9.2 Using DNA-Based Methods to Understand Protein Function

DNA libraries are organised collections of genetic information

A DNA library is a collection in which many DNA fragments are each preserved and recoverable
A genomic library represents the whole genome through fragments and therefore includes introns and regulatory regions
A cDNA library is made from mature mRNA, lacks introns and represents only expressed transcripts
Transcripts are not equally abundant, so sample conditions, RNA extraction and reverse transcription all bias the content
The chance of finding a sequence depends on its relative frequency and on how many clones are examined
With probability p per clone and n clones examined, detection is 1 minus (1 minus p) to the n
Example: p of 0.05 with n of 40 gives about 87.1 percent, leaving 12.9 percent undetected
A CRISPR guide library is a tool set for functional screening rather than a list of sequences

표 9.2-1Genomic and cDNA libraries

Figure 9.2-23How evidence accumulates across the stages of protein function research

Figure 9.2-24Genomic DNA and cDNA libraries compared

Figure 9.2-25Teaching model of the probability of detecting a transcript

Sequence and structural relationships inform protein function

Aligning an unknown sequence with studied proteins reveals conserved regions and variable positions
Sequence identity is the fraction of identical residues over a stated aligned region
Example: nine identical out of twelve gives 75 percent for that region only
An identity figure alone cannot establish whole-protein identity or equal activity
A domain folds independently or performs a defined function and is a strong clue to role
Known DNA-binding, ATP-binding or catalytic domains allow the role of an unknown protein to be inferred
Similar structures can still differ in substrate specificity or regulation
High local confidence in a predicted structure does not replace experimental proof of function

Figure 9.2-26Alignment of two short amino acid sequences and calculation of identity

Figure 9.2-27Forming a functional hypothesis by comparing domain architecture

Figure 9.2-28Levels of evidence from sequence and structure prediction to demonstrated function

Fusion proteins and immunofluorescence reveal localisation

Joining the coding sequence of a gene in frame with GFP lets the fluorescence report the location of the protein
Whether to tag the N or C terminus needs review because either can affect structure or function
Immunofluorescence uses antibodies, with indirect methods employing a primary antibody and a labelled secondary
Antibody specificity, secondary-only controls and fixation and permeabilisation must be validated
Colocalisation means the proteins share a compartment or lie closer than the resolution limit
It is not evidence of direct binding, which needs Co-IP, FRET or binding assays with purified protein
Overlap of GFP signal with nuclear staining supports a nuclear localisation

표 9.2-2Evidence from interaction and localisation methods

Figure 9.2-29GFP fusion protein compared with indirect immunofluorescence

Figure 9.2-30Simulated fluorescence image showing protein localisation in a cell

Figure 9.2-31Distinguishing correlated fluorescence intensity from direct binding

Protein interactions reveal how molecular functions connect

Yeast two-hybrid uses the principle that a bait fused to a DNA-binding domain and a prey fused to an activation domain bring transcription of a reporter when they interact
It preserves interaction in a cell but autoactivation and false positives must be excluded
Co-immunoprecipitation asks whether protein B is recovered when A is precipitated with an antibody
It raises the likelihood of a shared complex, though bridging proteins or nucleic acids can also cause recovery
A band in the IgG control changes the interpretation
FRET is energy transfer sensitive to the proximity and orientation of donor and acceptor
A simple model gives E = 1 over (1 plus (r/R0) to the sixth), where R0 is the half-efficiency distance
Example: with R0 of 5.4 nm, r of 5.4 nm gives E of 0.5 and r of 10 nm gives about 2.4 percent
Methods with different evidence levels should be combined to strengthen the conclusion

Figure 9.2-32Principle of the yeast two-hybrid assay

Figure 9.2-33Simulated co-immunoprecipitation and immunoblot bands

Figure 9.2-34Distance dependence of FRET efficiency with R0 assumed to be 5.4 nm

Figure 9.2-35Representative protein interaction methods compared

DNA microarrays report RNA expression patterns

A microarray immobilises known probes on a chip and measures hybridisation of labelled nucleic acid
Transcript analysis reverse-transcribes and labels RNA, sometimes with an amplification step
Signal intensities are compared only after background correction and within- and between-experiment normalisation
The fold change is T over C and the log change is log2 of that ratio
A log2 change of +1 is a doubling, -1 is a 50 percent fall and -2 means 25 percent remains
Example: 125 over 500 is 0.25, so log2 equals -2, a 75 percent fall
Lower transcript does not equal lower protein amount or activity
Many genes produce large apparent changes by chance, so biological replication and multiple-testing correction are required

