17
$\begingroup$

When reading my textbook I noticed that in all examples but one from eight the recognition site was an even number of bases.

I wondered if this was just a co-incidence, so I took the data from this site for over a thousand known recognition sites and put it into a spreadsheet (XLS uploaded here). The results are probably best summarised graphically:

restriction enzymes pie chart

restriction enzymes bar chart

Stats Test: Chi Squared Goodness of Fit.

Null Hypothesis: 
There is no significant difference between the number of restriction sequences 
that are an odd or even length.

|---------|----------|----------|-----|-------|----------|
|  Trait  | Observed | Expected | O-E |(O-E)^2|(O-E)^2 /E|
|---------|----------|----------|-----|-------|----------|
|  Odd    |    172   |  231.5   |-59.5|3540.25|  15.293  |
|  Even   |    291   |  231.5   | 59.5|3540.25|  15.293  |
|---------|----------|----------|-----|-------|----------|

Chi Squared Value = 30.586

P=0.05, 1 Degree of Freedom: Critical Value of 3.841

H0 rejected with 95% confidence (indeed with 99.9%+ confidence)

Can anyone explain or suggest why it is more common that restriction enzymes recognition sites have an even number of bases?

Updates

  • Expanded dataset to include new recognition sites from the resource that 96well linked to.
  • Removed all duplicate recognition sites leaving 465 distinct recognition sequences (my fault for not removing them in the first instance)
  • Ran stats test on the data
  • See Previous Version
Improve this question
$\endgroup$
1
  • 7
    $\begingroup$ I congratulate for your reflex: from the textbook to the database search. Nicely done! Your data is from a sample of some 1100 restriction enzymes. Rebase.neb.com reports some 6800 restriction enzymes. Are your conclusions still valid? $\endgroup$
    – Gianpaolo R
    Feb 23, 2012 at 21:40

4 Answers 4

Reset to default
14
$\begingroup$

I think this is due to the over-representation of recognition sites with length 6:

data<-c(16, 16, 12, 12, 6, 6, 6, 6, 4, 16, 6, 6, 6, 6, 15, 15, 6, 6, 6, 6, 11, 11, 6, 6, 4, 4, 6, 6, 11, 12, 6, 6, 23, 23, 6, 6, 6, 6, 9, 12, 4, 4, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7, 10, 10, 6, 4, 6, 6, 11, 11, 9, 9, 6, 6, 6, 6, 5, 5, 8, 8, 6, 6, 8, 8, 6, 9, 10, 10, 6, 6, 6, 5, 5, 6, 4, 6, 6, 5, 5, 14, 14, 6, 6, 6, 16, 6, 6, 6, 6, 15, 15, 6, 6, 6, 6, 6, 18, 18, 7, 7, 11, 11, 20, 20, 6, 13, 4, 4, 6, 6, 6, 6, 6, 6, 6, 11, 6, 6, 7, 7, 6, 6, 6, 6, 6, 12, 6, 6, 10, 10, 23, 23, 7, 7, 23, 23, 12, 12, 6, 6, 6, 6, 10, 10, 6, 6, 8, 8, 6, 6, 35, 35, 11, 11, 7, 7, 6, 6, 9, 9, 8, 8, 16, 16, 6, 6, 17, 17, 6, 6, 6, 6, 23, 23, 6, 6, 4, 4, 21, 21, 12, 12, 20, 20, 6, 6, 6, 6, 6, 7, 7, 6, 6, 6, 6, 5, 5, 11, 11, 6, 11, 11, 6, 6, 5, 5, 7, 7, 11, 11, 11, 11, 6, 6, 12, 12, 6, 6, 6, 21, 21, 9, 9, 8, 8, 7, 7, 16, 16, 4, 4, 6, 6, 6, 6, 7, 7, 18, 18, 6, 6, 6, 6, 6, 23, 23, 7, 34, 34, 39, 39, 6, 6, 12, 12, 5, 5, 19, 19, 8, 8, 8, 8, 4, 6, 6, 6, 6, 6, 5, 5, 6, 6, 6, 6, 7, 7, 4, 4, 15, 15, 7, 7, 7, 7, 14, 14, 11, 11, 27, 27, 12, 12, 4, 4, 10, 10, 6, 6, 8, 8, 7, 7, 8, 8, 7, 7, 5, 5, 7, 7, 6, 7, 7, 6, 6, 8, 8, 39, 39, 6, 6, 12, 12, 8, 8, 7, 13, 13, 8, 8, 8, 8, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 5, 5, 7, 17, 17, 17, 17, 11, 11, 15, 15, 6, 6, 6, 5, 5, 5, 6, 6, 5, 7, 7, 12, 6, 12, 6, 6, 6, 5, 6, 6, 7, 2, 5, 11, 5, 6, 4, 5, 7, 5, 4, 4, 6, 4, 5, 6, 7, 12, 6, 7, 7, 7, 6, 4, 4, 7, 5, 6, 6, 6, 7, 5, 12, 13, 5, 6, 6, 6, 5, 11, 11, 5, 6, 10, 5, 5, 11, 6, 5, 5, 6, 5, 6, 5, 6, 6, 7, 7, 6, 5, 5, 7, 6, 5, 6, 5, 6, 5, 5, 7, 6, 6, 6, 3, 5)
h<-hist(data, breaks=0.5:40.5)
df<-data.frame(counts=h$counts, mids=h$mids)
df$even <- (df$mids%%2 == 0)
ggplot(df, aes(x=mids, y=counts, fill=even))+geom_bar(stat="identity")

