating regions with multiple G’s or C’s, synthesize
using dGTP-yellow, dCTP-blue, other bases unlabeled

15. Segment from previous trace
Same data in graphical form
They at least can pick out
the G’s and C’s separated by
0-2 other bases in this region
of template, but note variability
in peak duration

16. Now try a circular 75b ss template
Their pol enzyme has “strand displacing” activity
What advantages might a circular template have?
Should you see a pattern repeating every 75 b?
What disadvantages would there be to having to use
circular DNA templates?

17. Now try all 4 bases on 150b ss template
Note variability in pulse widths

18. Error rates
Single run on 150 b template ~30% of reads incomplete
12 insertions, 8 deletions, 7 mismatches (~20%)
What causes apparent insertions?
what if base sticks to pol long enough to be detected
but falls off before being incorporated, then
sticks again and gets incorporated?
what could you do about this? – try to engineer
enzyme that binds bases tighter (might not work)

19. What causes apparent deletions?
what if base were photobleached before detection?
what if enzyme incorporates a base very quickly?
If enzyme has constant average rate of incorporation of
bound base/unit time, expect Poisson distribution of
“hold times”, with largest number of shortest durations

20. Expected Poisson distribution of hold times
They have problem detecting the shortest hold times
because of photon counting statistics and noise:
dyes produce ~5000 photons/sec -> 50/10ms

21. What could you do about this problem?
try to slow down incorporation rate chemically
(they say pH change has a small effect)
try to engineer enzyme that incorporates bases
more slowly (might not work)

22. Substitution errors = misreading dyes due to incomplete
spectral separation, esp. hard distinguish in short pulses
what could you do? – try to improve dyes
More immediate solution to high error rates
read each DNA template multiple times
to generate a consensus sequence
(circular templates would be useful…)

23. They used sequence info from 449 reads of same
150 base template in different wells. Generate a
consensus sequence based on random samples of data
in which ? each position appears in >15 sequences.
Repeat 100x to generate 100 consensus sequences.
Error rate in consensus sequences ~2-3%.
If they had 3000 wells,
why did they only use
449 reads? Suggests they
are getting fewer useful
wells than they say…

24. Over-sequencing makes sense if the errors are random
but what if the error rate depends on sequence?
Summary
Many technical hurdles have been overcome, but error
rate remains very high
Even if they got reliable sequence in real time at rate
of 3b/s in each of 3000 wells, would need > 15
days to do human genome 15-fold redundantly,
which is competitors’ claimed current rate; 2 yrs
ago they said a 1,000,000-well device was coming…

25. Nanopore sensor method – Bayley group

26. a-hemolysin: heptameric membrane pore-forming protein
(bacterial toxin that punches holes in red blood cells)
Spontaneously forms 7-mer and inserts into lipid membrane

27. When inserted in membrane, in electrolyte solution,
creates channel that allows ions to cross membrane
Can easily detect single channels in artificial membrane
as they cause step-like changes in current …how many
ions go thru pore/sec? pA
-20
How much current do you expect from 0.7nm radius pore
10nm long in 1M KCl at 100mV, if conductivity s = 14S/m?
I = V/R = VG = VsA/L = .1*14*p*(.7*10-9)2/10-8 = 200pA

28. How can you make membranes, introduce pores?
1 cm _ +
2 nm
Cl- K+
Add lipid to
solution, raise
and lower
meniscus over
hole, 5nm lipid
bilayer forms
spontaneously (!)
teflon
barrier
with
~50mm
hole
Add a-hemolysin protein to 1 chamber – it inserts itself!

29. b-cyclodextrin: heptameric ring of sugars
spontaneously inserts inside a-hemolysin pore
stabilized by coordination with 7 identical sites,
one in each a-hemolysin monomer

30. b-CD insertion lowers conductivity
Why might you want
to reduce pore
diameter?
Would you expect
charges in pore to
influence current?
As dNMPs go through pore, they further decrease
conductivity; can b-CD be modified so that different
bases will -> different decreases in conductivity?