표 9.2-3Calculating expression change from a microarray

Figure 9.2-36Workflow of a two-colour DNA microarray experiment

Figure 9.2-37Simulated two-colour spot image with relative signals

Figure 9.2-38Example calculation of log2 expression change per gene

Figure 9.2-39Simulated gene expression heat map

CRISPR perturbation tests protein function

With SpCas9 a guide RNA recognises the target and the enzyme cuts DNA next to a suitable PAM
Insertions and deletions during non-homologous end joining can shift the reading frame and abolish function
Knockout cuts to destroy function, interference uses dCas9 with a repressor to lower expression and activation raises it
Observed changes can also come from off-target effects, cell line differences or delivery
Non-targeting guide controls and several independent target guides should be included
Rescue by reintroducing the target gene strengthens causal inference
Example: control 99.67 percent, knockout 44.0 percent and rescue 93.0 percent means a fall of about 55.9 percent
Recovery in the rescue group supports dependence on the target

Figure 9.2-40SpCas9 target cleavage and the range of repair outcomes

Figure 9.2-41CRISPR knockout, interference and activation compared

Figure 9.2-42Simulated measurements from a CRISPR perturbation and rescue experiment

Figure 9.2-43Control experiments required to validate a CRISPR phenotype

Integrated study: from sequence to biological function

Alignment suggests domain similarity, fluorescence gives localisation, Co-IP suggests partners and perturbation tests causation
Each method supports a different proposition, so no single result fixes the whole function
Consistent evidence raises confidence while conflicting results call first for a check of conditions
The final judgement places sequence, localisation, interaction, perturbation and rescue evidence together

Figure 9.2-44Example of integrating evidence from sequence to gene perturbation

References

**표 9.2-1**Genomic and cDNA libraries
Feature	Genomic DNA library	cDNA library
Starting material	Fragments of genomic DNA	Reverse transcripts of mature mRNA
Introns	Present	Generally absent
Representation	Represents the whole genome	Represents expressed transcripts only
Main use	Finding genes and regulatory regions	Finding expressed genes and producing protein

**Figure 9.2-23**How evidence accumulates across the stages of protein function research

**Figure 9.2-24**Genomic DNA and cDNA libraries compared

**Figure 9.2-25**Teaching model of the probability of detecting a transcript

**Figure 9.2-26**Alignment of two short amino acid sequences and calculation of identity

**Figure 9.2-27**Forming a functional hypothesis by comparing domain architecture

**Figure 9.2-28**Levels of evidence from sequence and structure prediction to demonstrated function

**표 9.2-2**Evidence from interaction and localisation methods
Method	Signal measured	Required validation
Fluorescent fusion	Distribution of fluorescence in the cell	Confirm the tag does not alter function or location
Immunofluorescence	Antibody staining pattern	Antibody specificity and secondary controls
Colocalisation	Spatial overlap of two channels	Resolution limits; not direct binding
Yeast two-hybrid	Transcription of a reporter	Exclude autoactivation and false positives
Co-immunoprecipitation	Proteins recovered together	IgG control and possible nucleic acid bridging
FRET	Energy transfer between donor and acceptor	Control distance, orientation and labelling ratio

**Figure 9.2-29**GFP fusion protein compared with indirect immunofluorescence

**Figure 9.2-30**Simulated fluorescence image showing protein localisation in a cell

**Figure 9.2-31**Distinguishing correlated fluorescence intensity from direct binding

**Figure 9.2-32**Principle of the yeast two-hybrid assay

**Figure 9.2-33**Simulated co-immunoprecipitation and immunoblot bands

**Figure 9.2-34**Distance dependence of FRET efficiency with R0 assumed to be 5.4 nm

**Figure 9.2-35**Representative protein interaction methods compared

**표 9.2-3**Calculating expression change from a microarray
Gene	Control signal C	Treated signal T	FC = T/C	log2 FC
G1	100	400	4.00	+2.00
G2	250	250	1.00	0.00
G3	500	125	0.25	-2.00
G4	500	125	0.25	-2.00

**Figure 9.2-36**Workflow of a two-colour DNA microarray experiment

**Figure 9.2-37**Simulated two-colour spot image with relative signals

**Figure 9.2-38**Example calculation of log2 expression change per gene

**Figure 9.2-39**Simulated gene expression heat map

**Figure 9.2-40**SpCas9 target cleavage and the range of repair outcomes

**Figure 9.2-41**CRISPR knockout, interference and activation compared

**Figure 9.2-42**Simulated measurements from a CRISPR perturbation and rescue experiment

**Figure 9.2-43**Control experiments required to validate a CRISPR phenotype

**Figure 9.2-44**Example of integrating evidence from sequence to gene perturbation