Histogram of recognition site length

If you look at the histogram of lengths, there is no bias towards even recognition site lengths except for length 6.

Improve this answer
$\endgroup$
4
  • 1
    $\begingroup$ I would guess that the followup question is why is this the case? $\endgroup$
    – bobthejoe
    Feb 27, 2012 at 7:02
  • 1
    $\begingroup$ Interesting. Is there any motif in the length 6 sites? $\endgroup$
    – Gianpaolo R
    Feb 27, 2012 at 12:44
  • 1
    $\begingroup$ I think you've found the natural bias that exists for recognition sites. The distribution is symmetric for 4-8 bases, and obviously cannot get much smaller than 4 without losing a great deal of specificity and as a consequence, not being very useful. At the other extreme, the larger the stretch of DNA required, the larger the enzyme must be which is expensive, and while you gain specificity you sacrifice cutting efficiency for things like methylation and secondary structure. $\endgroup$
    – user560
    Apr 6, 2012 at 12:09
  • $\begingroup$ Surprising that there seems to be at least one restriction enzyme with a recognition sequence of length just 2! $\endgroup$
    – Vikki
    Apr 2, 2021 at 20:17
8
$\begingroup$

Not sure why Larry Parnell was down-voted, he was not technically wrong.

Crystal structures of the most popular restriction enzymes are already known and can easily be found the Protein Data Bank or Wikipedia for graphical reference. Any stretch of double-stranded DNA makes a complete 360 rotation (about it's helical axis) in 10-10.5 base pairs. A 180 degree rotation would then be half of that ~5 bases.

A palindrome is not strictly necessary for a restriction enzyme site, but exploiting them has its advantages. Firstly, synthesizing one subunit of the enzyme requires half the occupied space of the genome, therefore translation is more efficient (2 products from half the instruction set).

Some restriction enzymes take advantage of rotational symmetry, such as EcoRI and EcoRII (both odd and even palindromic enzymes). They recognize the same GAA (in the case of EcoRI) or NNGG (in the case of EcoRII), coming from each of the two DNA strands. For notation purposes we write the recognition site of the positive strand only (since the DNA code is symmetrical). Asymmetrical recognition sites are explained by enzymes binding to DNA as heterodimers.

New England Biolabs has a very good review of the kinds of restriction enzymes and the variations in their behaviour.

http://www.neb.com/nebecomm/tech_reference/restriction_enzymes/overview.asp#.T0wqHYcgerY

To sum up, palindromes are often, but not necessarily, comprised of even numbers of bases (domains bind to 2n bases). Many enzyme recognition sites are palindromic in the literal sense (CATTAC) or biological sense (GAATTC) because there is an axis of symmetry that is used. Palindromic sequences are more commonly used because enzymes are composed of homodimers.