31. Bayley’s group have made extensive mutations
in a-HL and tested many b-CD derivatives to try to make
pores that distinguish DNA bases by extent of decr. cond.
Here, devise covalent S-S linkage between a-HL and b-CD
based on single cys in a-HL and S in modified b-CD
in order to have stable small diam. pore
How do they get single cys in heptameric
a-HL? Mix 2 types of a-HL, 1 with 1 cys and
tail of 8-charged (asp) aa’s; other w/no cys
or tail; they form different hetero-7mers;
select desired 7mer by electrophoresis

32. b-CD with -SH group stably associates w/cys-modified a-HL
b-CD without S reversibly enters a-HL; with S, it inserts
stably but can be removed by reducing -S-S- with DTT

33. aHL with stably inserted b-CD senses mixture of bases
Residual pore currents come in 4 types – why?

34. More data, with higher conc bases
note variable dwell times
note minimal overlap
in residual pore current
histograms – good for
distinguishing bases but
requires extensive data
smoothing

35. Scatter plot of dwell time vs residual pore current
Dwell times have wide distribution, but averages
differ for different bases. What does this suggest?

36. Channel can also distinguish methyl-dCMP, a variant of
C associated with silenced gene expression: this system
can detect such “epigenetic” changes more easily than
other sequencing methods

37. Idea for sequencing – use exonuclease to degrade
template to dNMPs and read them going thru pore
in the order they are produced

38. Problem #1: exonuclease doesn’t work in high salt, which
is required
for good
base discrim-
ination; they
lower salt on
side w/exo
Test mixed salt system on simpler templates with
No A’s
shows they can
degrade temp-
lates with exo’ase
and read bases
produced
No T’s

39. Technical challenges
Reducing [KCL]cis to 200mM allowed exonuclease to work,
but decreased ability to distinguish A’s and T’s to ~90%,
not good enough for sequencing; they might be able to
select exo’s that can work in high salt (e.g. brine bacteria)
<tdwell> ~10ms, but Poisson distribution => most dwells
are short; for short dwells it is harder to distinguish
different bases; very short dwells may -> “deletions”
Ability to distinguish 4 bases enhanced by +charge on
linker arm; may help to trap bases electrostatically;
? more chemical modifications might improve
discrimination

40. No data yet that exonuclease can be held near enough
to pore (e.g. via aHL-exo fusion protein) that
that chewed off bases can be read sequentially
Will path of some bases (drift + diffusion) be such
that they are read out-of-order, or not read at all?
Can pore formation be automated and multiplexed?
Nevertheless, extraordinary accomplishment in terms of
chemical adaptation of nanopore to make real-time,
label-free, single-molecule detector that
distinguishes 5 bases

41. Some similarities between solid state FETs and ion channels:
charge on channel walls regulates rate charged objects
(ions, DNA molecules, peptides) go thru pore
if pore is small enough compared to transported object,
changes in charge -> exponential changes in transport
chips – use simple geometries, Coulomb interactions,
engineered structures ~ mm x 30nm laterally,
a few nm vertically
biology – more complex geometries, Coulomb + chemical
interactions (H-bonding, covalent bonds), greater
control of nanoscale positioning via protein & DNA
engineering but less control at mm scale

42. Some big ideas from course:
Biology at nano-scale is not very different from
chemistry, physics at similar scale
Biological macromolecules (DNA, protein) allow
new forms of engineering and control
over nanoscale phenomena
Some tools are novel to biology – e.g. replication via pcr
Biology provides new things we want to sense
(like DNA) and new tools to sense them

43. At nano-level, single things become detectable
Detecting single-molecule events can provide info abt.
molecular structure not obtainable in bulk
measurements – e.g. when replicate
molecules cannot be kept in same state for
bulk measurements (e.g. phasing problem in
DNA sequencing)
Aim for detailed and quantitative understanding:
how many molecules per sec, unit area, vol;
what do they stick to; for how long; how is
signal generated; how many photons, etc
Think critically as you learn!

44. Suggestions for student presentations
Go over paper with me beforehand if at all possible!
Stick to a few main points – you have only ~20 minutes
Try to teach us something interesting you learned