What do sequence numbers in PDB files actually mean and why don't they match the sequence?

Question

I am studying the 3D structure of the LDH from x-ray crystallographic imaging I was pointed to from Is there any stereospecific enzyme in PDB that catalyzes an anabolic reaction and has an entry showing both reactant ligands?. It's 4ND4 and 4ND3 in the PDB and the paper about it is Cook, Senkovich, et al. Biochemical and structural characterization of Cryptosporidium parvum Lactate dehydrogenase. And I am trying to exactly link the detailed molfile model that I have extracted (and touched up) with the help of How do I obtain a 3D MOLFILE for the lactose synthase structure, especially the catalytic center?.

The key picture in the article is this:

When trying to exactly relate this to the actual sequence of the LDH enzyme I am finding puzzling issues. 4ND3 and 4ND4 refer themselves to UniProt Q5CYZ2. But the sequence is different. This raises a bunch of questions that perhaps I do not understand about PDB files in general, hoping someone could shed a light on this. From the PDB file https://files.rcsb.org/view/4ND3.pdb:

REMARK 999 SEQUENCE                                                             
REMARK 999 PEPTIDE SEQUENCE FOR THIS ENTRY WAS TRANSLATED FROM DNA SEQUENCING   
REMARK 999 OF ACTUAL CLONE USED FOR PROTEIN EXPRESSION. AUTHORS DO NOT KNOW IF  
REMARK 999 THE SEQUENCE DIFFERENCES ARE PCR ERRORS OR MUTATIONS CORRESPONDING   
REMARK 999 DNA POLYMORPHISM OR THERE WERE ERRORS IN PROTEIN DATABASE.           
DBREF  4ND3 A   17   337  UNP    Q5CYZ2   Q5CYZ2_CRYPI    17    337             
DBREF  4ND3 B   17   337  UNP    Q5CYZ2   Q5CYZ2_CRYPI    17    337             
SEQADV 4ND3 ALA A  202  UNP  Q5CYZ2    VAL   200 SEE REMARK 999                 
SEQADV 4ND3 ALA B  202  UNP  Q5CYZ2    VAL   200 SEE REMARK 999                 
SEQRES   1 A  321  MET ILE GLU ARG ARG LYS ILE ALA VAL ILE GLY SER GLY          
SEQRES   2 A  321  GLN ILE GLY GLY ASN ILE ALA TYR ILE VAL GLY LYS ASP          
SEQRES   3 A  321  ASN LEU ALA ASP VAL VAL LEU PHE ASP ILE ALA GLU GLY          
SEQRES   4 A  321  ILE PRO GLN GLY LYS ALA LEU ASP ILE THR HIS SER MET  

So it references Q5CYZ2, and the sequence there is:

        10         20         30         40         50
SEFLFSKHLF IFFILKMIER RKIAVIGSGQ IGGNIAYIVG KDNLADVVLF 
        60         70         80         90        100
DIAEGIPQGK ALDITHSMVM FGSTSKVIGT NDYADISGSD VVIITASIPG 
       110        120        130        140        150
RPKDDRSELL FGNARILDSV AEGVKKYCPN AFVICITNPL DVMVSHFQKV 
       160        170        180        190        200
SGLPHNKVCG MAGVLDSSRF RTFIAQHFGV NASDVSANVI GGHGDGMVPV 
       210        220        230        240        250
TSSVSVGGVP LSSFIKQGLI TQEQIDEIVC HTRIAWKEVA DNLKTGTAYF 
       260        270        280        290        300
APAAAAVKMA EAYLKDKKAV VPCSAFCSNH YGVKGIYMGV PTIIGKNGVE 
       310        320        330 
DILELDLTPL EQKLLGESIN EVNTISKVLD NAPAAGA

and what I'm finding is none of these sequence numbers match exactly starting from I32. You can see in the UniProt sequence there is I at 31, not 32. But it's not a constant offset either, D53 is at 51. And what about the prefix SEFLFSKHLFIFFILK before the MIERRKIAVI... in the PDB file and associated FASTA sequence?

The answer seems to lie in the ATOM list, which has in the 6th column a sequence number, that starts with 17, and that means it skips right over the SEFLFSKHLFIFFILK to start with M. But why? UniProt isn't identifying SEFLFSKHLFIFFILK as a signal peptide or anything. But then Q5CYZ2 is also not really vetted, but it also seems long to me to be a signal peptide.

SCALE3      0.000000  0.000000  0.005376        0.00000                         
ATOM      1  N   MET A  17      32.275  60.760  98.001  1.00 65.51           N  
ATOM      2  CA  MET A  17      30.819  60.627  98.290  1.00 65.33           C  
ATOM      3  C   MET A  17      30.216  59.589  97.341  1.00 62.08           C  
ATOM      4  O   MET A  17      30.381  59.683  96.125  1.00 64.10           O  
ATOM      5  CB  MET A  17      30.122  61.990  98.139  1.00 68.91           C  
ATOM      6  CG  MET A  17      28.988  62.249  99.129  1.00 72.14           C  
ATOM      7  SD  MET A  17      27.333  61.845  98.525  1.00 76.22           S  
ATOM      8  CE  MET A  17      26.946  63.315  97.574  1.00 75.10           C  
ATOM      9  N   ILE A  18      29.536  58.595  97.907  1.00 56.99           N  

now, it gets even weirder, in this list there are gaps, like here it skips over 21 right from 20 to 22. And the same phenomenon happens at the 47, 82, 104, and who knows later too.