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ Could you expand on "domains bind to 2n bases" - I'm right in thinking that that is the reason for a preference of even bases? Why is six so popular as mentioned in Biocs' answer? $\endgroup$
    – Rory M
    Feb 28, 2012 at 18:22
  • 1
    $\begingroup$ If one half of a dimer recognizes "n" bases, and most of these enzymes (to the best of our knowledge of crystal structure/molecular weight analyses) are dimers, then that means they must be recognizing "2n" bases, which biases the total length of recognition site to be an even number. From your very well collected dataset, there is an overwhelming amount of enzymes that recognize 5-7 bases, with the mode being 6. My inference based on crystal structures is that the recognition of ~3 bases is about a quarter turn of a double-stranded helix and is a conveniently recognized conformation. $\endgroup$
    – user560
    Feb 28, 2012 at 19:10
2
$\begingroup$

Most restriction enzymes recognize palindromes, hence the recognition site of an even number of residues. Why palindromes? This allows for a vast increase in complexity (or rarity) from a DNA sequence standpoint, while requiring very little added complexity from the protein. In other words, if a protein domain recognizes 2 or 3 or 4 bp of DNA, simply adding a second copy of that domain to the peptide chain (via duplication of the DNA encoding that domain) does the trick. It is much easier to duplicate a module than to build a new or second (perhaps complimentary) module.

Improve this answer
$\endgroup$
6
  • 3
    $\begingroup$ How does being even or odd affect a palindrome? What would be the probability difference between CATATG (6) and CATGATG (7) i.e. why are even palindromes more likely? $\endgroup$
    – Rory M
    Feb 23, 2012 at 20:31
  • 2
    $\begingroup$ -1 for justifying your answer by defining palindromes as sequences with even number of bases. EcoRII recognizes the palindrome CCWGG (W being A or T). $\endgroup$
    – Gergana Vandova
    Feb 23, 2012 at 20:39
  • 2
    $\begingroup$ @RoryM CAT TAC is not a palindrome. CAT ATG is a palindrome. $\endgroup$
    – Gergana Vandova
    Feb 23, 2012 at 20:43
  • $\begingroup$ @GerganaVandova sorry I was thinking of palindrome in the common sense rather than the DNA sense, I'll make the edit! $\endgroup$
    – Rory M
    Feb 23, 2012 at 20:46
  • 1
    $\begingroup$ The misuse of the word palindrome for these sequences is a real pet hate of mine - they're rotationally symmetrical not palindromic! $\endgroup$
    – Jack Aidley
    Jan 23, 2013 at 13:45
-1
$\begingroup$

Are odd numbered overhangs actually palindromes? Somebody gave an example saying the following is a palindrome:

5'-CATGATG-3'

However, the opposite strand will have a C in the center position, that is, it will have the sequence:

5'-CATCATG-3'

In this example, the two center-most bases (G and C) are complimentary, but the two sequences are not strictly palindromic. This should be the case whenever there is an odd number of bases in the overhang.
Another example, a 3-base overhang.

5'-GTC-3'

The complement will be:

5'-GAC-3'

Which is not a palindrome of the upper sequence. This means that half of the time overhangs with odd number of bases cut by the same restriction enzyme from random fragments will not be complimentary (the 5'-GTC-3' overhangs can't anneal to other 5'-GTC-3' overhangs, only to the 5'-GAC-3' ones), and unless careful attention is paid to their sequences, odd-numbered overhangs in cloning projects can cause problems. Even numbered overhang example:

5'-AATT-3'

In this case the opposite strand will have an identical overhang:

5'-AATT-3'

and will always have homology with other 5'-AATT-3' overhangs. I suspect that the reason there are more even numbered restriction enzyme overhangs is to avoid this problem.

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ This is possibly better suited to be a follow-up question than an answer. $\endgroup$
    – Maximilian Press
    Jun 16, 2021 at 20:26
  • $\begingroup$ @MaximilinPress the last sentence does seem to address the problem, somewhat. To OP: I have added the 'add citation' banner, because the answer seems more like a long comment, or an educated guess at best. It's best to pen up science-based answers, including references to credible sources. $\endgroup$
    – AliceD ♦
    Jun 17, 2021 at 6:59

Not the answer you're looking for? Browse other questions tagged or ask your own question.