ATOM     35  NH1 ARG A  20      18.415  54.478  92.556  1.00 41.01           N  
ATOM     36  NH2 ARG A  20      18.134  54.811  94.808  1.00 41.94           N  
ATOM     37  N   ARG A  22      25.730  54.160  90.284  1.00 34.87           N  
ATOM     38  CA  ARG A  22      26.279  53.881  88.963  1.00 33.99           C  
...
ATOM    212  O   ALA A  46      20.706  50.735  87.589  1.00 25.85           O  
ATOM    213  CB  ALA A  46      23.964  51.086  86.776  1.00 28.99           C  
ATOM    214  N   ASP A  48      21.675  52.759  87.757  1.00 29.47           N  
ATOM    215  CA  ASP A  48      20.536  53.481  87.203  1.00 30.89           C  

But why? Why would it say that A is at 46 and D at 48, when UniProt and their own FASTA sequence say that D follows A directly at that point? What is the meaning of this gap? Is it merely an arbitrary numbering in the PDB file? So that if I want to associate to a UniProt sequence, I would have to re-number and have to say I31 and D51 instead of I32 and D53? Is that it? Or are they claiming there are actually a missing amino-acid here which they haven't identified?

I would guess I'd have to re-number and there is no actual missing AA.

But now it gets even stranger:

ATOM    400  CE  MET A  72       4.872  44.082  88.838  1.00 33.49           C  
ATOM    401  N   PHE A  73A     11.326  42.924  91.678  1.00 31.09           N  
ATOM    402  CA  PHE A  73A     12.685  42.370  91.742  1.00 31.48           C  
ATOM    403  C   PHE A  73A     13.463  42.779  92.989  1.00 31.81           C  
ATOM    404  O   PHE A  73A     14.583  42.317  93.195  1.00 32.21           O  
ATOM    405  CB  PHE A  73A     13.489  42.810  90.520  1.00 30.78           C  
ATOM    406  CG  PHE A  73A     12.939  42.321  89.212  1.00 30.96           C  
ATOM    407  CD1 PHE A  73A     12.580  40.990  89.041  1.00 30.80           C  
ATOM    408  CD2 PHE A  73A     12.831  43.184  88.130  1.00 31.13           C  
ATOM    409  CE1 PHE A  73A     12.090  40.542  87.827  1.00 31.23           C  
ATOM    410  CE2 PHE A  73A     12.347  42.740  86.913  1.00 30.95           C  
ATOM    411  CZ  PHE A  73A     11.978  41.416  86.761  1.00 31.09           C  
ATOM    412  N   GLY A  73B     12.888  43.658  93.806  1.00 32.52           N  
ATOM    413  CA  GLY A  73B     13.569  44.166  94.988  1.00 31.46           C  
ATOM    414  C   GLY A  73B     14.728  45.089  94.656  1.00 31.53           C  
ATOM    415  O   GLY A  73B     15.667  45.211  95.440  1.00 31.87           O  
ATOM    416  N   SER A  74      14.668  45.740  93.496  1.00 30.43           N  

suddenly there are two at position 73, F and G. What does that mean?

I am thinking that the truth is these numbers are all meaningless and I can just re-number them and in reality we have M, F and G, S in that sequence, no variance or alternative at position 73, the FASTA sequence has -VMFGST-.

So what is going on here? Numbers are arbitrary and if I want actual meaningful sequence numbers I just re-number. Or is there a meaning to these gaps and double assignments 73A, 73B? And even if there is no actual meaning, I'd sure like to know why they did it in this complicated way? I guess it is an artifact of the technology somehow?

Answer 1

As @Andrew mentions in the comments, there is a weird tradition for depositing enzyme structures where the authors adjust the residue numbers so that the active site residues have the same residue numbers as famous related enzymes. For instance for serine proteases, the catalytic triad would always have the same numbering. If you do that, you have to introduce gaps in the numbering or residue numbers like your 73A and 73B (for deletions and insertions with respect to the aligned reference sequence).

Here is the active site of a related thermophilus structure. Notice that the active site residues are R171 and H195 as well. If you look at the underlying sequences, the numbering might be different.

Source: https://hal.archives-ouvertes.fr/hal-03011827/document (googled "R171 N140 H195 LDH")

The reference sequence is that of Squalus acanthias, shifted by one (see Abad-Zapatero, C., Griffith, J.P., Sussman, J.L., Rossmann, M.G. (1987) J Mol Biol 198: 445-467). The historical PDB structures are 1LDM, 3LDH, and 6LHD (and obsolete 2LDH and 4LDH). There is also a historic structure of protein from pig (5LDH) from the same lab. These are the structures that led to the discovery of the Rossmann fold.

As for the mismatch of the Uniprot sequence and the PDB sequence at the N-terminus, the source of the Uniprot sequence is a shotgun DNA sequencing project, and they did not know where the translation start is. Usually, these protein sequences should start with a methionine.

Here is the sequence alignment with a different related sequence obtained with a different method which supports methionine as the first amino acid in the sequence. Notice the Iowa strain has two mutations, R->K and K->E near the end of the sequence.